Detecting presence of tumors based on methylation status of cell-free nucleic acid molecules
The methylation state in the nucleotide sequence is analyzed through computer systems and machine learning algorithms, combined with MBD protein and restriction enzyme technology, the complexity of cell-free nucleic acid detection is solved and high sensitivity detection is achieved for cancer.
Patent Information
- Application Number
- CN202380032966.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-04-29
- Filing Date
- 2023-04-07
- Publication Date
- 2024-11-15
AI Technical Summary
The prior art is difficult to effectively detect cell-free nucleic acids in cancer, especially in bodily fluids, due to low nucleic acid amounts and heterogeneous forms, resulting in complex detection, and lack of high sensitivity systems and methods.
Using computer systems and machine learning algorithms, the tumor-derived DNA is determined by analyzing the methylation status in the nucleotide sequence, and the model is generated using training data to indicate the presence of cancer based on the methylation level, combining MBD protein and restriction penetration. Enzyme technology performs segmentation and analysis of sequencing reads.
It improves the sensitivity to cancer detection, can accurately identify the presence of cancer in cell-free nucleic acids, reduces the false positive rate, and achieves early detection of various cancer types.
Smart Images

Figure CN118974279A_ABST
Abstract
Description
[0001] CROSS-REFERENCE TO RELATED APPLICATIONS
[0002] This application claims the benefit of U.S. Provisional Application No. 63 / 328,602, filed on April 7, 2022, and U.S. Provisional Application No. 63 / 336,852, filed on April 29, 2022. Both are incorporated by reference in their entirety for all purposes.
[0003] background
[0004] Cancer is the leading cause of disease worldwide. Every year, tens of millions of people around the world are diagnosed with cancer, and more than half of them eventually die from it. In many countries, cancer ranks as the second most common cause of death after cardiovascular disease. Early detection is associated with improved outcomes for many cancers.
[0005] Cancer can be caused by the accumulation of genetic variations within an individual's normal cells, at least some of which result in improperly regulated cell division. Such variations typically include copy number variations (CNVs), single nucleotide variations (SNVs), gene fusions, insertions and / or deletions (indels), epigenetic variations including 5-methylation of cytosine (5-methylcytosine) and association of DNA with chromatin and transcription factors.
[0006] Cancer is usually detected by tumor biopsy, followed by analysis of cells, markers or DNA extracted from cells. However, it has recently been proposed that cancer can also be detected based on cell-free nucleic acid in body fluids such as blood or urine. Such tests have the advantage that they are non-invasive and can be performed without the need to identify suspicious cancer cells in a biopsy. However, due to the very low amount of nucleic acids in body fluids and the fact that the nucleic acids present are heterogeneous in form (e.g., RNA and DNA, single-stranded and double-stranded, and post-replication modifications and various states associated with proteins such as histones), such tests are complex.
[0007] Therefore, there is a need for improved systems and methods for improving cancer detection using liquid biopsy assays.It is therefore an object of the present disclosure to provide computer-implemented systems and methods having an improved ability to classify a sample as containing tumor-derived DNA with increased sensitivity. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] The accompanying drawings, which are incorporated into and constitute a part of this specification, illustrate certain embodiments and are used together with the written description to explain certain principles of the methods, computer-readable media, and systems disclosed herein. The description provided herein can be better understood when read in conjunction with the accompanying drawings, which are included by way of example and not by way of limitation. It should be understood that, unless the context otherwise indicates, the same reference numerals represent the same components in all the drawings. It should also be understood that some or all of the drawings may be schematic representations for illustrative purposes and do not necessarily depict the actual relative sizes or positions of the elements shown.
[0010] Figure 1 is a schematic diagram of an example environment 100 for identifying a nucleic acid corresponding to a classification region of a reference sequence, wherein the classification region has at least a threshold number of CpGs.
[0011] Figure 2 is a schematic diagram of an example architecture for determining a tumor metric based on analyzing one or more models of methylation status of cell-free nucleic acid molecules, according to one or more embodiments.
[0012] Figure 3 is a schematic diagram of an example architecture for training one or more machine learning models to determine a cancer metric based on the methylation state of cell-free nucleic acid molecules, according to one or more embodiments.
[0013] Figure 4 is a flow chart of an example process for determining a tumor metric associated with a methylation level of a classification region of a reference sequence, according to one or more embodiments.
[0014] Figure 5 is a block diagram illustrating components of a machine in the form of a computer system that may read and execute instructions from one or more machine-readable media to perform any one or more of the methodologies described herein, according to one or more example embodiments.
[0015] Figure 6 is a block diagram illustrating a representative software architecture that may be used in conjunction with one or more hardware architectures described herein, according to one or more example implementations.
[0016] Fig. 7A , Figure 7B and Figure 7C is a graphical representation showing promoter methylation calls in training and testing samples for 88 TSG+HRD genes.
[0017] Fig. 8A is a graphical representation showing the cancer prediction scores in cancer-free samples with >1 call. Figure 8B is a table showing genes most frequently called for promoter methylation in cancer-free donors. Figure 8C is a table showing in silico LoD estimates for selected genes from cell line KM12.
[0018] Fig. 9 is a graphical representation showing the prevalence of promoter methylation in different types of cancer, our test dataset (total N=559) and TCGA public data (total N=2,380) for TSG and HRD genes. Limited to genes with promoter methylation in TCGA.
[0019] Fig. 10A is a graphical representation showing MLH1 promoter methylation in cancer-free donors and CRC patients (MSI-H and MSS). Fig. 10B is a table showing MLH1 promoter methylation and calling of BRAF-V600E in CRC patients.
[0020] Fig.11 is a table showing an overview of the training and test data sets of Example 5.
[0021] Fig. 12A is a graphical representation showing the performance of the model for predicting CRC / cancer-free status in the training set. The shading indicates changes in iterations. Fig. 12B is a table showing the performance of cancer prediction models on an independent test dataset.
[0022] Fig.13A is a table showing the CV of TF estimation from genome calling and methylation in the in vitro dataset. Fig. 13B is a graphical representation showing the TF model performance (black line represents the diagonal) in the training set of CRC and cancer-free samples (cross validation). Fig. 13C is a graphical representation of a computer simulated data set showing a lower true TF.
[0023] Fig.14 is a graphical representation showing the distribution of predicted TFs for CRC patients in the training set with and without driver mutations.
[0024] Fig.15A is a graphical representation showing the individual positivity rate for lung cancer detection in stage I / II patients and stage III / IV patients.
[0025] Fig. 15B is a graphical representation showing the individual positivity rates for various cancer tests (bladder, gastric, ovarian, pancreatic, and liver) in stage I / II patients and stage III / IV patients.
[0026] Fig.16is a graphical representation showing the individual positivity rates for various cancer tests (bladder cancer, gastric cancer, ovarian cancer, pancreatic cancer, and liver cancer) in stage I patients, stage II patients, stage III patients, and stage IV patients.
[0027] Fig.17 is a graphical representation of epigenomic MAF relative to target MAF for colorectal, lung, and breast cancer.
[0028] Fig.18 is a table showing that the quantitative accuracy of epigenomic cTFs can reach LoQ less than 0.1% in CRC, lung and breast clinical samples.
[0029] Fig.19A is a graphical representation showing that somatic mutation-based cTFs are robust to duplications within the same cTF level, particularly at cTF levels of 0.5% or higher.
[0030] Fig.19B is a graphical representation showing that epigenomic cTFs can maintain 100% evaluation rate and have LoQ as low as 0.1% cTF.
[0031] Fig. 20A is a graphical representation of the methylation signals and somatic mutations of the first repeats titrated clinically.
[0032] Fig. 20B is a graphical representation of the methylation signal and somatic mutations of the clinically titrated second repeat.
[0033] Fig.21 is a table indicating the changes in ctDNA levels of the first and second replicates calculated using the genomic-only method and the methylation method.
[0034] Fig. 22 is a graphical representation of epigenomic cTFs versus genomic cTFs on clinical samples (one dot represents one sample).
[0035] Fig.23 is a graphical representation of the distribution of epiMAF in early and advanced cancer patients for breast, colorectal, lung, and a group of other cancers.
[0036] Fig.24 is a graphical representation showing probability distributions indicating the numbers of methylated cytosines included in the three partitions.
[0037] Fig.25A Included is a graphical representation showing a change in a metric for a first classification region for a first set of samples treated with MBD using a first set of reagents and a second set of samples treated with MBD using a second set of reagents.
[0038] Fig.25BIncluded is a graphical representation showing a change in a metric for a second classification region for a first set of samples treated with MBD using a first set of reagents and a second set of samples treated with MBD using a second set of reagents.
[0039] Overview
[0040] In one or more aspects, a method includes obtaining training sequence data by a computing system having one or more hardware processors and memory, the training sequence data including training sequencing reads from more than one sample derived from more than one subject, a single training sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in a sample in the more than one sample, and the single training sequencing read corresponds to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The method also includes analyzing the training sequencing reads by the computing system to determine a first quantitative measure corresponding to a single classification region in more than one classification region derived from the training sequencing reads, at least a portion of the single classification region in more than one classification region corresponding to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least the threshold cytosine-guanine content in a subject in which cancer was detected. The method also includes analyzing the training sequencing reads by a computing system to determine a second quantitative measure corresponding to more than one control region derived from the training sequencing reads, wherein a single control region in the more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and having at least a threshold amount of methylated cytosine in the subject in which the cancer was detected and another subject in which the cancer was not detected. The method also includes determining, by the computing system, a measure of a single classification region in more than one classification region based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region. The method also includes generating training data by a computing device, the training data including measures of single classification regions in more than one classification region of the training sequence reads of the sample of the training subject. The method also includes implementing one or more machine learning algorithms by the computing system and using the training data to generate a model to determine an indication of the presence of cancer in the subject based on the amount of methylated cytosine in at least a portion of the more than one classification region, the model including weights of single classification regions in more than one classification region, and at least a portion of the weights of the single classification regions are different from each other.
[0041] In one or more aspects, the method includes obtaining, by a computing system, test sequence data from an additional subject that is not included in the more than one subject, the test sequence data including test sequencing reads from a sample derived from the additional subject, a single test sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in the additional sample, and the single test sequencing read corresponding to a molecule having at least a threshold amount of methylated cytosine in a region of the nucleotide sequence having at least a threshold cytosine-guanine content; and determining an indication of the presence of cancer in the additional subject using the model and the additional sequence data.
[0042] In one or more aspects, the method includes analyzing, by a computing system, a test sequencing read to determine a first quantitative measure corresponding to a single classification region among more than one classification region derived from the test sequencing read; analyzing, by the computing system, a test sequencing read to determine a second quantitative measure corresponding to a single control region among more than one control region derived from the test sequencing read; determining, by the computing system, a metric for the single classification region based on the first quantitative measure for the single classification region and the second quantitative measure for the more than one control region; and generating, by the computing system, an input vector comprising the metric for the single classification region, wherein the model uses the input vector to determine an indication of the presence of cancer in another subject.
[0043] In one or more aspects, the one or more machine learning algorithms comprise one or more classification algorithms, and the indication of the presence of cancer corresponds to a probability of the presence of cancer in the additional subject.
[0044] In one or more aspects, the one or more machine learning algorithms include one or more regression algorithms, and the indicator corresponds to an estimate of the tumor fraction of the additional sample.
[0045] In one or more aspects, the training sequencing reads constitute a first portion of training sequence data, and the additional training sequencing reads constitute a second portion of training sequence data, wherein the additional training sequencing reads are different from the training sequencing reads, and the method includes analyzing, by a computing system, at least one of the first portion of the training sequence data or the second portion of the training sequence data to determine a single frequency of more than one variant present in a single sample of more than one samples; determining, by the computing system and for the single sample, a variant having a maximum frequency among the more than one variants, the maximum frequency corresponding to a single frequency having a maximum value among the single frequencies derived from the single sample; and determining, by the computing system, a single measure of the tumor score for the single sample based on the maximum value of the single frequencies derived from the single sample.
[0046] In one or more aspects, the training data includes a single measure of tumor fraction for a single sample of the more than one samples, and the model is generated based on the single measure of tumor fraction for the single sample of the more than one samples.
[0047] In one or more aspects, a metric for a single classification region is determined based on a scaling factor and an error correction factor.
[0048] In one or more aspects, more than one classification region individually corresponds to a genomic region where the methylation rate of the genomic region in nucleic acid derived from cells obtained from a subject in which cancer is present is different from the methylation rate of the genomic region in nucleic acid derived from cells obtained from a subject in which cancer is not present.
[0049] In one or more aspects, the more than one classification regions correspond to first more than one classification regions for a first cancer type, and a model can be generated for a second cancer type based on second more than one classification regions that are different from the first more than one classification regions.
[0050] In one or more aspects, the more than one sample and the additional samples include cell-free nucleic acid.
[0051] In one or more aspects, the method includes performing, by a computing system, a training process using training data to generate a model, wherein the training process includes determining, by the computing system, one or more additional weights for a single sample included in the training data based on an indication of cancer for the single sample being within a threshold confidence level.
[0052] In one or more aspects, the indication of cancer for the single sample is outside a threshold confidence level, and the method includes applying, by the computing system during the training process, a penalty to the weight of the single sample.
[0053] In one or more aspects, the method includes performing, by a computing system and using one or more machine learning algorithms, one or more first iterations of a training process for a model using a portion of the training data; and generating, by the computing system, first output data for the model based on the one or more first iterations of the training process, the first output data corresponding to one or more first additional indications of the presence of cancer in a first single subject among more than one subject, the first single subject corresponding to the portion of the training data.
[0054] In one or more aspects, the method includes combining, by a computing system, the first output data and the training data to produce additional training data; performing, by the computing system, one or more second iterations of the training process for the model using a portion of the additional training data; and generating, by the computing system, second output data for the model based on the one or more second iterations of the training process, the second output data indicating one or more second additional indications of the presence of cancer in a second single subject among the more than one subjects, the second single subject corresponding to a portion of the additional training data.
[0055] In one or more aspects, a weight of a single classification region of the more than one classification region is determined based on the first output data and the second output data.
[0056] In one or more aspects, the method includes determining, by the computing system, that the number of indications of the presence of cancer determined during one or more iterations of a training process is at least a threshold of one or more samples included in the training data; and determining, by the computing system, that modification of one or more weights of the model is not modified or is modified by a minimal amount.
[0057] In one or more aspects, the method includes determining, by the computing system, that an additional number of indications of the presence of cancer determined during one or more iterations of the training process is less than a threshold of one or more additional samples included in the training data; and determining, by the computing system, that a modification to one or more additional weights of the model is modified by an amount that exceeds a minimum amount.
[0058] In one or more aspects, the method comprises combining more than one nucleic acid derived from at least one of the blood or tissue of the subject with a solution comprising an amount of a methyl binding domain (MBD) protein to produce a nucleic acid-MBD protein solution; and washing the nucleic acid-MBD protein solution more than once with a salt solution to produce a plurality of nucleic acid fractions, a single nucleic acid fraction having a threshold number of methylated cytosines in a region having at least a threshold cytosine-guanine content in more than one nucleic acid.
[0059] In one or more aspects, the washes in the more than one wash are performed with a solution having a concentration of sodium chloride (NaCl) and produce a nucleic acid fraction having a range of binding strengths to the MBD protein in the plurality of nucleic acid fractions.
[0060] In one or more aspects, the method includes determining that a first nucleic acid fraction is associated with a first partition among more than one nucleic acid partitions, the first partition corresponding to a first range of binding intensities to an MBD protein; attaching a first molecular barcode to nucleic acid in the first nucleic acid fraction, the first molecular barcode being included in a first set of molecular barcodes associated with the first partition; determining that a second nucleic acid fraction is associated with a second partition among more than one nucleic acid partitions, the second partition corresponding to a second range of binding intensities to the MBD protein that is different from the first range of binding intensities to the MBD protein; and attaching a second molecular barcode to nucleic acid in the second nucleic acid fraction, the second molecular barcode being included in a second set of molecular barcodes associated with the second partition.
[0061] In one or more aspects, the method includes combining at least a portion of the plurality of nucleic acid fractions with an amount of one or more methylation-sensitive restriction endonucleases that cleave molecules having one or more unmethylated cytosines to produce at least a portion of more than one sample for generating sequencing reads.
[0062] In one or more aspects, the method includes combining at least a portion of the plurality of nucleic acid fractions with an amount of one or more methylation-dependent restriction endonucleases that cleave molecules having one or more methylated cytosines to produce at least a portion of more than one sample for generating sequencing reads.
[0063] In one or more aspects, the limit of detection of the model used to determine the tumor fraction of a sample is no greater than 0.05%.
[0064] In one or more aspects, a computing system includes: one or more hardware processors; and one or more non-transitory computer-readable storage media including computer-readable instructions, which, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations including: obtaining training sequence data, the training sequence data including training sequencing reads from more than one sample derived from more than one subject, including a single training sequencing read corresponding to a nucleotide sequence of a nucleic acid fragment included in a sample of the more than one sample and a single training sequencing read corresponding to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The operations also include analyzing the training sequencing reads to determine a first quantitative measure corresponding to a single classification region in more than one classification region derived from the training sequencing reads, at least a portion of the single classification region in more than one classification region corresponding to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer is detected. The operation also includes analyzing the training sequencing reads to determine a second quantitative measure corresponding to more than one control region derived from the training sequencing reads, wherein a single control region in the more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and at least a threshold amount of methylated cytosine in the subject in which the cancer was detected and another subject in which the cancer was not detected. The operation also includes determining a measure of a single classification region in more than one classification region based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region. The operation also includes generating training data, which includes measures of single classification regions in more than one classification region of the training sequence reads of the sample of the training subject. The operation also includes implementing one or more machine learning algorithms using the training data to generate a model to determine an indication of the presence of cancer in the subject based on the amount of methylated cytosine in at least a portion of the more than one classification region, the model including weights of single classification regions in more than one classification region, and at least a portion of the weights of the single classification regions are different from each other.
[0065] In one or more aspects, the computing system can obtain test sequence data from another subject that is not included in more than one subject. The test sequence data can include test sequencing reads of a sample derived from another subject, the test sequencing reads having a single test sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in the other sample and a single test sequencing read corresponding to a molecule having at least a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content included in the nucleotide sequence. The operation can also include using the model and the additional sequence data to determine an indication of the presence of cancer in the additional subject.
[0066] In one or more aspects, the computing system can analyze the test sequencing reads to determine a first quantitative measure corresponding to a single classification region in more than one classification region derived from the test sequencing reads. The computing system can also analyze the test sequencing reads to determine a second quantitative measure corresponding to a single control region in more than one control region derived from the test sequencing reads. The computing system can also determine the measure of a single classification region based on the first quantitative measure of the single classification region and the second quantitative measure of more than one control region. The computing system can also generate an input vector comprising a measure of a single classification region, wherein the model uses the input vector to determine an indication of the presence of cancer in another subject.
[0067] In one or more aspects of the computing system, the one or more machine learning algorithms include one or more classification algorithms, and the indication of the presence of cancer corresponds to a probability of the presence of cancer in the additional subject.
[0068] In one or more aspects of the computing system, the one or more machine learning algorithms include one or more regression algorithms, and the indicator corresponds to an estimate of a tumor fraction for the additional sample.
[0069] In one or more aspects of the computing system, the training sequencing reads constitute a first portion of the training sequence data, and the additional training sequencing reads constitute a second portion of the training sequence data, wherein the additional training sequencing reads are different from the training sequencing reads, and the computing system can analyze at least one of the first portion of the training sequence data or the second portion of the training sequence data to determine a single frequency of more than one variant in a single sample present in more than one sample. The computing system can also determine a variant with a maximum frequency among more than one variants for a single sample, the maximum frequency corresponding to a single frequency with a maximum value among the single frequencies derived from the single sample. The computing system can also determine a single measure of the tumor score of a single sample based on the maximum value of the single frequency derived from the single sample.
[0070] In one or more aspects of the computing system, the training data includes a single measure of tumor score for a single sample of the more than one samples, and the model is generated based on the single measure of tumor score for the single sample of the more than one samples.
[0071] In one or more aspects of the computing system, a metric for a single classification region is determined based on a scaling factor and an error correction factor.
[0072] In one or more aspects of the computing system, more than one classification region individually corresponds to a genomic region where the methylation rate of the genomic region in nucleic acid derived from a cell obtained from a subject in which cancer is present is different from the methylation rate of the genomic region in nucleic acid derived from a cell obtained from a subject in which cancer is not present.
[0073] In one or more aspects of the computing system, the more than one classification regions correspond to first more than one classification regions for a first cancer type, and a model can be generated for a second cancer type based on second more than one classification regions different from the first more than one classification regions.
[0074] In one or more aspects, the computing system can perform a training process using the training data to generate a model, wherein the training process includes determining one or more additional weights for a single sample included in the training data based on an indication of cancer for the single sample being within a threshold confidence level.
[0075] In one or more aspects of the computing system, the indication of cancer for the single sample is outside a threshold confidence level, and the method includes applying, by the computing system during the training process, a penalty to the weight of the single sample.
[0076] In one or more aspects, the computing system can use one or more machine learning algorithms to perform one or more first iterations of a training process for the model using a portion of the training data. The computing system can also generate first output data for the model based on the one or more first iterations of the training process, the first output data corresponding to one or more first additional indications of the presence of cancer in a first single subject of the more than one subject, wherein the first single subject corresponds to the portion of the training data.
[0077] In one or more aspects, the computing system may combine the first output data and the training data to produce additional training data; perform one or more second iterations of the training process for the model using a portion of the additional training data; and generate second output data for the model based on the one or more second iterations of the training process, wherein the second output data indicates one or more second additional indications of the presence of cancer in a second single subject among the more than one subjects, and the second single subject corresponds to the portion of the additional training data.
[0078] In one or more aspects of the computing system, a weight of a single classification region of the more than one classification region is determined based on the first output data and the second output data.
[0079] In one or more aspects, the computing system can determine that the number of indications of the presence of cancer determined during one or more iterations of the training process is at least a threshold of one or more samples included in the training data. The computing system can also determine that modification of one or more weights of the model is not modified or is modified by a minimal amount.
[0080] In one or more aspects, the computing system can determine that an additional number of indications of the presence of cancer determined during one or more iterations of the training process is less than a threshold of one or more additional samples included in the training data; and determine that the modification of one or more additional weights of the model is modified by an amount that exceeds a minimum amount.
[0081] In one or more aspects of the computing system, the limit of detection of the model used to determine the tumor score of the sample is no greater than 0.05%.
[0082] In one or more aspects, one or more computer-readable storage media include computer-readable instructions that, when executed by one or more processors of a computing system, cause the computing system to perform operations including obtaining training sequence data, the training sequence data including training sequencing reads from more than one sample derived from more than one subject, a single training sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in a sample in the more than one sample, and the single training sequencing read corresponds to a molecule having a threshold amount of methylated cytosine in a region of the nucleotide sequence having at least a threshold cytosine-guanine content. The operations also include analyzing the training sequencing reads to determine a first quantitative measure corresponding to a single classification region in the more than one classification region derived from the training sequencing reads, at least a portion of the single classification region in the more than one classification region corresponding to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer was detected. The operation also includes analyzing the training sequencing reads to determine a second quantitative measure corresponding to more than one control region derived from the training sequencing reads, wherein a single control region in the more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and at least a threshold amount of methylated cytosine in the subject in which the cancer was detected and another subject in which the cancer was not detected. The operation also includes determining a measure of a single classification region in more than one classification region based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region. The operation also includes generating training data, which includes measures of single classification regions in more than one classification region of the training sequence reads of the sample of the training subject. The operation also includes implementing one or more machine learning algorithms using the training data to generate a model to determine an indication of the presence of cancer in the subject based on the amount of methylated cytosine in at least a portion of the more than one classification region, the model including weights of single classification regions in more than one classification region, and at least a portion of the weights of the single classification regions are different from each other.
[0083] In one or more aspects, a method includes obtaining, by a computing system having one or more hardware processors and memory, sequencing reads derived from a sample obtained from a subject, wherein a single sequencing read includes a nucleotide sequence corresponding to a nucleic acid fragment included in the sample and corresponds to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The method may also include determining, by the computing system, a first quantitative measure corresponding to a single classification region in more than one classification region derived from the sequencing read, wherein at least a portion of the single classification region in more than one classification region corresponds to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer is detected. The method may also include analyzing the sequencing reads by the computing system to determine a second quantitative measure corresponding to more than one control region derived from the sequencing read, wherein a single control region in more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and having at least a threshold amount of methylated cytosine in a subject in which cancer is detected and another subject in which cancer is not detected. The method may also include determining, by the computing system, more than one metric based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region, wherein a single metric in the more than one metric corresponds to a single classification region in the more than one classification region. The method may also include determining, by the computing system, an indication of the presence of cancer in the subject based on at least a portion of the more than one metric.
[0084] In one or more aspects, the method may further include: determining, by the computing system and using the sequencing data, a distribution of sequence representations of the differentially methylated region, and determining, by the computing system, that at least a threshold amount of the sequence representations included in the distribution overlap with a sub-region of the differentially methylated region. The method may further include determining, by the computing system, that the sub-region of the differentially methylated region is a classification region in more than one classification region.
[0085] In one or more aspects, the method may further include determining, by the computing system, an order of values of the more than one metric, and determining, by the computing system, a subset of classification regions from the more than one classification regions based on the order, wherein a portion of the more than one metric corresponding to the subset of classification regions is used to determine an indication of the presence of cancer in the subject.
[0086] In one or more aspects, the indication of the presence of cancer in the subject is an initial indication of the presence of cancer in the subject, and the method may further include applying, by the computing system, the scaling factor to the initial indication of the presence of cancer in the subject to determine the modified indication of the presence of cancer in the subject.
[0087] In one or more aspects, the indication of the presence of cancer in the subject corresponds to a tumor score.
[0088] In one or more aspects, the sample is a first sample collected at least one of before or at the start of cancer treatment, and the method may also include obtaining, by a computing system having one or more hardware processors and memory, additional sequencing reads derived from a second sample obtained from a subject, wherein a single additional sequencing read includes additional nucleotide sequences corresponding to nucleic acid fragments included in the second sample, and corresponds to additional molecules having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content included in the additional nucleotide sequence. The method may also include determining, by the computing system, an additional first quantitative measure corresponding to a single classification region in more than one classification region derived from the additional sequencing reads, wherein at least a portion of the single classification region in more than one classification region corresponds to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer was detected. The method may also include analyzing additional sequencing reads by a computing system to determine additional second quantitative measurements corresponding to more than one control region derived from additional additional sequencing reads, wherein a single control region in the more than one control region corresponds to an additional genomic region in the reference genome having at least a threshold cytosine-guanine content and at least a threshold amount of methylated cytosine in a subject in which cancer was detected and another subject in which cancer was not detected. The method may also include determining more than one additional metric by a computing system based on an additional first quantitative measurement of a single classification region and an additional second quantitative measurement of more than one control region using a single additional measurement corresponding to a single classification region in more than one classification region. The method may also include determining, by a computing system, an additional indication of the presence of cancer in a subject based on at least a portion of more than one additional measurement.
[0089] In one or more aspects, a method includes obtaining test sequence data from a subject by a computing system having one or more hardware processors and memory, wherein the test sequence data includes test sequencing reads of a sample derived from the subject, a single test sequencing read includes a nucleotide sequence corresponding to a nucleic acid fragment included in another sample, and the single test sequencing read corresponds to a molecule having a threshold amount of methylated cytosine in a region of the nucleotide sequence having at least a threshold cytosine-guanine content. The method may also include analyzing the test sequencing reads by the computing system to determine a first quantitative measure corresponding to a single classification region of more than one classification region derived from the test sequencing reads, at least a portion of the single classification region of more than one classification region corresponding to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least the threshold cytosine-guanine content in a subject in which cancer was detected. The method may also include analyzing the test sequencing reads by a computing system to determine a second quantitative measure corresponding to a single control region in more than one control region derived from the test sequencing reads, wherein the single control region in more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and at least a threshold amount of methylated cytosine in a subject in which cancer is detected and another subject in which cancer is not detected. The method may also include determining a measure of a single classification region by a computing system based on a first quantitative measure of a single classification region and a second quantitative measure of more than one control region. The method may also include generating an input vector comprising a measure of a single classification region by a computing system. The method also includes determining an indication of the presence of cancer in a subject by a computing system by providing an input vector to a model that implements one or more machine learning techniques to generate an indication of the presence of cancer in a subject, the model comprising weights of a single classification region in more than one classification region, and at least a portion of the weights of the single classification regions are different from each other.
[0090] In one or more aspects, the method may also include obtaining training sequence data by a computing system having one or more hardware processors and a memory, the training sequence data including training sequencing reads from more than one sample derived from more than one training subject, wherein a single training sequencing read includes a nucleotide sequence corresponding to a nucleic acid fragment included in a sample in more than one sample, and a single training sequencing read corresponds to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The method may also include analyzing the training sequencing reads by the computing system to determine an additional first quantitative measure corresponding to a single classification region in more than one classification region derived from the training sequencing reads. The method may also include analyzing the training sequencing reads by the computing system to determine an additional second quantitative measure corresponding to more than one control region derived from the training sequencing reads. The method may also include determining an additional measure of a single classification region in more than one classification region based on an additional first quantitative measure of a single classification region and an additional second quantitative measure of more than one control region by the computing system. The method may also include generating, by a computing device, training data including additional metrics for a single classification region in more than one classification region of training sequence reads from samples of more than one training subject. The method may also include implementing, by a computing system and using the training data, one or more machine learning algorithms to generate a model to determine an indication of the presence of cancer in a subject based on the amount of methylated cytosine in at least a portion of the more than one classification region.
[0091] In one or more aspects, the one or more machine learning algorithms include one or more classification algorithms, and the indication of the presence of cancer corresponds to a probability of the presence of cancer in the subject.
[0092] In one or more aspects, the one or more machine learning algorithms include one or more regression algorithms and indicate an estimate of a tumor fraction corresponding to the sample.
[0093] In one or more aspects, the subject's sample and more than one sample from more than one training subject comprise cell-free nucleic acid.
[0094] In one or more aspects, the threshold amount of sequence representation is at least about 70% of the sequence representation included in the distribution.
[0095] In one or more aspects, the second sample is obtained at least one week after the cancer therapy is administered to the subject.
[0096] In one or more aspects, the method can further include analyzing, by the computing system, the indication that cancer is present in the subject relative to other indications that cancer is present in the subject to determine the subject's response to treatment.
[0097] In one or more aspects, the training sequencing reads constitute a first portion of the training sequence data, and the additional training sequencing reads constitute a second portion of the training sequence data, wherein the additional training sequencing reads are different from the training sequencing reads; and the method may also include analyzing at least one of the first portion of the training sequence data or the second portion of the training sequence data by a computing system to determine a single frequency of more than one variant present in a single sample in more than one sample, and determining by the computing system and for the single sample a variant with a maximum frequency among the more than one variants, the maximum frequency corresponding to a single frequency with a maximum value among the single frequencies derived from the single sample. The method may also include determining by the computing system a single measure of the tumor score of the single sample based on the maximum value of the single frequency derived from the single sample.
[0098] In one or more aspects, the training data includes a single measure of tumor fraction for a single sample of the more than one samples, and the model is generated based on the single measure of tumor fraction for the single sample of the more than one samples.
[0099] In one or more aspects, a computing system comprises: a processor; and a memory storing instructions, which, when executed by the processor, configure the computing system to: obtain sequencing reads derived from a sample obtained from a subject, a single sequencing read comprising a nucleotide sequence corresponding to a nucleic acid fragment included in the sample, and corresponding to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The computing system can also determine a first quantitative measure corresponding to a single classification region in more than one classification region derived from the sequencing read, at least a portion of the single classification region in more than one classification region corresponds to a genomic region having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer is detected in a reference genome. The computing system can also analyze the sequencing reads to determine a second quantitative measure corresponding to more than one control region derived from the sequencing reads, wherein a single control region in more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and having at least a threshold amount of methylated cytosine in a subject in which cancer is detected and another subject in which cancer is not detected. The computing system may also determine more than one metric based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region, wherein a single metric in the more than one metrics corresponds to a single classification region in the more than one classification region. The computing system may also determine an indication of the presence of cancer in the subject based on at least a portion of the more than one metrics.
[0100] In one or more aspects of the computing system, the threshold amount of sequence representations is at least about 70% of the sequence representations included in the distribution.
[0101] In one or more aspects, the computing system can use the sequencing data to determine a distribution of sequence representations of the differentially methylated region. The computing system can also determine that at least a threshold amount of the sequence representations included in the distribution overlap with a sub-region of the differentially methylated region. The computing system can also determine by the computing system that a sub-region of the differentially methylated region is a classification region in more than one classification region.
[0102] In one or more aspects, the computing system can determine an order of values of more than one metric; and determine a subset of classification regions from the more than one classification regions based on the order; wherein a portion of the subset of classification regions corresponding to the more than one metric is used to determine an indication of the presence of cancer in the subject.
[0103] In one or more aspects of the computing system, the indication of the presence of cancer in the subject is an initial indication of the presence of cancer in the subject, and the computing system may apply the scaling factor to the initial indication of the presence of cancer in the subject to determine the modified indication of the presence of cancer in the subject.
[0104] In one or more aspects of the computing system, the indication of the presence of cancer in the subject corresponds to a tumor score.
[0105] In one or more aspects of the computing system, the sample is a first sample collected at least one of before or at the start of cancer treatment, and the computing system can obtain additional sequencing reads derived from a second sample obtained from the subject, wherein the single additional sequencing read includes additional nucleotide sequences corresponding to nucleic acid fragments included in the second sample and corresponds to additional molecules having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content included in the additional nucleotide sequence. The computing system can also determine an additional first quantitative measure corresponding to a single classification region of more than one classification region derived from the additional sequencing reads, wherein at least a portion of the single classification region of more than one classification region corresponds to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in the subject in which cancer was detected. The computing system can also analyze additional sequencing reads to determine additional second quantitative measurements corresponding to more than one control region derived from additional sequencing reads, wherein a single control region in the more than one control region corresponds to an additional genomic region in the reference genome having at least a threshold cytosine-guanine content and at least a threshold amount of methylated cytosine in a subject in which cancer is detected and another subject in which cancer is not detected. The computing system can also determine more than one additional metric based on an additional first quantitative measurement of a single classification region and an additional second quantitative measurement of more than one control region using a single additional metric corresponding to a single classification region in more than one classification region in more than one additional metric. The computing system can also determine additional indications of the presence of cancer in a subject based on at least a portion of more than one additional metric.
[0106] In one or more aspects of the computing system, the second sample is obtained at least one week after the cancer therapy is administered to the subject.
[0107] In one or more aspects, the computing system can analyze indications of the presence of cancer in the subject relative to other indications of the presence of cancer in the subject to determine the subject's response to treatment.
[0108] In one or more aspects, a computing system includes: a processor; and a memory storing instructions that, when executed by the processor, configure the system to: obtain test sequence data from a subject, the test sequence data including test sequencing reads of a sample derived from the subject, a single test sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in another sample, and the single test sequencing read corresponds to a molecule having a threshold amount of methylated cytosine in a region of the nucleotide sequence having at least a threshold cytosine-guanine content. The computing system can also analyze the test sequencing reads to determine a first quantitative measure corresponding to a single classification region of more than one classification region derived from the test sequencing reads, at least a portion of the single classification region of more than one classification region corresponding to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer was detected. The computing system can also analyze the test sequencing reads to determine a second quantitative measure corresponding to a single control region in more than one control region derived from the test sequencing reads, wherein the single control region in more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and at least a threshold amount of methylated cytosine in the subject in which the cancer was detected and another subject in which the cancer was not detected. The computing system can also determine the measure of the single classification region based on the first quantitative measure of the single classification region and the second quantitative measure of more than one control region. The computing system can also generate an input vector including the measure of the single classification region. The computing system can also determine the indication of the presence of cancer in the subject by providing an input vector to a model that implements one or more machine learning techniques to generate an indication of the presence of cancer in the subject, the model including the weights of the single classification regions in more than one classification region, and at least a portion of the weights of the single classification regions are different from each other.
[0109] In one or more aspects, a computing system can obtain training sequence data, the training sequence data including training sequencing reads from more than one sample of more than one training subject, wherein a single training sequencing read includes a nucleotide sequence corresponding to a nucleic acid fragment included in a sample of more than one sample, and a single training sequencing read corresponds to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The computing system can also analyze the training sequencing reads to determine an additional first quantitative measure corresponding to a single classification region in more than one classification region derived from the training sequencing reads. The computing system can also analyze the training sequencing reads to determine an additional second quantitative measure corresponding to more than one control region derived from the training sequencing reads. The computing system can also determine an additional measure of a single classification region in more than one classification region based on an additional first quantitative measure of a single classification region and an additional second quantitative measure of more than one control region. The computing system can also generate training data, the training data including additional measures of a single classification region in more than one classification region of training sequence reads from samples of more than one training subject. The computing system may also implement one or more machine learning algorithms using the training data to generate a model to determine an indication of the presence of cancer in a subject based on the amount of methylated cytosine in at least a portion of more than one classification region.
[0110] In one or more aspects of the computing system, the one or more machine learning algorithms include one or more classification algorithms, and the indication of the presence of cancer corresponds to a probability of the presence of cancer in the subject.
[0111] In one or more aspects of the computing system, the one or more machine learning algorithms include one or more regression algorithms and indicate an estimate of a tumor fraction corresponding to the sample.
[0112] In one or more aspects of the computing system, the training sequence reads constitute a first portion of the training sequence data, and the additional training sequencing reads constitute a second portion of the training sequence data, wherein the additional training sequencing reads are different from the training sequencing reads, and the computing system can analyze at least one of the first portion of the training sequence data or the second portion of the training sequence data to determine a single frequency of more than one variant in a single sample present in more than one sample. The computing system can also determine, for a single sample, a variant with a maximum frequency among more than one variants, the maximum frequency corresponding to a single frequency with a maximum value among the single frequencies derived from the single sample. The computing system can also determine a single measure of the tumor score of a single sample based on the maximum value of the single frequency derived from the single sample.
[0113] In one or more aspects of the computing device, the training data includes a single measure of tumor score for a single sample of the more than one samples, and the model is generated based on the single measure of tumor score for the single sample of the more than one samples.
[0114] In one or more aspects, a method includes obtaining sequencing data from more than one subject by a computing system having one or more hardware processors and memory, the sequencing data including sequencing reads from more than one sample derived from more than one subject, a single sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in another sample, and the single sequencing read corresponds to a molecule having a threshold amount of methylated cytosine in a promoter region having at least a threshold cytosine-guanine content included in the nucleotide sequence. The method may also include analyzing the sequencing reads by the computing system to determine a first quantitative measure corresponding to the promoter region derived from the sequencing reads. The method may also include analyzing the sequencing reads by the computing system to determine a second quantitative measure corresponding to a single control region in more than one control region derived from the sequencing reads, the single control region in more than one control region corresponding to another genomic region in a reference genome having at least a threshold cytosine-guanine content and having at least a threshold amount of methylated cytosine in a subject in which cancer was detected and another subject in which cancer was not detected. The method may also include determining, by the computing system, a metric of the promoter region based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region. The method may also include generating, by the computing system, an indication of the methylation state of the promoter region based on the metric having at least the threshold value.
[0115] In one or more aspects, the computing system may include one or more hardware processors; and one or more non-transitory computer-readable storage media including computer-readable instructions, which, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations including the following: obtaining sequencing data from more than one subject, the sequencing data including sequencing reads from more than one sample derived from more than one subject, a single sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in another sample, and the single sequencing read corresponding to a molecule having a threshold amount of methylated cytosine in a promoter region having at least a threshold cytosine-guanine content in the nucleotide sequence. The system may also analyze the sequencing reads to determine a first quantitative measure corresponding to the promoter region derived from the sequencing reads. The system may also analyze the sequencing reads to determine a second quantitative measure corresponding to a single control region in more than one control region derived from the sequencing reads, the single control region in more than one control region corresponding to another genomic region having at least a threshold cytosine-guanine content and at least a threshold amount of methylated cytosine in a subject in which cancer is detected and another subject in which cancer is not detected in a reference genome. The system can also analyze the sequencing reads to determine a second quantitative measure corresponding to a single control region in more than one control region derived from the sequencing reads, the single control region in more than one control region corresponding to another genomic region in the reference genome having at least a threshold cytosine-guanine content and having at least a threshold amount of methylated cytosine in the subject in which the cancer was detected and another subject in which the cancer was not detected. The system can also determine a measure of the promoter region based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region. The system can also generate an indication of the methylation state of the promoter region based on the measure having at least a threshold.
[0116] definition
[0117] In order to more easily understand the present disclosure, some terms are first defined below. The following terms and other definitions of other terms can be set forth through the specification. If the definition of the term set forth below is inconsistent with the definition in the application or patent incorporated by reference, the definition set forth in the present application should be used to understand the meaning of the term.
[0118] As used in this specification and the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to "a method" includes one or more methods and / or steps of the type described herein and / or which will become apparent to one of ordinary skill in the art upon reading this disclosure and so forth.
[0119] It should also be understood that the terms used herein are only for the purpose of describing a particular implementation and are not intended to be limiting. In addition, unless otherwise defined, all technical and scientific terms used herein have the same meanings as those commonly understood by those of ordinary skill in the art to which the present disclosure belongs. When describing and claiming these methods, computer-readable media, and systems, the following terms and their grammatical variants will be used according to the definitions set forth below.
[0120] About: As used herein, "about" or "approximately" applied to one or more values or elements of interest refers to a value or element similar to the reference value or element. In certain embodiments, the term "about" or "approximately" refers to a value or element that falls within a range of 25%, 20%, 19%, 18%, 17%, 16%, 15%, 14%, 13%, 12%, 11%, 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1% or less in either direction (greater than or less than) of the reference value or element, unless otherwise stated or obvious from the context (unless the number would exceed 100% of the possible value or element).
[0121] Administration: As used herein, "administering" or "administering" a therapeutic agent (e.g., an immunotherapeutic agent) to a subject refers to giving, providing, or contacting a composition with a subject. Administration can be accomplished by any of a number of routes, including, for example, topical, oral, subcutaneous, intramuscular, intraperitoneal, intravenous, intrathecal, and intradermal.
[0122] Adaptor: As used herein, "adaptor" refers to a short nucleic acid (e.g., less than about 500 nucleotides, less than about 100 nucleotides, or less than about 50 nucleotides in length) that can be at least partially double-stranded and used to connect to either end or both ends of a given sample nucleic acid molecule. An adaptor can include primer binding sites that allow amplification of nucleic acid molecules flanked by adaptors at both ends, and / or sequencing primer binding sites, including primer binding sites for sequencing applications such as various next-generation sequencing (NGS) applications. An adaptor can also include binding sites for capture probes, such as oligonucleotides attached to a flow cell support, etc. An adaptor can also include a nucleic acid tag as described herein. The nucleic acid tag can be positioned relative to the amplification primer and sequencing primer binding sites so that the nucleic acid tag is included in the amplicon and sequence read of a given nucleic acid molecule. The same or different adaptors can be connected to the corresponding ends of a nucleic acid molecule. In some embodiments, in addition to different nucleic acid tags, the same adaptor is connected to the corresponding ends of a nucleic acid molecule. In some embodiments, the adapter is a Y-shaped adapter, one end of which is flat-ended or tailed as described herein for connection to a nucleic acid molecule, which is also flat-ended or tailed with one or more complementary nucleotides. In yet other example embodiments, the adapter is a bell-shaped adapter, comprising a flat-ended or tailed end for connection to a nucleic acid molecule to be analyzed. Other examples of adapters include T-tailed and C-tailed adapters.
[0123] Alignment: As used herein, "alignment" or "alignment" refers to determining whether at least two sequence representations have at least a threshold amount of homology. In one or more examples, the threshold amount of homology can be at least about 90%, at least about 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, at least about 99%, at least about 99.5%, or at least about 99.9%. In the case where two sequence representations have at least a threshold amount of homology, the two sequence representations can be referred to as "aligned."
[0124] Amplification: As used herein, "amplify" or "amplification" in the context of nucleic acids refers to the production of multiple copies of a polynucleotide or a portion of a polynucleotide starting from a small amount of polynucleotide (e.g., a single polynucleotide molecule), wherein the amplification product or amplicon is typically detectable. Amplification of polynucleotides includes various chemical and enzymatic processes.
[0125] Barcode: As used herein, "barcode" or "molecular barcode" in the context of nucleic acids refers to a nucleic acid molecule that contains a sequence that can be used as a molecular identifier. For example, during next-generation sequencing (NGS) library preparation, a single "barcode" sequence can be added to each DNA fragment so that each read can be identified and sorted prior to final data analysis.
[0126] Cancer type: As used herein, "cancer type" refers to a type or subtype of cancer defined, for example, by histopathology. Cancer type can be defined by any conventional criteria, such as based on occurrence in a given tissue (e.g., blood cancer, central nervous system (CNS) cancer, brain cancer, lung cancer (small cell and non-small cell), skin cancer, nose cancer, laryngeal cancer, liver cancer, bone cancer, lymphoma, pancreatic cancer, intestinal cancer, rectal cancer, thyroid cancer, bladder cancer, kidney cancer, oral cancer, stomach cancer, breast cancer, prostate cancer, ovarian cancer, lung cancer, intestinal cancer, soft tissue cancer, neuroendocrine cancer, gastroesophageal cancer, head and neck cancer, gynecological cancer, colorectal cancer, Cancer can also be classified by stage (e.g., 1, 2, 3 or 4 phases) and whether it is primary or secondary origin.
[0127] Carrier Signal: As used herein, “carrier signal” refers to any intangible medium capable of storing, encoding, or carrying transitory or non-transitory instructions 502 for execution by machine 500, and includes a digital or analog communications signal or other intangible medium that facilitates communication of such instructions 502. Instructions 502 may be sent or received over network 534 via a network interface device using a transitory or non-transitory transmission medium and using any of a number of well-known transmission protocols.
[0128] Cell-free nucleic acid: As used herein, "cell-free nucleic acid" refers to nucleic acids that are not contained in cells or otherwise bound to cells, or in some embodiments, refers to nucleic acids that remain in a sample after the removal of intact cells. Cell-free nucleic acids may include, for example, all unencapsulated nucleic acids derived from body fluids from a subject (e.g., blood, plasma, serum, urine, cerebrospinal fluid (CSF), etc.). Cell-free nucleic acids include DNA (cfDNA), RNA (cfRNA), and hybrids thereof, including genomic DNA, mitochondrial DNA, circulating DNA, siRNA, miRNA, circulating RNA (cRNA), tRNA, rRNA, small nucleolar RNA (snoRNA), Piwi-interacting RNA (piRNA), long non-coding RNA (long ncRNA), and / or fragments of any of these. Cell-free nucleic acids may be double-stranded, single-stranded, or hybrids thereof. Cell-free nucleic acids may be released into body fluids by secretion or cell death procedures, such as cell necrosis, apoptosis, and the like. Some cell-free nucleic acids are released into body fluids from cancer cells, such as circulating tumor DNA (ctDNA). Others are released from healthy cells. ctDNA can be unencapsulated tumor-derived fragmented DNA. The cell-free nucleic acid can have one or more epigenetic modifications, for example, the cell-free nucleic acid can be acetylated, 5-methylated, ubiquitinated, phosphorylated, sumoylated, ribosylated, and / or citrullinated.
[0129] Cellular nucleic acid: As used herein, "cellular nucleic acid" refers to nucleic acids that are distributed within one or more cells at least at the time a sample is taken or collected from a subject, even if such nucleic acids are subsequently removed as part of a given analytical procedure.
[0130] Classification region: As used herein, "classification region" refers to a genomic region that can show changes that are not dependent on sequence in neoplastic cells (e.g., tumor cells and cancer cells) or can show changes that are not dependent on sequence in cfDNA from subjects with cancer relative to cfDNA from subjects in which cancer is not present. Examples of changes that are not dependent on sequence include, but are not limited to, changes in methylation rate (increase or decrease), nucleosome distribution, CTCF binding, transcription start site, and regulatory protein binding region. In one or more examples, changes that are not dependent on sequence in the classification region can indicate the presence of a single form of cancer in the subject. In one or more additional examples, changes that are not dependent on sequence in the classification region can correspond to the presence of multiple forms in the subject. The classification region can be enriched by one or more probes. In addition, the classification region can be defined by a pair of primer binding sites. In addition, the classification region can be defined by a predetermined starting genomic locus and a predetermined ending genomic locus. The classification region can include about 25 nucleotides to about 250 nucleotides, about 50 nucleotides to about 200 nucleotides, or about 75 nucleotides to about 150 nucleotides. For example, the classification region can be a differentially methylated region. "Differentially methylated region" or "DMR" refers to a region of DNA that has a detectably different degree of methylation in at least one cell or tissue type relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type; or a region of DNA that has a detectably different degree of methylation in at least one cell or tissue type obtained from a subject with a disease or disorder relative to the degree of methylation in the same region of DNA in the same cell or tissue type obtained from a healthy subject. In some embodiments, a differentially methylated region has a detectably higher degree of methylation (e.g., a hypermethylated region / hypermethylated target region) in at least one cell or tissue type relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type that contributes to cfDNA in a healthy individual or from the same cell or tissue type of a healthy subject. In some embodiments, the differentially methylated region has a detectably lower degree of methylation (e.g., hypomethylated region / hypomethylated target region) in at least one cell or tissue type relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type (such as other immune cell types and / or cell types that contribute to cfDNA in healthy individuals) or the same cell or tissue type from healthy subjects. In some embodiments, the classified regions include hypermethylated target regions and / or hypomethylated target regions.
[0131] Communication network: As used herein, "communication network" refers to one or more portions of a network 114, 1034, which may be an ad hoc network, an intranet, an extranet, a virtual private network (VPN), a local area network (LAN), a wireless LAN (WLAN), a wide area network (WAN), a wireless WAN (WWAN), a metropolitan area network (MAN), the Internet, a portion of the Internet, a portion of the public switched telephone network (PSTN), a plain old telephone service (POTS) network, a cellular telephone network, a wireless network, 1034, or a combination of two or more such networks. For example, the network 114, 1034, or a portion of the network may include a wireless or cellular network, and the coupling may be a code division multiple access (CDMA) connection, a global system for mobile communications (GSM) connection, or other type of cellular or wireless coupling. In this example, the coupling may implement any of various types of data transmission technologies, such as single carrier radio transmission technology (1xRTT), evolution data optimized (EVDO) technology, general packet radio service (GPRS) technology, enhanced data rates for GSM evolution (EDGE) technology, the third generation partnership project (3GPP) including 3G, fourth generation wireless (4G) networks, universal mobile telecommunications system (UMTS), high speed packet access (HSPA), world interoperability for microwave access (WiMAX), long term evolution (LTE) standards, others defined by various standards development organizations, other remote protocols, or other data transmission technologies.
[0132] Confidence interval: As used herein, a "confidence interval" refers to a range of values so defined that the value of a given parameter lies within the range of values with a certain probability.
[0133] Control sample: As used herein, a "control sample" or "reference sample" refers to a sample obtained from an individual without a known copy number variation.
[0134] Coverage: As used herein, "coverage" or "coverage metric" refers to the number of nucleic acid molecules or sequencing reads that correspond to a particular genomic region of a reference sequence.
[0135] Deoxyribonucleic acid or ribonucleic acid: As used herein, "deoxyribonucleic acid" or "DNA" refers to a natural or modified nucleotide having a hydrogen group at the 2'-position of the sugar portion. DNA can include a nucleotide chain comprising four types of nucleotide bases: adenine (A), thymine (T), cytosine (C), and guanine (G). As used herein, "ribonucleic acid" or "RNA" refers to a natural or modified nucleotide having a hydroxyl group at the 2'-position of the sugar portion. RNA can include a nucleotide chain comprising four types of nucleotides: A, uracil (U), G, and C. As used herein, the term "nucleotide" refers to a natural nucleotide or a modified nucleotide. Certain pairs of nucleotides specifically bind to each other in a complementary manner (called complementary base pairing). In DNA, adenine (A) pairs with thymine (T) and cytosine (C) pairs with guanine (G). In RNA, adenine (A) pairs with uracil (U) and cytosine (C) pairs with guanine (G). When a first nucleic acid chain is combined with a second nucleic acid chain consisting of nucleotides complementary to the nucleotides in the first chain, the two chains combine to form a double strand. As used herein, "nucleic acid sequencing data", "nucleic acid sequencing information", "sequence information", "sequence representation", "nucleic acid sequence", "nucleotide sequence", "genomic sequence", "genetic sequence", "fragment sequence", "sequencing reads" or "nucleic acid sequencing reads" represent any information or data indicating the order and identity of nucleotide bases (e.g., adenine, guanine, cytosine and thymine or uracil) in a molecule (e.g., a whole genome, a whole transcriptome, an exome, an oligonucleotide, a polynucleotide or a fragment) of a nucleic acid such as DNA or RNA. It should be understood that the present teachings contemplate sequence information obtained using all available various techniques, platforms or technologies including but not limited to: capillary electrophoresis, microarrays, ligation-based systems, polymerase-based systems, hybridization-based systems, direct or indirect nucleotide identification systems, pyrophosphate sequencing, ion-based or pH-based detection systems, and electronic signature-based systems.
[0136] Differentially methylated regions: As used herein, "differentially methylated regions" refer to regions of DNA that have a detectably different degree of methylation in at least one cell or tissue type relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type; or regions of DNA that have a detectably different degree of methylation in at least one cell or tissue type obtained from a subject with a disease or disorder relative to the degree of methylation in the same region of DNA in the same cell or tissue type obtained from a healthy subject. In some embodiments, a differentially methylated region has a detectably higher degree of methylation (e.g., a hypermethylated region) in at least one cell or tissue type (such as at least one immune cell type) relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type (such as other immune cell types and / or cell types that contribute to cfDNA in a healthy individual) or from the same cell or tissue type of a healthy subject. In some embodiments, a differentially methylated region has a detectably lower level of methylation (e.g., a hypomethylated region) in at least one cell or tissue type, such as at least one immune cell type, relative to the level of methylation in the same region of DNA from at least one other cell or tissue type, such as other immune cell types and / or cell types that contribute to cfDNA in a healthy individual, or from the same cell or tissue type in a healthy subject.
[0137] Driver Mutation: As used herein, "driver mutation" refers to a mutation that drives cancer progression.
[0138] Epigenetic target region: As used herein, "epigenetic target region" refers to a target region that can show sequence-independent differences in different cell or tissue types (e.g., different types of immune cells) or in neoplastic cells (e.g., tumor cells and cancer cells) relative to normal cells; or a target region that can show sequence-independent differences (i.e., where there is no change in the nucleotide sequence, such as differences in methylation, nucleosome distribution or other epigenetic characteristics) in DNA from different cell types or from subjects with cancer (such as cfDNA) relative to DNA from healthy subjects (such as cfDNA), or in cfDNA derived from different cell or tissue types (e.g., immune, lung, colon, etc.) that do not normally contribute significantly to cfDNA relative to background cfDNA (e.g., cfDNA derived from hematopoietic stem cells). Examples of sequence-independent changes include, but are not limited to, methylation (increase or decrease), nucleosome distribution, cfDNA fragmentation pattern, CCCTC binding factor ("CTCF") binding, transcription start site (e.g., any one or more of binding of RNA polymerase components, binding of regulatory proteins, fragmentation characteristics, and nucleosome distribution) and changes in regulatory protein binding regions. Thus, epigenetic target region sets include, but are not limited to, hypermethylation target region sets, hypomethylation target region sets, and fragmentation variable target region sets, such as CTCF binding sites and transcription start sites. For purposes of the present invention, loci susceptible to focal amplification and / or gene fusions associated with neoplasms, tumors, or cancers may also be included in the epigenetic target region set, because detection of copy number changes by sequencing or detection of fusion sequences that map to more than one locus in a reference genome tends to be more similar to detection of the exemplary epigenetic changes discussed above than detection of nucleotide substitutions, insertions, or deletions, e.g., in that focal amplification and / or gene fusions can be detected at relatively shallow sequencing depths because their detection does not rely on the accuracy of base calls at one or a few single positions. An epigenetic target region set is a group of epigenetic target regions.
[0139] Hypermethylation: As used herein, "hypermethylation" refers to an increase in the level or degree of methylation of one or more nucleic acid molecules relative to other nucleic acid molecules within a population (e.g., a sample) of nucleic acid molecules from the same genomic locus. In some embodiments, hypermethylated DNA can include DNA molecules comprising at least 1 methylated cytosine, at least 2 methylated cytosines, at least 3 methylated cytosines, at least 5 methylated cytosines, or at least 10 methylated cytosines.
[0140] Hypomethylation: As used herein, "hypomethylation" refers to a decrease in the methylation level or degree of one or more nucleic acid molecules relative to other nucleic acid molecules within a population (e.g., a sample) of nucleic acid molecules from the same genomic locus. In some embodiments, hypomethylated DNA includes unmethylated DNA molecules. In some embodiments, hypomethylated DNA may include a DNA molecule comprising 0 methylated cytosines, at most 1 methylated cytosines, at most 2 methylated cytosines, at most 3 methylated cytosines, at most 4 methylated cytosines, or at most 5 methylated cytosines.
[0141] Immunotherapy: As used herein, "immunotherapy" refers to treatment with one or more agents that act to stimulate the immune system to kill cancer cells or at least inhibit the growth of cancer cells, and preferably reduce the further growth of cancer, reduce the size of cancer, and / or eliminate cancer. Some such agents bind to targets presented on cancer cells; some bind to targets presented on immune cells rather than cancer cells; some bind to targets presented on both cancer cells and immune cells. Such agents include, but are not limited to, checkpoint inhibitors and / or antibodies. Checkpoint inhibitors are inhibitors of pathways of the immune system that maintain self-tolerance and regulate the duration and amplitude of physiological immune responses in peripheral tissues to minimize para-tissue damage (see, e.g., Pardoll, Nature Reviews Cancer 12, 252-264 (2012)). Exemplary agents include antibodies for any of the following: PD-1, PD-2, PD-L1, PD-L2, CTLA-40, OX40, B7.1, B7He, LAG3, CD137, KIR, CCR5, CD27, or CD40. Other exemplary agents include proinflammatory cytokines such as IL-1β, IL-6, and TNF-α. Other exemplary agents are T cells activated against tumors, such as T cells activated by expressing chimeric antigens that target tumor antigens recognized by the T cells.
[0142] Insertion / deletion (Indel): As used herein, "Indel" refers to a mutation involving the insertion or deletion of nucleotides in the genome of a subject.
[0143] Limit of Detection (LoD): As used herein, "limit of detection" refers to the minimum amount of a substance (eg, nucleic acid) in a sample that can be measured by a given assay or analytical method.
[0144] Machine-readable medium: As used herein, "machine-readable medium" refers to a component, device, or other tangible medium capable of temporarily or permanently storing instructions 502 and data, and may include, but is not limited to, random access memory (RAM), read-only memory (ROM), buffer memory, flash memory, optical media, magnetic media, cache memory, other types of storage devices (e.g., erasable programmable read-only memory (EEPROM)), and / or any suitable combination thereof. The term "machine-readable medium" may be understood to include a single medium or more than one medium (e.g., a centralized or distributed database, or associated caches and servers) capable of storing instructions 502. The term "machine-readable medium" should also be understood to include any medium or combination of more than one medium capable of storing instructions 502 (e.g., code) for execution by the machine 500, such that when the instructions 502 are executed by one or more processors 504 of the machine 500, the machine 500 performs any one or more of the methods described herein. Therefore, "machine-readable medium" refers to a single storage device or device, as well as a "cloud-based" storage system or storage network that includes more than one storage device or device. The term "machine-readable medium" does not include the signal itself.
[0145] Maximum MAF: As used herein, "maximum MAF" or "max MAF" refers to the maximum MAF (mutant allele fraction) of all somatic variants in a sample.
[0146] Methylation: As used herein, "methylation" or "DNA methylation" refers to the addition of a methyl group to a nucleotide base in a nucleic acid molecule. In some embodiments, methylation refers to the addition of a methyl group to a cytosine at a CpG site (cytosine-phosphate-guanine site (i.e., a cytosine followed by a guanine in the 5'→3' direction of a nucleic acid sequence). In some embodiments, DNA methylation refers to the addition of a methyl group to an adenine, such as at N 6-methyladenine. In some embodiments, DNA methylation is 5-methylation (modification of the 5th carbon of the 6-carbon ring of cytosine). In some embodiments, 5-methylation refers to adding a methyl group to the 5C position of cytosine to generate 5-methylcytosine (5mC). In some embodiments, methylation includes derivatives of 5mC. Derivatives of 5mC include but are not limited to 5-hydroxymethylcytosine (5-hmC), 5-formylcytosine (5-fC) and 5-carboxylcytosine (5-caC). In some embodiments, DNA methylation is 3C methylation (modification of the 3rd carbon of the 6-carbon ring of cytosine). In some embodiments, 3C methylation includes adding a methyl group to the 3C position of cytosine to generate 3-methylcytosine (3mC). Methylation can also occur at non-CpG sites, for example, methylation can occur at CpA, CpT or CpC sites. DNA methylation can change the activity of methylated DNA regions. For example, when DNA in the promoter region is methylated, transcription of the gene can be inhibited. DNA methylation is essential for normal development, and abnormalities in methylation can disrupt epigenetic regulation. Disruptions in epigenetic regulation, such as inhibition, can cause diseases such as cancer. DNA methylation of promoters can indicate cancer.
[0147] Methylation-dependent nuclease: As used herein, "methylation-dependent nuclease" refers to a nuclease that preferentially cleaves methylated DNA relative to unmethylated DNA. For example, a methylation-dependent nuclease can cut at or near a recognition sequence (such as a restriction site) in a manner that depends on the methylation of at least one nucleobase (such as cytosine) in the recognition sequence. In some embodiments, in a standard nucleic acid cleavage (nucleolysis) assay, the nucleic acid cleavage activity of the methylation-dependent nuclease on the methylated recognition site is at least 10 times, 20 times, 50 times, or 100 times higher than that of an unmethylated control. Methylation-dependent nucleases include methylation-dependent restriction endonucleases.
[0148] Methylation-dependent restriction endonuclease: As used herein, "methylation-dependent restriction endonuclease" or "MDRE" refers to a restriction endonuclease that depends on DNA methylation (e.g., cytosine methylation), i.e., the presence or absence of a methyl group in a nucleotide base changes the rate at which the enzyme cleaves the target DNA. In some embodiments, the methylation-dependent restriction endonuclease does not cleave DNA if the specific nucleotide base at the recognition sequence is not methylated. For example, MspJI is a methylation-dependent restriction endonuclease with a recognition sequence of "mCNNR (N9)", and if there is no methylated cytosine (mC) in the recognition sequence, MspJI does not cleave DNA.
[0149] Methylation-sensitive nuclease: As used herein, "methylation-sensitive nuclease" refers to a nuclease that preferentially cleaves unmethylated DNA relative to methylated DNA. For example, a methylation-sensitive nuclease can cut at or near a recognition sequence (such as a restriction site) in a manner that depends on the lack of methylation of at least one nucleobase (such as cytosine) in the recognition sequence. In some embodiments, in a standard nucleic acid cleavage assay, the nucleic acid cleavage activity of the methylation-sensitive nuclease on the unmethylated recognition site is at least 10 times, 20 times, 50 times, or 100 times higher than that of the methylated control. Methylation-sensitive nucleases include methylation-sensitive restriction endonucleases.
[0150] Methylation-sensitive restriction endonucleases: As used herein, "methylation-sensitive restriction endonucleases" or "MSREs" refer to restriction endonucleases that are sensitive to the methylation state of DNA (e.g., cytosine methylation), i.e., the presence or absence of a methyl group in a nucleotide base changes the rate at which the enzyme cleaves the target DNA. In some embodiments, a methylation-sensitive restriction endonuclease does not cleave DNA if a particular nucleotide base at the recognition sequence is methylated. For example, HpaII is a methylation-sensitive restriction endonuclease with a recognition sequence of "CCGG", and if the second cytosine in the recognition sequence is methylated, HpaII does not cleave DNA.
[0151] Methylation rate: As used herein, "methylation rate" refers to the probability, possibility or percentage of a given base (e.g., cytosine residues in CpG) being methylated on a DNA molecule in a specific genomic region analyzed in a sample. In some embodiments, the methylation rate can be applied to a defined region comprising one or more potential methylated bases. In some embodiments, the methylation rate refers to the percentage of methylated CpG residues in a DNA molecule. In some embodiments, the methylation rate refers to the percentage of methylated CpG residues in a molecule aligned with a specific genomic position or genomic region. The methylation rate can be measured by a variety of methods, including but not limited to using bisulfite sequencing (any single base resolution, such as TAPS, EM-SEQ, etc.) or using partitions (DNA molecular resolution). The methylation rate can be measured in different ways. In the case of bisulfite sequencing or any other base level resolution sequencing method, an estimation can be made by counting how many DNA fragments end up in each methylation-dependent partition, or by counting the number of CpGs converted by each fragment. In addition, in the case of methylation-dependent partitioning, the methylation rate calculation can be normalized using a set of predefined regions (i.e., positive control regions and / or negative control regions) with known methylation status or spiked synthetic DNA with known methylation status, deriving a parameterized partition distribution of methylation rates and estimating methylation rates using a maximum likelihood method. In one or more instances, the methylation rate can be determined by determining the abundance of a portion of the sequencing reads corresponding to the genomic region. The portion of the genomic region can include a number of genomic locations where at least a threshold number of sequencing reads of the genomic region overlap.
[0152] Methylation status: As used herein, "methylation status" or "methylation state" can refer to the presence or absence of a methyl group on a DNA base (e.g., cytosine) at a specific genomic position in a nucleic acid molecule. It can also refer to the degree of methylation in a nucleic acid sequence (e.g., a highly methylated, hypomethylated, moderately methylated, or unmethylated nucleic acid molecule). Methylation state can also refer to the number of methylated nucleotides in a particular nucleic acid molecule.
[0153] Modified nucleotide-specific binding reagent: As used herein, refers to a binding reagent that is specific for a modified nucleotide or targets a modified nucleotide. For example, the modified nucleotide can be a nucleotide that has been methylated, and therefore, the binding reagent can be specific for the methylated nucleotide. Examples of binding reagents include, but are not limited to, a methyl binding domain (MBD) of a methylation binding protein ("MBP") or a variant thereof, an antibody (and antibody variants, such as a single-chain antibody), an aptamer, or a combination thereof. Therefore, as disclosed throughout the text, the use of MBD can be replaced with any other modified nucleotide-specific binding reagent, provided that the modified nucleotide-specific binding reagent has the desired specificity and affinity for the specific modified base of interest in the selected embodiment.
[0154] Mutant allele fraction: As used herein, "mutant allele fraction", "mutation dose" or "MAF" refers to the fraction of nucleic acid molecules containing allelic changes or mutations at a given genomic position in a given sample. MAF is usually expressed as a fraction or percentage. For example, MAF can be less than about 0.5, 0.1, 0.05 or 0.01% (i.e., less than about 50%, 10%, 5% or 1%) of all somatic variants or alleles present at a given site.
[0155] Mutation: As used herein, "mutation" refers to a variation from a known reference sequence, and includes mutations such as single nucleotide variants (SNVs), copy number variants or variations (CNVs) / aberrations, insertions or deletions (indels), gene fusions, transversions, translocations, frameshifts, duplications, repeat amplifications, and epigenetic variants. Mutations can be germline mutations or somatic mutations. In some instances, the reference sequence for comparison purposes is the wild-type genomic sequence of the species of the subject providing the test sample, typically the human genome.
[0156] Mutation caller: As used herein, "mutation caller" refers to an algorithm (implemented in software or other computer-implemented manner) for identifying mutations in test sample data (eg, sequence information obtained from a subject).
[0157] Mutation count: As used herein, "mutation count" or "mutational count" refers to the number of somatic mutations in the entire genome or exome or targeted region of a nucleic acid sample.
[0158] Negative control region: As used herein, "negative control region" refers to a genomic region that is expected to be unmethylated or hypomethylated in substantially all samples, regardless of whether the DNA is derived from cancer cells or normal cells.
[0159] Neoplasm: As used herein, the terms "neoplasm" and "tumor" are used interchangeably. They refer to an abnormal growth of cells in a subject's body. A neoplasm or tumor can be benign, potentially malignant, or malignant. Malignant tumors are called cancer or cancerous tumors.
[0160] Next Generation Sequencing: As used herein, "Next Generation Sequencing" or "NGS" refers to sequencing technologies with increased throughput compared to traditional Sanger and capillary electrophoresis-based methods, e.g., the ability to generate hundreds of thousands of relatively small sequencing reads at a time. Some examples of next generation sequencing technologies include, but are not limited to, sequencing by synthesis, sequencing by ligation, and sequencing by hybridization.
[0161] Nucleic acid tag: As used herein, "nucleic acid tag" refers to a short nucleic acid (e.g., less than about 500 nucleotides, about 100 nucleotides, about 50 nucleotides, or about 10 nucleotides in length) used to distinguish nucleic acids from different samples (e.g., representing a sample index), or different nucleic acid molecules of different types or different treatments in the same sample (e.g., representing a molecular barcode). Nucleic acid tags contain predetermined, fixed, non-random, random, or semi-random oligonucleotide sequences. Such nucleic acid tags can be used to label different nucleic acid molecules or different nucleic acid samples or subsamples. Nucleic acid tags can be single-stranded, double-stranded, or at least partially double-stranded. Nucleic acid tags optionally have the same length or different lengths. Nucleic acid tags can also include double-stranded molecules with one or more blunt ends, including 5' or 3' single-stranded regions (e.g., overhangs), and / or include one or more other single-stranded regions at other positions within a given molecule. Nucleic acid tags can be attached to one or both ends of other nucleic acids (e.g., sample nucleic acids to be amplified and / or sequenced). Nucleic acid tags can be decoded to reveal information, such as the sample source, form or treatment of a given nucleic acid. For example, nucleic acid tags can also be used to realize the collection and / or parallel processing of more than one sample containing nucleic acids with different molecular barcodes and / or sample indexes, wherein the nucleic acid is subsequently deconvoluted by detecting (e.g., reading) nucleic acid tags. Nucleic acid tags can also be referred to as identifiers (e.g., molecular identifiers, sample identifiers). Additionally or alternatively, nucleic acid tags can be used as molecular identifiers (e.g., for distinguishing different molecules or amplicons of different parental molecules in the same sample or subsample). For example, this includes uniquely labeling different nucleic acid molecules in a given sample, or non-uniquely labeling these molecules. In the case of non-uniquely labeling applications, a limited number of tags (i.e., molecular barcodes) can be used to label each nucleic acid molecule, so that different molecules can be distinguished based on their endogenous sequence information (e.g., they are mapped to the starting and / or ending position of the selected reference sequence, the subsequence and / or sequence length at one end or both ends of the sequence) in conjunction with at least one molecular barcode. A sufficient number of different molecular barcodes are used so that the probability that any two molecules may have the same endogenous sequence information (e.g., start and / or end position, subsequence at one or both ends of the sequence, and / or length) and also have the same molecular barcode is low (e.g., less than about 10%, less than about 5%, less than about 1%, or less than about 0.1%).
[0162] Partition: As used herein, "partition" refers to the physical separation or classification of a mixture of nucleic acid molecules in a sample based on the characteristics of the nucleic acid molecules. Partitions can be physical partitions of molecules. Partitions can include dividing nucleic acid molecules into groups (groups) or sets based on epigenetic features (e.g., methylation) levels. For example, nucleic acid molecules can be partitioned based on the methylation levels of nucleic acid molecules. In some embodiments, methods and systems for partitioning can be found in PCT patent application No. PCT / US2017 / 068329, which is hereby incorporated by reference in its entirety.
[0163] Partition group: As used herein, "partition group" or "partition" refers to a nucleic acid molecule group that is partitioned into a set or group based on the different binding affinities of a nucleic acid molecule or a protein associated with a nucleic acid molecule and a binding agent. Partition groups may also be referred to as subsamples. Binding agents preferentially bind to nucleic acid molecules comprising nucleotides with epigenetic modifications. For example, if the epigenetic modification is methylation, the binding agent may be a methyl binding domain (MBD) protein. In some embodiments, partition groups may include nucleic acid molecules belonging to a specific epigenetic feature (e.g., methylation) level or degree. For example, nucleic acid molecules may be partitioned into three groups-one group is a highly methylated nucleic acid molecule (a first subsample, a high partition, a high partition group, or a high methylation partition group), a second group is a low methylated nucleic acid molecule (a second subsample, a low partition, a low partition group, or a low methylation partition group), and a third group is a medium methylated nucleic acid molecule (a third subsample, a medium partition group, a medium methylation partition group, a residual partition, or a residual partition group). In another example, nucleic acid molecules can be partitioned based on the number of methylated nucleotides - one partition group can have nucleic acid molecules with 9 methylated nucleotides and another partition group can have unmethylated nucleic acid molecules (zero methylated nucleotides).
[0164] Polynucleotide: As used herein, "polynucleotide", "nucleic acid", "nucleic acid molecule", "polynucleotide molecule" or "oligonucleotide" refers to a linear polymer of nucleosides (including deoxyribonucleosides, ribonucleosides or their analogs) connected by internucleoside linkages. A polynucleotide may contain at least three nucleosides. Oligonucleotides typically range in size from a few monomeric units, e.g., 3-4, to several hundred monomeric units. Unless otherwise noted, whenever a polynucleotide is represented by a letter sequence such as "ATGCCTG", it is understood that the nucleotides are in the order 5'→3' from left to right, and in the case of DNA, "A" represents deoxyadenosine, "C" represents deoxycytidine, "G" represents deoxyguanosine, and "T" represents deoxythymidine. The letters A, C, G, and T may be used to refer to the base itself, to the nucleoside or nucleotide comprising the base, as is standard in the art.
[0165] Positive control region: As used herein, a "positive control region" refers to a genomic region that is expected to be methylated or hypermethylated in substantially all samples, regardless of whether the DNA is derived from cancer cells or normal cells.
[0166] Probe: As used herein, "probe" refers to a polynucleotide that contains functionality. The functionality can be a detectable label (fluorescence), a binding moiety (biotin), or a solid support (magnetically attracted particles or chips). The probe can include a single-stranded DNA / RNA polynucleotide or a double-stranded DNA polynucleotide that hybridizes to a target nucleic acid sequence (e.g., Probes, Agilent Technologies. Sequence capture using probes generally depends in part on the number of consecutive nucleotides in at least a portion of the target nucleic acid sequence that are complementary (or nearly complementary) to the sequence of the probe. In some examples, the probe can correspond to a driver mutation.
[0167] Processing: As used herein, the terms "processing," "calculating," and "comparing" are used interchangeably. In certain applications, these terms refer to determining differences, e.g., differences in number or sequence. For example, gene expression, copy number variation (CNV), indels, and / or single nucleotide variant (SNV) values or sequences can be processed.
[0168] Processor: As used herein, a "processor" refers to any circuit or virtual circuit (a physical circuit simulated by logic executed on an actual processor) that manipulates data values according to control signals (e.g., "commands," "opcodes," "machine codes," etc.) and produces corresponding output signals for operating a machine. A processor may be, for example, a CPU, a RISC processor, a CISC processor, a GPU, a DSP, an ASIC, an RFIC, or any combination thereof. A processor may also be a multi-core processor with two or more independent processors (sometimes called "cores") that can execute instructions simultaneously.
[0169] Promoter region: As used herein, "promoter region" refers to a DNA sequence recognized by the synthetic machinery of the cell or introduced synthetic machinery required to initiate specific transcription of a gene.
[0170] Quantitative measure: As used herein, "quantitative measure" refers to an absolute measure or a relative measure. A quantitative measure can be, but is not limited to, a number, a statistical measure (e.g., frequency, mean, median, standard deviation, or quantile) or a degree or relative amount (e.g., high, medium, and low). A quantitative measure can be the ratio of two quantitative measures. A quantitative measure can be a linear combination of quantitative measures. A quantitative measure can be a normalized measure.
[0171] Reference sequence: As used herein, "reference sequence" refers to a known sequence for the purpose of comparing with an experimentally determined sequence. For example, a known sequence can be a whole genome, a chromosome, or any segment thereof. The reference sequence can include at least about 20, at least about 50, at least about 100, at least about 200, at least about 250, at least about 300, at least about 350, at least about 400, at least about 450, at least about 500, at least about 1000, or more nucleotides. The reference sequence can be aligned with a single continuous sequence of a genome or chromosome, or can include non-continuous segments aligned with different regions of a genome or chromosome. Exemplary reference sequences include, for example, human genome reference sequences, such as hG19 and hG38.
[0172] Sample: As used herein, "sample" means anything capable of being analyzed by the methods and / or systems disclosed herein.
[0173] Sensitivity: As used herein, "sensitivity" refers to the probability of detecting the presence of single nucleotide variants, insertions and deletions at a given MAF and coverage, and the probability of detecting the presence of copy number variants at a given tumor fraction and coverage.
[0174] Sequencing: As used herein, "sequencing" refers to any of a number of techniques used to determine the sequence (eg, the identity and order of monomeric units) of a biological molecule, eg, a nucleic acid such as DNA or RNA. Exemplary sequencing methods include, but are not limited to, targeted sequencing, single molecule real-time sequencing, exon or exome sequencing, intron sequencing, electron microscopy-based sequencing, panel sequencing, transistor-mediated sequencing, direct sequencing, random shotgun sequencing, Sanger dideoxy terminator sequencing, whole genome sequencing, hybridization sequencing, pyrosequencing, capillary electrophoresis, duplex sequencing, cycle sequencing, single base extension sequencing, solid phase sequencing, high throughput sequencing, massively parallel signature sequencing, emulsion PCR, low denaturing temperature co-amplification PCR (COLD-PCR), multiplex PCR, reversible dye terminator sequencing, paired-end sequencing, near-term sequencing, exonuclease sequencing, ligation sequencing, short read sequencing, single molecule sequencing, sequencing by synthesis, real-time sequencing, reverse terminator sequencing, nanopore sequencing, 454 sequencing, Solexa Genome Analyzer sequencing, SOLiD TMSequencing, MS-PET sequencing, and combinations thereof. In some embodiments, sequencing can be performed by a genetic analyzer, such as, for example, a genetic analyzer commercially available from Illumina, Inc., Pacific Biosciences, Inc., or Applied Biosystems / Thermo Fisher Scientific, and the like.
[0175] Single nucleotide variant: As used herein, "single nucleotide variant" or "SNV" refers to a mutation or variation of a single nucleotide occurring at a specific position in the genome.
[0176] Somatic mutation: As used herein, "somatic mutation" means a genomic mutation that occurs after conception. Somatic mutations can occur in any cell of the body except germ cells and are therefore not passed on to future generations.
[0177] Specific binding: As used herein, "specific binding" in the context of a probe or other oligonucleotide and a target sequence means that under appropriate hybridization conditions, the oligonucleotide or probe hybridizes to its target sequence or a copy thereof to form a stable probe: target hybrid, while minimizing the formation of stable probe: non-target hybrids. Thus, the probe hybridizes to a target sequence or a copy thereof to a greater extent than to a non-target sequence to achieve capture or detection of the target sequence. Appropriate hybridization conditions are well known in the art and can be predicted based on sequence composition or can be determined by use of routine testing procedures (see, e.g., Sambrook et al., Molecular Cloning, A Laboratory Manual, 2nd ed. (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1989) at §§1.90-1.91, 7.37-7.57, 9.47-9.51 and 11.47-11.57, especially §§9.50-9.51, 11.12-11.13, 11.45-11.47 and 11.55-11.57, incorporated herein by reference).
[0178] Subject: As used herein, "subject" refers to an animal, such as a mammalian species (e.g., human), or an avian (e.g., bird) species, or other organisms, such as plants. More specifically, the subject can be a vertebrate, e.g., a mammal such as a mouse, a primate, an ape, or a human. Animals include farm animals (e.g., production cattle, dairy cows, poultry, horses, pigs, etc.), sports animals, and companion animals (e.g., pets or support animals). The subject can be a healthy individual, an individual having or suspected of having a disease or a tendency to suffer from the disease, or an individual in need of treatment or suspected of needing treatment. The terms "individual" or "patient" are intended to be used interchangeably with "subject".
[0179] For example, the subject can be an individual who has been diagnosed with cancer, will receive cancer treatment and / or has received at least one cancer treatment. The subject can be in remission of cancer. As another example, the subject can be an individual diagnosed with an autoimmune disease. As another example, the subject can be a female individual who is pregnant or planning to become pregnant, who may have been diagnosed or suspected of having a disease, such as cancer, an autoimmune disease.
[0180] Target region: As used herein, "target region" refers to a genomic locus that is targeted for identification and / or capture, e.g., by using a probe (e.g., by sequence complementarity). "Target region set" or "set of target regions" refers to more than one genomic locus that is targeted for identification and / or capture, e.g., by using a set of probes (e.g., by sequence complementarity).
[0181] Threshold: As used herein, "threshold" refers to a predetermined value of experimentally determined values used to characterize the same parameter of different samples, the characterization of the experimentally determined values depending on their relationship to the threshold.
[0182] Tumor score: As used herein, "tumor score" refers to an estimate of the fraction of nucleic acid molecules derived from a tumor in a given sample. For example, the tumor score of a sample can be a measure derived from the maximum MAF of the sample or the sequencing coverage pattern of the sample or the length of cfDNA fragments in the sample or any other selected feature of the sample. In some cases, the tumor score of a sample is equal to the maximum MAF of the sample.
[0183] Variant: As used herein, "variant" may be referred to as an allele. Variants typically occur at a frequency of 50% (0.5) or 100% (1), depending on whether the allele is heterozygous or homozygous. For example, germline variants are inherited and typically have a frequency of 0.5 or 1. However, somatic variants are acquired variants and typically have a frequency of <0.5. The major allele and minor allele of a genetic locus refer to nucleic acids having a locus in which the locus is occupied by nucleotides of a reference sequence and variant nucleotides that are different from the reference sequence, respectively. Measurements at a locus may take the form of an allele fraction (AF), which measures the frequency at which an allele is observed in a sample.
[0184] Details
[0185] Cancer is usually caused by the accumulation of mutations in the genes of individual cells, at least some of which cause improperly regulated cell division. Such mutations may include single nucleotide variations (SNVs), gene fusions, insertions, inversions, translocations and inversions. These mutations may also include copy number variations, which correspond to an increase or decrease in the copy number of a gene in a tumor genome relative to an individual's non-cancerous cell. The amount of the cell-free nucleic acid mutated in the degree of mutation present in the cell-free nucleic acid and in the sample can be used as a biomarker to determine tumor progression, predict patient outcomes and refine treatment options. In various examples, the degree of mutation present in the cell-free nucleic acid may be indicated by the tumor cell copy number and tumor score of a given sample.
[0186] In addition, cancer can be indicated by non-sequence modifications such as methylation. Examples of methylation changes in cancer include local increases in DNA methylation in CpG islands at the TSS of genes involved in normal growth control, DNA repair, cell cycle regulation and / or cell differentiation. This increase in the amount of methylation can be associated with an abnormal loss of transcriptional capacity of the genes involved, and occurs at least as frequently as point mutations and deletions that cause changes in gene expression.
[0187] Therefore, DNA methylation profiling can be used to detect abnormal methylation in sample DNA. The DNA may correspond to certain genomic regions ("differentially methylated regions" or "DMRs") that are normally hypermethylated or hypomethylated in a given sample type (e.g., cfDNA from the bloodstream), but which may show abnormal methylation levels associated with a neoplasm or cancer, e.g., due to abnormally increased contribution of tissue to the sample type (e.g., due to increased DNA shedding in or around a neoplasm or cancer) and / or due to genomic methylation levels that are altered during development or perturbed by a disease (e.g., cancer or any cancer-related disease).
[0188] Some methods of measuring DNA methylation can make it difficult to accurately determine the amount of DNA methylation. The accuracy of determining DNA methylation can affect the accuracy of the estimate of the tumor score of a sample. Since the tumor score can be used to determine whether a sample is derived from a subject in which a tumor is present or absent, determining the accuracy of the estimate of the tumor score can affect the diagnosis and / or treatment decisions of an individual.
[0189] The methods and systems described herein relate to accurately generating information indicating the amount of nucleic acid methylation using data indicating the amount of binding of nucleic acids to methyl binding domains (MBDs). In various instances, the present application relates to systems and processes for determining an estimate of a tumor score of a sample. In one or more instances, the amount of methylation of a nucleic acid can be determined based on the binding strength of the nucleic acid to a methyl binding domain (MBD). The nucleic acid can be partitioned according to the binding strength to the MBD. Additionally, the number of cytosine-guanine (CG) regions of the nucleic acid can be determined. The amount of methylation of a classification region of a nucleic acid can be determined based on the partition information associated with the nucleic acid and the number of cytosine-guanine regions of the nucleic acid. The classification region can have different amounts of methylation in tumor cells and non-tumor cells. The estimate of the tumor score of a sample can be determined according to the amount of methylation of the classification region.
[0190] In at least some embodiments, methods, systems, techniques, and architectures can implement a model configured to have at least one of a parameter or a weight that can be modified to more accurately fit the methylation data provided to the model. The methods, systems, techniques, and architectures also involve implementing a number of optimization procedures during training of the model to generate a model that more accurately predicts a metric indicating the presence or absence of a tumor than other systems, methods, techniques, and architectures. In addition, the methods, techniques, and procedures for generating information for generating methylation data reduce the amount of noise present in the methylation data, which results in a metric that more accurately predicts the presence or absence of a tumor than other methods, techniques, and procedures.
[0191] Figure 1is a schematic diagram of an example environment 100 for identifying nucleic acids corresponding to a classification region of a reference sequence, wherein the classification region has at least a threshold number of CpGs, according to one or more embodiments. In one or more examples, the disease under consideration is a type of cancer. Non-limiting examples of such cancers include biliary tract cancer, bladder cancer, transitional cell carcinoma, urothelial carcinoma, brain cancer, glioma, astrocytoma, breast cancer, metaplastic carcinoma, cervical cancer, cervical squamous cell carcinoma, rectal cancer, colorectal cancer, colon cancer, hereditary nonpolyposis colorectal cancer, colorectal adenocarcinoma, gastrointestinal stromal tumor (GIST), endometrial cancer, endometrial stromal sarcoma, esophageal cancer, esophageal squamous cell carcinoma, esophageal adenocarcinoma, ocular melanoma, uveal melanoma, gallbladder cancer, gallbladder adenocarcinoma, renal cell carcinoma, clear cell renal cell carcinoma, transitional cell carcinoma, urothelial carcinoma, Wilms tumor, leukemia, acute lymphocytic leukemia (ALL), acute myeloid leukemia (AML), chronic lymphocytic leukemia (CLL), chronic myeloid leukemia (CML), chronic myelomonocytic leukemia (CMML), liver cancer, liver epithelial carcinoma (liver carcinoma), hepatoma, hepatocellular carcinoma, cholangiocarcinoma, hepatoblastoma, lung cancer, non-small cell lung cancer (NSCLC), mesothelioma, B-cell lymphoma, non-Hodgkin's lymphoma, diffuse large B-cell lymphoma, mantle cell lymphoma, T-cell lymphoma, non-Hodgkin's lymphoma, precursor T-lymphoblastic lymphoma / leukemia, peripheral T-cell lymphoma, multiple myeloma, nasopharyngeal carcinoma (NPC), neuroblastoma, oropharyngeal cancer, oral squamous cell carcinoma, osteosarcoma, ovarian cancer, pancreatic cancer, pancreatic ductal adenocarcinoma, pseudopapillary tumor, acinar cell carcinoma, prostate cancer, prostate adenocarcinoma, skin cancer, melanoma, malignant melanoma, cutaneous melanoma, small intestine cancer, stomach cancer, gastric carcinoma, gastrointestinal stromal tumor (GIST), uterine cancer or uterine sarcoma. Prostate cancer, prostate adenocarcinoma, skin cancer, melanoma, malignant melanoma, skin melanoma, small intestine cancer, stomach cancer, gastric carcinoma, gastrointestinal stromal tumor (GIST), uterine cancer or uterine sarcoma.
[0192] Environment 100 may include sample 102. Sample 102 may be derived from a biological fluid obtained from a subject. For example, sample 102 may be derived from blood obtained from a subject. In one or more other examples, sample 102 may be derived from a tissue of a subject. In various examples, sample 102 may be derived from more than one source. For illustration, sample 102 may be derived from one or more fluids of a subject and / or from a tissue of a subject. In one or more illustrative examples, a subject may be a mammal. In one or more other illustrative examples, a subject may be a human. In one or more other illustrative examples, a subject may be a non-human mammal.
[0193] Sample 102 may include many nucleic acids 104. Single nucleic acid 104 may include many regions of cytosine molecules and guanine molecules with at least a threshold number. In one or more examples, single nucleic acid 104 may include a region of cytosine-guanine dinucleotides with at least a threshold number. In various examples, at least a portion of the cytosine-guanine pairs included in the region may be located sequentially in the sequence of nucleic acid 104. In one or more illustrative examples, the nucleic acid region with at least a threshold amount of cytosine-guanine pairs may be referred to as "CG region" or "CpG region" herein. In one or more examples, the CG region may include at least 200 CpG dinucleotides. In one or more illustrative examples, the CG region may include 200 CpG dinucleotides to 5000 CpG dinucleotides, 300 CpG dinucleotides to 3000 CpG dinucleotides, 200 CpG dinucleotides to 2500 CpG dinucleotides, or 500 CpG dinucleotides to 1500 CpG dinucleotides. Additionally, the CG region can have a GC percentage of at least 50% and an observed to expected CpG ratio of at least 60%. The observed to expected CpG ratio can be calculated, where the observed CpG is the number of CpGs identified in a given genomic region, and the expected CpG is the number of cytosines multiplied by the number of guanines divided by the number of bases in the genomic region. The expected CpG can also be calculated by:
[0194] ((number of cytosines + number of guanines) / 2) 2 / length of the genomic region.
[0195] For example, CG regions can be determined using the techniques described in Gardiner-Garden M, Frommer M (1987). "CpG islands in vertebrate genomes". Journal of Molecular Biology. 196(2): 261-282. and / or Saxonov S, Berg P, Brutlag DL (2006). "A genome-wide analysis of CpG dinucleotides in the human genome distinguishes two distinct classes of promoters". Proc Natl Acad Sci USA. 103(5): 1412-1417.
[0196] exist Figure 1 In the illustrative example of , a portion of the sequence of the example nucleic acid 104 may include a first CG region 106, a second CG region 108, and a third CG region 110. Figure 1 The illustrative example shows a portion of the sequence of the nucleic acid 104 having three CG regions, but the nucleic acid 104 included in the sample 102 may have a different number of CG regions. For example, a single nucleic acid 104 included in the sample 102 may include at least 1 CG region, at least 5 CG regions, at least 10 CG regions, at least 25 CG regions, at least 50 CG regions, at least 100 CG regions, at least 250 CG regions, at least 500 CG regions, or at least 1000 CG regions.
[0197] A single CG region can correspond to many molecules with one or more methylated cytosines. Figure 1 In the illustrative example of , CG region 106 may include molecules having methylated cytosine 112. Figure 1In the illustrative example of , the molecule with methylated cytosine 112 is 5-methylcytosine. A single CG region may also correspond to many molecules with unmethylated cytosine. For example, CG region 106 may include molecules with unmethylated cytosine 116. In various examples, at least a portion of the CG region of nucleic acid 104 may correspond to a classification region of a reference genome. The classification region may correspond to a genomic region of a reference genome corresponding to a non-sequence difference consistent with one or more biological conditions (such as one or more types of cancer). In at least some examples, the non-sequence difference may include one or more mutations consistent with one or more biological conditions. In one or more examples, the classification region may correspond to a genomic region of a molecule of a reference sequence derived from a subject suffering from at least one form of cancer. In at least some examples, a nucleic acid molecule (e.g., a hypermethylated molecule) having at least a threshold amount of methylated cytosine in at least one CG region may be derived from a subject in which cancer is present and corresponds to a classification. In one or more additional examples, a nucleic acid molecule having less than a threshold amount of methylated cytosine (eg, a hypomethylated molecule) in at least one CG region can be derived from a subject in which cancer is present and corresponds to a classification region.
[0198] In addition to the classification region, the CG region may include one or more positive control regions, such as positive control region 118. Positive control region 118 may be mapped to nucleic acid molecules having at least a threshold number of methylated cytosine molecules in at least one CG region and derived from a subject without cancer and from a subject in which cancer is present. In various examples, positive control region 118 may be hypermethylated in cells derived from a subject without cancer and also in cells derived from a subject in which cancer is present. The CG region may also include one or more negative control regions, such as negative control region 120. Negative control region 120 may be mapped to nucleic acid molecules having less than a threshold number of methylated cytosine molecules in at least one CG region and derived from a subject without cancer and from a subject in which cancer is present. In one or more illustrative examples, negative control region 120 may be hypomethylated in a subject without cancer and also in a subject in which cancer is present. In various examples, positive control regions and negative control regions may be used to perform normalization calculations. Normalization calculations may be performed to generate input data for one or more models implemented to determine a tumor metric for a given sample 102 .
[0199] A first molecular separation process 122 may be performed. The first molecular separation process 122 may separate the nucleic acids 104 included in the sample 102 based on the amount of methylated cytosine of the individual nucleic acids 104. In one or more examples, the first molecular separation process may separate the nucleic acids 104 included in the sample 102 based on the amount of methylated cytosine included in the CG region of the individual nucleic acids 104. In various examples, the first molecular separation process 122 may separate the nucleic acids 104 into more than one group, wherein the individual groups correspond to respective amounts of methylated cytosine of the nucleic acids 104.
[0200] exist Figure 1 In the illustrative example of , the first molecule separation process 122 can be performed relative to a first methylation threshold 124. Performing the first molecule separation process 122 for the first methylation threshold 124 can produce a first partition 126 of nucleic acids. In one or more examples, the first methylation threshold 124 can indicate a first threshold number of molecules with methylated cytosine located in a CG region of the nucleic acid 104. The first molecule separation process 122 can identify a number of nucleic acids 104 having fewer molecules with methylated cytosine in the CG region than the first methylation threshold 124. In various examples, the first methylation threshold 124 can correspond to a first methylation rate.
[0201] The first molecular separation process 122 can also be performed relative to a second methylation threshold 128. The second methylation threshold 128 can indicate that the amount of methylated cytosine in one or more genomic regions of the nucleic acid 104 is greater than the amount of methylated cytosine in one or more regions corresponding to the first methylation threshold 124. The second methylation threshold 124 can indicate the number of molecules having methylated cytosine per a certain number of nucleic acids. In one or more additional examples, the second methylation threshold 124 can correspond to a methylation rate of nucleic acids that is greater than the methylation rate corresponding to the first methylation threshold 124. Performing the first molecular separation process 122 relative to the second methylation threshold 128 can produce a second partition 130 of nucleic acids. In one or more examples, the first molecular separation process 122 can identify nucleic acids 104 having a greater amount of methylated cytosine than the first methylation threshold 124 and having a lower amount of methylated cytosine than the second methylation threshold 128 to produce a second partition 130 of nucleic acids.
[0202] Additionally, the first molecular separation process 122 may also be performed relative to a third methylation threshold 132. The third methylation threshold 132 may indicate that the amount of methylated cytosine in one or more genomic regions of the nucleic acid 104 is greater than the amount of methylated cytosine in one or more regions corresponding to the first methylation threshold 124 and greater than the amount of methylated cytosine in one or more regions corresponding to the second methylation threshold 128. The third methylation threshold 132 may indicate the number of molecules having methylated cytosine per a certain number of nucleic acids. In one or more additional examples, the third methylation threshold 132 may correspond to a methylated cytosine rate that is greater than the methylation rate corresponding to the first methylation threshold 124 and greater than the methylation rate corresponding to the second methylation threshold 128. Executing the first molecular separation process 122 relative to the third methylation threshold 132 may produce a third partition 134 of nucleic acids. In one or more examples, the first molecular separation process 122 may identify nucleic acids 104 having a greater amount of methylated cytosine than nucleic acids 104 contained in the second partition of nucleic acids 128. In this manner, the amount of methylated cytosine of the nucleic acids contained in the first partition 122, the second partition 126, and the third partition 130 increases from the first partition 122 to the second partition 126, and increases from the second partition 126 to the third partition 130. In one or more illustrative examples, the first partition of nucleic acids 126 can be referred to as a low methylation partition, the second partition of nucleic acids 130 can be referred to as a medium partition, and the third partition of nucleic acids 134 can be referred to as a high methylation partition.
[0203] In one or more examples, the amount of methylated cytosine of the nucleic acid can correspond to the strength of binding to the methyl binding domain (MBD). In these cases, the first partition 126, the second partition 130, and the third partition 134 can be generated based on the different binding strengths of nucleotides with different amounts of methylated cytosine to the MBD. In one or more examples, the first molecule separation process 122 can include a series of washes in which the nucleic acid 104 is contacted with solutions having different concentrations of sodium chloride (NaCl).
[0204] Partitioning of nucleic acids can be performed by contacting the nucleic acids with a modified nucleotide-specific binding reagent such as MBD of MBP. The modified nucleotide-specific binding reagent can bind to 5-methylcytosine (5mC). The modified nucleotide-specific binding reagent such as MBD can be attached to paramagnetic beads such as M-280 streptavidin coupling. Partitioning into fractions with varying degrees of methylation can be performed by increasing the NaCl concentration in a series of washes. The sequence eluted from the modified nucleotide-specific binding reagent is partitioned into two or more fractions (e.g., low, high) according to the wash (e.g., NaCl concentration) of the elution sequence. The resulting partitions may include one or more of the following nucleic acid forms: double-stranded DNA (dsDNA), shorter DNA fragments, and longer DNA fragments.
[0205] The combination of nucleic acid and modified nucleotide specific binding reagent can change with the number of methylation (or modification) sites of each molecule, wherein the molecule with more methylation is eluted under the increased salt concentration. In order to elute DNA into different populations based on the degree of methylation, people can use a series of elution buffers with increasing NaCl concentration. In one or more embodiments, the salt concentration range can be from about 100nM to about 2500mM NaCl. In various embodiments, the process produces three (3) partitions. The molecule is contacted with a solution of the first salt concentration, and the solution comprises a molecule containing a methyl binding domain, which can be attached to a capture part such as streptavidin. At the first salt concentration, a molecular population will be combined with MBD, and a population will remain unbound. Unbound populations can be separated into "hypomethylation" populations (low partitions). For example, the first partition 126 representing a DNA hypomethylation form can be a partition that remains unbound at a low salt concentration. In one or more illustrative examples, the NaCl concentration of the solution used to generate the first partition 126 can be about 100 nM, about 120 nM, about 140 nM, about 160 nM, about 180 nM, about 200 nM, or about 250 nM. The second partition 130 can be referred to as the "remaining partition" or "intermediate partition" and can represent medium methylated DNA eluted using an intermediate salt concentration (e.g., between 100 mM and 2000 mM concentration). In one or more additional illustrative examples, the NaCl concentration of the solution used to produce the second partition 130 can be about 100mM to about 500mM, about 100mM to about 1000mM, about 100mM to about 1500mM, about 250mM to about 1000mM, about 250mM to about 1500mM, about 500mM to about 1500mM, about 250mM to about 2000mM, about 500mM to about 2000mM, or about 1000mM to about 2000mM. This is also separated from the sample. The third partition 134 can represent a high methylated form of DNA (high partition) and is eluted using a high salt concentration (e.g., at least about 2000mM). In one or more additional illustrative examples, the NaCl concentration of the solution used to create third partition 134 can be about 2000 mM to about 5000 mM, about 2000 mM to about 4000 mM, about 2000 mM to about 3500 mM, about 2000 mM to about 3000 mM, or about 2500 mM to about 4000 mM.
[0206] In various examples, the first partition 126 may correspond to a first range of binding strengths and a first range of methylated CG regions of nucleic acids to MBD, and the second partition 130 may correspond to a second range of binding strengths and a second range of methylated CG regions of nucleic acids to MBD. The first range of binding strengths may be less than the second range of binding strengths. In one or more cases, a first solution having a first NaCl concentration may separate a first group of nucleic acids having a first range of binding strengths from MBD, and a second solution having a second NaCl concentration may separate a second group of nucleic acids having a second range of binding strengths from MBD, wherein the second NaCl concentration is greater than the first NaCl concentration. Additionally, the third partition 134 may correspond to a third range of binding strengths and a third range of methylated CG regions. The third range of binding strengths may be greater than the first range of binding strengths and the second range of binding strengths. In one or more cases, a third solution having a third NaCl concentration may separate a third group of nucleic acids having a third range of binding strengths from NaCl. The third NaCl concentration may be greater than the first NaCl concentration and the second NaCl concentration.
[0207] In one or more illustrative examples, more than one nucleic acid derived from at least one of the blood or tissue of the subject can be combined with a solution comprising an amount of MBD to produce a nucleic acid-MBD solution. A first wash of the nucleic acid-MBD solution can be performed with a first solution comprising a first NaCl concentration to produce a first nucleic acid fraction and a first remaining solution. The first nucleic acid fraction can include a first portion of more than one nucleic acid, and the first remaining solution can include a second portion of more than one nucleic acid. In one or more examples, the first portion of more than one nucleic acid can have a first range of binding strengths with MBD that is less than a second range of binding strengths with MBD of the second portion of more than one nucleic acid.
[0208] Additionally, a second wash of the first remaining solution may be performed with a second solution comprising a second NaCl concentration greater than the first NaCl concentration to produce a second nucleic acid fraction and a second remaining solution. The second nucleic acid fraction may include a first subset of the second portion of more than one nucleic acid, and the second remaining solution may include a second subset of the second portion of more than one nucleic acid. The first subset of the second portion of more than one nucleic acid may have a third range binding strength with MBD that is less than a fourth range binding strength with MBD of the second subset of the second portion of more than one nucleic acid. In addition, a third wash of the second remaining solution may be performed with a third solution comprising a third NaCl concentration greater than the second NaCl concentration to produce a third nucleic acid fraction comprising a second subset of the second portion of more than one nucleic acid.
[0209] After the first wash, the second wash, and the third wash, a determination can be made that the first portion of more than one nucleic acid is associated with the first partition 126. The first portion of more than one nucleic acid can be attached to a molecular barcode from a first group of molecular barcodes indicating the first partition 126. In this way, the sequencing read corresponding to the first partition 126 can be identified based on the determination that the sequencing read includes the first molecular barcode. In addition, a determination can be made that the first subset of the second portion of more than one nucleic acid is associated with another partition in more than one partition. In these cases, a second group of molecular barcodes different from the first group of molecular barcodes can be attached to the second portion of more than one nucleic acid, wherein the second molecular barcode indicates another partition. Therefore, the sequencing read corresponding to the other partition can be identified based on the determination that the sequencing read includes one or more molecular barcodes from the second group of molecular barcodes. In addition, a determination can be made that the second subset of the second portion of more than one nucleic acid is associated with the second partition 130. A third set of molecular barcodes that are different from the first set of molecular barcodes and the second set of molecular barcodes can then be attached to a second subset of the second portions of the more than one nucleic acid, wherein the third set of molecular barcodes is indicative of the second partition 130. In these cases, the sequencing reads corresponding to the second partition 130 can be identified based on determining that the sequencing reads include the third molecular barcode from the third set of molecular barcodes.
[0210] In at least some examples, the first molecular separation process 122 can cause nucleic acids present in at least one of the first partition 126, the second partition 130, or the third partition 134 to have an amount of methylation that is different from the amount of methylation of other nucleic acids in the corresponding partition. For example, the first partition 126 can contain a number of nucleic acids having an amount of methylation corresponding to the amount of methylation of nucleic acids contained in at least one of the second partition 130 or the third partition 134. Additionally, at least one of the second partition 130 or the third partition 134 can contain nucleic acids having an amount of methylation corresponding to the amount of methylation of nucleic acids contained in the first partition 126. When performing computational operations on sequence reads generated from nucleic acids contained in the first partition 126, the second partition 130, and the third partition 134, the presence of nucleic acids in at least one of the first partition 126, the second partition 130, or the third partition 134 that do not correspond to at least the majority of the methylation of other nucleic acids contained in the corresponding partition can result in data noise. Data noise can result in inaccuracies in computations performed based on sequence reads derived from nucleic acids contained in the first partition 126, the second partition 130, and the third partition 134.
[0211] In order to reduce or eliminate data noise associated with nucleic acids present in at least one of the first partition 126, the second partition 130, or the third partition 134 having an amount of methylation inconsistent with the amount of methylation of at least most other molecules contained in the corresponding partition, a second molecular separation process 136 can be performed after the first molecular separation process 122. The second molecular separation process 136 can be performed on nucleic acids contained in the first partition 126, nucleic acids contained in the second partition 130, and nucleic acids contained in the third partition 134. In one or more examples, the second molecular separation process 136 can include digesting the nucleic acids contained in the first partition 126 using a methylation-dependent restriction endonuclease (MDRE), and the nucleic acids contained in the second partition 130 and the third partition 134 can be digested using a methylation-sensitive restriction endonuclease (MSRE). Digesting the nucleic acids contained in the first partition 126 with MDRE can result in the separation of nucleic acids contained in the first partition having an amount of methylation corresponding to the second partition 130 and the third partition 134 from nucleic acids having an amount of methylation corresponding to the first partition. Additionally, digesting the nucleic acids contained in the second partition 130 and the third partition 134 with MSRE can result in the separation of nucleic acids having an amount of methylation corresponding to the first partition 126 from nucleic acids of the second partition 130 and nucleic acids of the third partition 134. By removing nucleic acids having an amount of methylation corresponding to the second partition 130 and the third partition 134 from the first partition 126, and by removing nucleic acids having an amount of methylation corresponding to the first partition 126 from the second partition 130 and the third partition 134, a group 138 of additional nucleic acids can be generated. The group 138 of additional nucleic acids can include nucleic acids having an amount of methylation corresponding to the second partition 130 and the third partition 134, wherein minimal amounts of nucleic acids or no nucleic acids have an amount of methylation corresponding to the first partition 126.For example, less than 50% of the nucleic acids contained in the additional group 138 may have an amount of methylation corresponding to the second partition 130 and the third partition 134, at least 50% of the nucleic acids contained in the additional group 138 may have an amount of methylation corresponding to the second partition 130 and the third partition 134, at least 60% of the nucleic acids contained in the additional group 138 may have an amount of methylation corresponding to the second partition 130 and the third partition 134, at least 70% of the nucleic acids contained in the additional group 138 may have an amount of methylation corresponding to the second partition 130 and the third partition 134, at least 90% of the nucleic acids contained in the additional group 138 may have an amount of methylation corresponding to the second partition 130 and the third partition 134, At least 95% of the nucleic acids contained in the additional group 138 may have an amount of methylation corresponding to the second partition 130 and the third partition 134, at least 97% of the nucleic acids contained in the additional group 138 may have an amount of methylation corresponding to the second partition 130 and the third partition 134, at least 99% of the nucleic acids contained in the additional group 138 may have an amount of methylation corresponding to the second partition 130 and the third partition 134, at least 99.5% of the nucleic acids contained in the additional group 138 may have an amount of methylation corresponding to the second partition 130 and the third partition 134, or at least 99.9% of the nucleic acids contained in the additional group 138 may have an amount of methylation corresponding to the second partition 130 and the third partition 134.
[0212] The architecture 100 may include a sequencing machine 140. In one or more examples, the sequencing machine 140 may be any of a plurality of sequencing machines capable of performing one or more sequencing operations to amplify nucleic acids present in the sample 104. In various examples, the sequencing machine 140 may perform next generation sequencing operations. In one or more examples, the sample 104 may include a quantity of at least one bodily fluid extracted from a subject. In one or more additional examples, the sample 104 may include a tissue sample obtained from a subject.
[0213] In one or more instances, before sequencing, the extracted polynucleotides can be partitioned into two or more partitions based on the binding strength of the polynucleotides to the MBD. Blunt end connection can be performed on the polynucleotides and adapters of the partitions, and labels (e.g., molecular barcodes) can be added to the polynucleotides of the partitions. The tagged polynucleotides in one or more partitions (e.g., high partitions and / or medium partitions) can be processed with one or more methylation-sensitive restriction endonucleases (MSREs). In some instances, low partitions can be processed with one or more methylation-dependent restriction endonucleases (MDREs). After MSRE and / or MDRE processing, molecules can also be enriched by causing hybridization between the extracted polynucleotides and the probes corresponding to the target regions of the reference sequences. The enrichment process can identify thousands, hundreds of thousands, up to millions of polynucleotides, which correspond to the target regions associated with the probes.
[0214] After and / or before the enrichment process, the molecule can be amplified according to one or more amplification processes. One or more amplification processes can produce thousands, up to millions of copies of a single nucleic acid molecule. In one or more instances, in some cases, a portion of the unenriched polynucleotide can be amplified, but can not reach the extent to which the enriched polynucleotide is amplified. One or more amplification processes can produce an amplified product that undergoes one or more sequencing operations. After one or more sequencing operations are performed to the sample 104, the sequencing machine 140 can generate sequencing data 142.
[0215] Sequencing data 142 may include alphanumeric representations of nucleic acids contained in the amplified product. For example, for a single nucleic acid of an amplified product, sequencing data 142 may include data corresponding to a string of letters representing a corresponding nucleotide chain, the letters corresponding to the single nucleic acid.
[0216] Sequencing data 142 can be stored in one or more data files.For example, sequencing data 142 can be stored in a FASTQ file, which includes a text-based sequencing data file format for storing raw sequence data and quality scores.In one or more other instances, sequencing data 142 can be stored in a data file according to a binary base call (BCL) sequence file format.In one or more other instances, sequencing data 142 can be stored in a BAM file.In one or more instances, sequencing data 142 can include at least about 1 gigabyte (GB), at least about 2GB, at least about 3GB, at least about 4GB, at least about 5GB, at least about 8GB or at least about 10GB.The single sequence representation included in sequencing data 106 can be referred to as "read" or "sequencing read" herein.In various instances, as a result of a single first nucleic acid amplification, the single first nucleic acid included in pool 138 can correspond to more than one sequence representation included in sequencing data 142. In one or more additional examples, due to the absence of amplification of the single second nucleic acid, the single second nucleic acid included in pool 138 can correspond to a single sequence representation included in sequencing data 142 .
[0217] Figure 2 2 is an example architecture 200 for analyzing sequencing data to determine one or more metrics indicating the presence of a tumor in a subject, according to one or more embodiments. The architecture 200 may include one or more sequencing machines 202 that perform one or more sequencing operations for a number of samples 204. The one or more samples 204 may be obtained from a subject 206. In one or more illustrative examples, a first portion of the subject 206 may be free of cancer. That is, no tumor was detected in the first portion of the subject 206. Additionally, a tumor may be present in a second portion of the subject 206.
[0218] One or more molecular separation processes 208 can be performed for sample 204. One or more separation processes 208 can correspond to the separation of nucleic acid molecules into many partitions based on the characteristics of nucleic acid molecules. The example of the feature that can be used for the partition of nucleic acid molecules includes more than one different nucleotide modification, methylation level, nucleosome binding, sequence mismatch, immunoprecipitation and / or protein combined with DNA. In one or more illustrative examples, the heterogeneity population partition of nucleic acid molecules can be divided into nucleic acid molecules with one or more epigenetic modifications and without one or more epigenetic modifications. The example of epigenetic modification includes but is not limited to the presence or absence of methylation; the level of methylation, hydroxymethylation and methylation type (5' cytosine or 6 methyl adenine).
[0219] Prior to one or more molecular separation processes 208, nucleic acid molecules can be extracted from sample 204. In one or more embodiments, nucleic acid molecules include cell-free nucleic acids (e.g., cell-free DNA). In various embodiments, sample 204 can be a sample selected from one or more of the following: blood, plasma, serum, urine, feces, saliva samples, combinations thereof, and / or the like. In one or more additional examples, sample 204 can include a sample selected from one or more of the following: whole blood, blood fractions, tissue biopsy, pleural fluid, pericardial fluid, cerebrospinal fluid, and peritoneal fluid. In one or more illustrative examples, cell-free nucleic acid molecules can be extracted from sample 204, wherein sample 204 is obtained from a subject 206 (e.g., a cancer patient) known to have cancer or a subject 206 suspected of having cancer.
[0220] Extracting nucleic acid molecules from sample 204 may include performing one or more cell lysis techniques to lyse cell membranes included in sample 204, and applying one or more proteases to break down proteins included in sample 204. Extracting nucleic acid molecules from sample 204 may also include a number of washing and / or elution techniques to separate nucleic acid molecules from other components included in sample 204. In various examples, thousands, up to millions, up to billions of nucleic acid molecules may be extracted from sample 204 before undergoing one or more separation processes 208.
[0221] The nucleic acid molecules extracted from the sample 204 may include molecules with different methylation levels. Methylation may occur by any one or more post-replication or post-transcriptional modifications. Post-replication modifications include modifications to the nucleotide cytosine, including but not limited to 5-methylcytosine, 5-hydroxymethylcytosine, 5-formylcytosine, and 5-carboxylcytosine. One or more molecular separation processes 208 may separate the nucleic acid molecules extracted from the sample 204 into a plurality of partitions, wherein a single partition corresponds to a different methylation level. For example, the molecular separation process 208 may produce a first partition of nucleic acid molecules having a first methylation level, a second partition of nucleic acid molecules having a second methylation level, and a third partition of nucleic acid molecules having a third methylation level. In various examples, the second methylation level may be greater than the first methylation level, and the third methylation level may be greater than the first methylation level and the second methylation level. In one or more illustrative examples, the one or more molecular separation processes 208 may include Figure 1 The first molecule separation process 122 and the second molecule separation process 136.
[0222] The one or more molecular separation processes 208 can generate a pool 210 that includes a portion of nucleic acid molecules extracted from one or more samples 204 and subjected to the one or more molecular separation processes 208. For example, the pool 210 can include a plurality of nucleic acid molecules having a second methylation level and a plurality of nucleic acid molecules having a third methylation level. Thus, the nucleic acid molecules included in the pool 210 can have at least a threshold amount of methylation. In one or more illustrative examples, the nucleic acid molecules included in the pool 210 can have at least a threshold amount of methylation in a CG region of the nucleic acid molecules.
[0223] One or more sequencing machines 202 may perform one or more sequencing operations to generate sequencing data 212 corresponding to the pool 210. The architecture 200 may include a computing system 214 that obtains the sequencing data 212 from one or more sequencing machines 202 and analyzes the sequencing data 212. For example, the computing system 214 may analyze the sequencing data 212 to determine one or more metrics indicating that a tumor may be present in the subject 206 providing the at least one sample 204. The computing system 214 may include one or more computing devices 216. The one or more computing devices 216 may include at least one of one or more desktop computing devices, one or more mobile computing devices, or one or more server computing devices. In various examples, at least a portion of the one or more computing devices 216 may be included in a remote computing environment, such as a cloud computing environment. In one or more examples, the computing system 214 and the sequencing machine 202 may be owned, operated, maintained, and / or controlled by a single organization. In one or more additional examples, the computing system 214 and the sequencing machine 202 may be owned, operated, maintained, and / or controlled by more than one organization.
[0224] At operation 218, the computing system 214 can analyze the sequencing data 212. Analyzing the sequencing data 212 can include determining one or more first sequence representations 220 of one or more classification regions corresponding to the reference sequence included in the sequencing data 212. One or more classification regions can correspond to a genomic region of the reference sequence mapped to such a nucleic acid molecule, which has a methylation amount in the cfDNA obtained from a subject in which cancer is present relative to the methylation amount of the molecule mapped to the same genomic region of the reference sequence in the cfDNA obtained from a subject in which no tumor is present. In at least some instances, the amount of methylation present in the nucleic acid molecule mapped to the classification region and derived from the subject in which cancer is present is less than the amount of methylation present in the nucleic acid molecule mapped to the classification region and derived from the subject in which cancer is not present. In one or more additional instances, the amount of methylation present in the nucleic acid molecule mapped to the classification region and derived from the subject in which cancer is present is greater than the amount of methylation present in the nucleic acid molecule mapped to the classification region and derived from the subject in which cancer is not present. One or more classification regions can also include at least a threshold amount of cytosine-guanine content. In various instances, one or more classification regions can include a series of cytosine-guanine (CG) pairs (CpG sites) in the 5'→3' direction, such as at least 3 CpG sites, at least 5 CpG sites, at least 8 CpG sites, at least 10 CpG sites, at least 12 CpG sites, at least 15 CpG sites, at least 18 CpG sites, or at least 20 CpG sites.
[0225] In addition, the computing system 214 can analyze the sequencing data 212 to determine one or more second sequence representations 222 corresponding to one or more control regions of the reference sequence. One or more control regions may include one or more positive control regions and / or one or more negative control regions. In various examples, the positive control region may include a genomic region of the reference sequence having at least a threshold amount of molecules with methylated cytosine and including at least a threshold number of CpG sites. The positive control region may correspond to a nucleic acid molecule having at least a threshold amount of methylation in one or more CG regions and obtained from a subject in which cancer is present and in a sample obtained from a subject in which no tumor is present. In at least some examples, the threshold amount of methylation may correspond to at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 11, at least 12, at least 15 or more CpGs methylated in the nucleic acid molecule. In one or more illustrative examples, the positive control region may be mapped to a nucleic acid molecule that is highly methylated in one or more CG regions and is derived from a sample obtained from a subject in which cancer is present and a subject in which no cancer is present. In one or more instances, the negative control region may include a genomic region of a reference sequence having less than a threshold amount of molecules with methylated cytosine and at least a threshold number of CpG sites. The negative control region may correspond to a nucleic acid molecule having less than an additional threshold amount of methylation in one or more CG regions and obtained from a subject in which cancer is present and in a sample obtained from a subject in which a tumor is not present. In various instances, the additional threshold amount of methylation may correspond to no more than 1, no more than 2, no more than 3, no more than 4, no more than 5, no more than 6, or no more than 7 CpGs that are methylated in the nucleic acid molecule. In one or more additional illustrative instances, the negative control region may be mapped to a nucleic acid molecule that is hypomethylated in one or more CG regions and is derived from a sample obtained from a subject in which cancer is present and a subject in which cancer is not present.
[0226] In one or more illustrative examples, the first sequence representation 220 can be determined by aligning the sequence representation included in the sequencing data 212 with one or more classification regions of the reference sequence. Additionally, the second sequence representation 222 can be determined by aligning the sequence representation included in the sequencing data 212 with one or more control regions of the reference sequence. The alignment process can identify the first sequence representation 220 by determining the number of sequence representations included in the sequencing data 212 that correspond to the one or more classification regions of the reference sequence. Additionally, the alignment process can identify the second sequence representation 222 by determining the number of sequence representations that correspond to the one or more control regions of the reference sequence.
[0227] In one or more illustrative examples, the alignment process can determine an amount of homology between a single sequence representation included in sequence data 212 and a portion of a reference sequence. The amount of homology between a given sequence representation and a reference sequence can indicate the number of positions in the reference sequence that have the same nucleotide as a corresponding position of the given sequence representation. The computing system 214 can determine that the sequence representation is aligned with a portion of the reference sequence based on determining that the sequence representation and the portion of the reference sequence have at least a threshold amount of homology. In the event that the sequence representation has at least a threshold amount of homology with respect to more than one portion of the reference sequence, it can be determined that the portion of the reference sequence that has the greatest amount of homology with the sequence representation is aligned with the sequence representation.
[0228] The amount of homology between a given sequence representation and a portion of a reference sequence can be determined by using the BLAST program (Basic Local Alignment Search Tool) and the PowerBLAST program (Altschul et al., J. Mol. Biol., 1990, 215, 403-410; Zhang and Madden, Genome Res., 1997, 7, 649-656) or by using the Gap program (Wisconsin Sequence Analysis Package, Genetics Computer Group, University Research Park, Madison Wis.), using the default settings, which uses the algorithm of Needleman and Wunsch (J. Mol. Biol. 48; 443-453 (1970)). The amount of homology between a sequence representation and a portion of a reference sequence can also be determined using the Burrows-Wheeler aligner (Li, H., & Durbin, R. (2009). Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics, 25(14), 1754-1760).
[0229] In one or more instances, after the sequence representations included in the sequencing data 212 have been aligned with the reference sequence, the aligned sequence representations can be analyzed to identify one or more groups of sequence representations. For example, a single aligned sequence representation can correspond to a single sequencing read included in the sequencing data 212. In these cases, the aligned sequence representations can include more than one read corresponding to a single nucleic acid molecule included in the sample pool 210. In one or more additional instances, the aligned sequence representations can correspond to a single nucleic acid molecule included in the pool 210. In these cases, the computing system can determine the read group corresponding to the single nucleic acid molecule included in the pool 210 included in the sequence data 212 based on the molecular barcodes common to each group of sequencing reads. That is, the single nucleic acid molecule included in the pool 210 can be encoded with a molecular barcode that uniquely identifies the single nucleic acid molecule, and in at least some cases, the single nucleic acid molecule can be represented by more than one sequencing read included in the sequencing data 212. Therefore, when there are more than one sequence representations in the sequencing data 212 corresponding to a single nucleic acid molecule included in the pool 210, the computing system 214 can group the more than one sequence representations together. In various examples, the group of sequence representations corresponding to individual nucleic acid molecules included in pool 210 may be referred to herein as a "family." Additionally, the start and stop positions of aligned sequence representations having a common molecular barcode relative to a reference sequence may be used to group sequence representations corresponding to individual nucleic acids included in pool 210. In one or more illustrative examples, a single sequence representation representing a family of sequence representations corresponding to individual nucleic acid molecules included in pool 210 may be referred to herein as a "consensus sequence representation."
[0230] At operation 224, the computing system 214 may analyze the first sequence representation and the second sequence representation 222 to generate a metric corresponding to a single classification region. Figure 2In the illustrative example of , the computing system 214 can analyze the first sequence representation 220 and the second sequence representation 222 to generate a classification region metric 226. The classification region metric 226 can include a quantitative measure determined based on the number of first sequence representations 220 having at least a threshold amount of methylated cytosines. In one or more illustrative examples, the classification region metric 226 can include a quantitative measure determined based on the number of sequencing reads corresponding to the number of first sequence representations 220 with at least a threshold amount of methylated cytosines located. In one or more additional illustrative examples, the classification region metric 226 can include a quantitative measure determined based on the number of nucleic acid molecules corresponding to the number of first sequence representations 220. In various examples, the classification region metric 226 can include a quantitative measure determined based on the number of first sequence representations 220 having at least a threshold amount of methylated cytosines and the number of second sequence representations 222 corresponding to a control region of a reference sequence. In one or more additional illustrative examples, classification region metric 226 may include a quantitative measure associated with a ratio of the number of first sequence representations 220 having at least a threshold amount of methylated cytosines relative to the number of second sequence representations. In at least some examples, the sequence representation of second sequence representation 222 used by computing system 214 to generate the quantitative measure included in classification region metric 226 may include a sequence representation corresponding to a positive control region of a reference sequence.
[0231] The classification region metric 226 may also be determined by performing one or more normalization operations relative to a quantitative metric generated by the computing system 214 using at least one of the first sequence representation 220 and the second sequence representation 222. For example, a logarithmic calculation may be performed relative to a quantitative metric generated by the computing system 214 using at least one of the first sequence representation 220 or the second sequence representation 222. Additionally, the classification region metric 226 may be determined by adding pseudo counts to a quantitative metric determined by the computing system 214 using at least one of the first sequence representation 220 or the second sequence representation 222. In one or more illustrative examples, the one or more normalization operations may include determining a quantitative metric of a ratio of the number of first sequence representations 220 corresponding to a single classification region relative to the number of second sequence representations 222 corresponding to a positive control region of a reference sequence.
[0232] In one or more illustrative examples, the computing system 214 can determine the number of first sequence representations 220 corresponding to a single classification region of a reference sequence and having at least a threshold amount of methylated cytosines located in the single classification region. In these cases, the computing system 214 can determine a single classification region metric 226 for the single classification region. In addition, the computing system 214 can determine the number of second sequence representations 222 corresponding to the positive control region. In at least some examples, the computing system 214 can determine, for the single classification region, a ratio of the number of first sequence representations 220 including molecules having at least a threshold amount of methylated cytosines in the classification region corresponding to the single classification region relative to the total number of second sequence representations 222 corresponding to the positive control region of the reference sequence. In one or more examples, the computing system 214 can add a value of a pseudo count to the ratio to determine the classification region metric 226 for the single classification region. The value of the pseudo count can be at least 1, at least 1.2, at least 1.4, at least 1.6, at least 1.8, or at least 2. Additionally, computing system 214 may perform a logarithm base 10 operation on the combination of the ratio and the pseudo count to determine a classification region metric 226 for the individual classification regions. In at least some illustrative examples, computing system 214 may determine at least a portion of the classification region metric according to the following equation:
[0233]
[0234] where x i is the total number of first sequence representations 220 of a single classification region i having at least a threshold amount of methylated cytosines included in region i, and x 阳性_对照 is the total number of second sequence representations 222 corresponding to the positive control region of the reference sequence.
[0235] At operation 228, the computing system 214 may execute the model to determine an indication of cancer based on the classified region metrics 226. Figure 2 In the illustrative example of , computing system 214 can execute the model using classification region metrics 226 to generate model output 230. In one or more examples, model output 230 can indicate a tumor detected state 232 or a tumor not detected state 234 associated with sample 204 provided by subject 206. In one or more additional examples, computing system 214 can execute the model to determine an estimate of a tumor score 236 for sample 204. In one or more additional examples, computing system 214 can execute the model to determine a probability that a tumor is present in subject 206 providing sample 204.
[0236] In one or more instances, the model may include a classification model that implements one or more machine learning techniques. In one or more illustrative instances, the model may include a linear regression model. In various instances, the model may be executed to determine the probability 238 of the presence of a tumor in a subject 206 providing a sample 204 based on the classification region metric 226. In one or more illustrative instances, the computing system 214 may execute the model to determine the weight of a single classification region. The weights of a single classification region may be different. For example, the computing system 214 may determine that the first weight of the first classification region metric 226 of the first classification region is different from the second weight of the second classification region metric 226 of the second classification region. In at least some illustrative instances, the probability 238 of the presence of a tumor in a subject 206 providing a sample 204 may be determined by the computing system 214 by executing a model corresponding to the following equation:
[0237]
[0238] where w i is the weight of a single classification region, the score of region i is calculated using equation (1), and b is the slope corresponding to the linear regression model. In at least some instances, the probability 238 of the presence of a tumor can be used to generate a tumor detected state 232 or a tumor not detected state 234. In one or more additional illustrative examples, the computing system 214 can analyze the probability 238 of the presence of a tumor relative to a threshold probability to determine the tumor detected state 232 or the tumor not detected state 234 of the sample 204. In response to determining that the probability 238 of the presence of a tumor in the sample 204 is at least the threshold probability, the computing system 214 can determine that the sample 204 corresponds to the tumor detected state 232. Additionally, in response to determining that the probability 238 of the presence of a tumor in the sample 204 is less than the threshold probability, the computing system 214 can determine that the sample 204 corresponds to the tumor not detected state 234.
[0239] In one or more other examples, computing system 214 can perform a model for determining maximum mutant allele fraction (MAF). In various examples, computing system 214 can use maximum MAF value to perform model to determine the tumor score 236 of sample 204. In one or more illustrative examples, computing system 214 can use classification region metric 226 to perform model to determine the maximum MAF value of logit conversion, and then computing system 214 can use this maximum MAF value to estimate the tumor score of sample 204. In various examples, computing system 214 can analyze maximum MAF value to determine the probability 238 of cancer in the subject 206 providing sample 204. In various examples, Huber regression (Huber, PJ1964. "Robust Estimation of a Location Parameter." Annals of MathematicalStatistics35 (1): 73-101) can be performed to determine maximum MAF value based on classification region metric 226.
[0240] In various examples, the model output 230 may also include a tumor tissue indication 240. The tumor tissue indication 240 may indicate one or more tissues from which the cancer cells that produced the genomic material detected in the sample 204 originated. In one or more examples, the tumor tissue indication 240 may correspond to one or more source tissues from which the cancer cells that produced the genomic material detected in the sample 204 originated. In these cases, the computing system 214 may generate more than one model, where a single model corresponds to a given tissue type. The output from the single model may be analyzed to determine additional metrics indicating the tissue from which the cancer cells that produced the genomic material detected in one or more samples originated. In at least some examples, the output of the single model may indicate at least one of a tumor score 236 or a probability of the presence of a tumor 238. The computing system 214 may analyze the corresponding model output to determine that the model has at least one of the maximum value of the tumor score or the maximum probability of the presence of cancer. Then, the computing system 214 may generate a tumor tissue indication 240 corresponding to the model with the maximum value of the tumor score and / or the maximum probability of the presence of cancer.
[0241] For example, sample 204 can be obtained from a subject 206 in which different types of cancer are present. For illustration, a first sample can be obtained from a first group of subjects in which a first classification of cancer is present, and a second sample can be obtained from a second group of subjects in which a second classification of cancer is present. The sequencing data generated from the first sample can be analyzed by a computing system 214 to generate a first metric corresponding to a classification region of the first classification of cancer, and the first metric can be used to generate a first model corresponding to the first classification of cancer. Additionally, the sequencing data generated from the second sample can be analyzed by a computing system 214 to generate a second metric corresponding to a classification region of the first classification of cancer, and the second metric can be used to generate a second model corresponding to the second classification of cancer. After the model has been trained, the computing system 214 can analyze the sequencing data obtained from one or more additional subjects that are not included in the training subject to determine the classification region metric of one or more additional subjects. The classification region metric can then be analyzed using different tumor classification models to generate model outputs. The model output can be analyzed by a computing system to determine a model with a maximum value of a corresponding model output, and to determine a tumor tissue classification corresponding to the model.
[0242] In one or more additional illustrative embodiments, the model output 230 may also indicate the methylation state of one or more genomic regions of the reference sequence. For example, the computing system 214 may analyze the classification region metric 226 to determine the methylation state of one or more promoter regions of the reference sequence. In various instances, the one or more promoter regions may include at least one promoter region associated with the presence of a tumor in a subject. In one or more illustrative instances, the classification region metric 226 may indicate the number of sequence representations with at least a threshold amount of methylation relative to the one or more promoter regions. In these cases, the computing system 214 may determine that the promoter region is methylated in response to determining that the number of sequence representations of molecules with at least a threshold amount of methylated cytosine in the promoter region is greater than the threshold number.
[0243] In other illustrative embodiments, computing system 214 can analyze sequencing data 212 to determine the quantitative measure corresponding to the number of sequence representations corresponding to the promoter region, and determine the other quantitative measure corresponding to the other number of sequence representations corresponding to the number of positive control regions. Normalization metrics can be determined based on quantitative measures and other quantitative measures. Normalization metrics can be analyzed relative to threshold values to determine the methylation state of the promoter region. In various examples, for different promoter regions, the threshold value can be different. The threshold value can be determined using a cancer-free data set, and the threshold value is set so that the false positive rate is not more than 5%, not more than 4%, not more than 3%, not more than 2% or not more than 1%, and the specificity is at least 80%, at least 90%, at least 95%, at least 98% or at least 99%. In at least some instances, computing system 214 may analyze a promoter region relative to a threshold value where at least 4 sequence indications correspond to a promoter region, where at least 5 sequence indications correspond to a promoter region, where at least 6 sequence indications correspond to a promoter region, where at least 7 sequence indications correspond to a promoter region, where at least 8 sequence indications correspond to a promoter region, where at least 9 sequence indications correspond to a promoter region, where at least 10 sequence indications correspond to a promoter region, where at least 11 sequence indications correspond to a promoter region, or where at least 12 sequence indications correspond to a promoter region.
[0244] In one or more additional illustrative examples, the computing system 214 can combine the results from more than one model to determine the model output 230. For example, the computing system 214 can execute a model about one or more epigenetic signals (such as methylation of a classification region) to determine one or more first tumor metrics. Regarding methylation, the computing system 214 can execute both a classification model (such as a logistic regression model) that generates an indication of the presence of cancer in a subject providing a sample and another model that predicts the tumor score of the sample. In one or more instances, the epigenetic signal can also correspond to the fragment length of the sequence representation generated from the sample. In addition, the computing system 214 can execute one or more additional models for the genomic signal to generate additional tumor metrics for the sample. In various instances, the genomic signal can correspond to the presence and / or insertion or deletion of one or more single nucleotide variants (SNVs) at one or more genomic regions of the reference sequence. In at least some instances, the computing system may include an integrated system that combines tumor metrics generated by executing many models corresponding to data of genomic signals and epigenetic signals to generate an aggregate tumor metric for a given sample. In some embodiments, the quantitative measure obtained from the model output 230 can be used to analyze relative to a threshold value. In the case where the quantitative measure obtained from the model output is at least the threshold value, the computing system 214 can determine an indication of the presence of cancer in the subject. In the case where the quantitative measure obtained from the model output is less than the threshold value, the computing system 214 can determine an indication that cancer is not present or not detected in the subject. In some embodiments, the threshold value for determining the presence or absence of an indication of cancer is calculated using a set of normal samples and is set to a specific value that provides high specificity.
[0245] In various additional embodiments, the computing system 214 can determine the methylation state of a single genomic region. In one or more illustrative examples, the computing system 214 can determine the methylation state of one or more promoter regions. In one or more examples, the sequencing data 212 can be analyzed to determine sequence representations corresponding to the one or more genomic regions. For example, the sequencing data 212 can be analyzed to determine the number of sequence representations corresponding to the one or more promoter regions. In at least some examples, the computing system 214 can determine the number of sequence representations corresponding to a single promoter region having at least a threshold amount of methylated cytosines.
[0246] For each genomic region and a single sample, the computing system 214 can determine the number of sequence representations corresponding to polynucleotide molecules having at least a threshold number of methylated cytosines in the genomic region. The computing system 214 can use the counts of polynucleotide molecules or sequence reads corresponding to the genomic region and having at least a threshold number of methylated cytosines to perform one or more normalization operations to generate a normalized metric. For illustration, the computing system 214 can divide the counts of polynucleotide molecules or reads corresponding to the genomic region and having at least a threshold number of methylated cytosines by the number of molecules or sequencing reads corresponding to the control region (such as a positive control region). In another example, the computing system 214 can divide the counts of polynucleotide molecules or reads corresponding to the genomic region and having at least a threshold number of methylated cytosines by the number of molecules or sequencing reads corresponding to the same genomic region and having at least the same threshold number of methylated cytosines in the control data set (i.e., the control data set includes a sample in which no tumor is detected) to perform a normalized metric.
[0247] The normalized metric can be analyzed relative to a threshold. The threshold can correspond to a given genomic region, such as a given promoter region. In various instances, the threshold can be different for different promoter regions. In these cases, a first promoter region can have a first threshold, and a second promoter region can have a second threshold. In the case where the normalized metric is at least the threshold, the computing system 214 can determine that the genomic region has a first methylation state. In the case where the normalized metric is less than the threshold, the computing system 214 can determine that the genomic region has a second methylation state. In one or more illustrative examples, the first methylation state can be labeled "methylated", and the second methylation state can be labeled "unmethylated".
[0248] The threshold for a given genomic region can be determined based on training data obtained from samples of individuals in which cancer was not detected. In one or more examples, the sequence representation obtained from the training samples can be analyzed to determine a z-score relative to the number of polynucleotide molecules corresponding to the genomic region and having at least a threshold amount of methylated cytosine. In one or more illustrative examples, the threshold for a promoter region used to determine a normalized metric for a promoter region can be obtained from a z-score calculated based on the training samples for the promoter region.
[0249] although Figure 2The illustrative examples of describe that a model can be generated to determine the number of indicators about the presence or absence of cancer in a given subject, but in at least some other examples, the sequencing data 212 can be analyzed by the computing system 214 to determine the indicators of the presence of cancer without training a specific model. In one or more examples, the computing system 214 can determine the tumor score value based on the sequencing data 212 generated from one or more samples obtained from a single subject in which the presence or absence of cancer in the subject is unknown. In one or more examples, the computing system 214 can determine the change in the tumor score value based on the sequencing data 212 generated from one or more samples obtained from a single subject at two or more time points. The change in the tumor score value can be used to monitor the subject's response to treatment. In one or more other examples, a first sample can be obtained from a subject before or at the beginning of at least one administration of a treatment or procedure related to cancer, and one or more second samples can be obtained from a subject after at least one administration of a treatment or procedure related to cancer. In one or more illustrative examples, the one or more second samples can be obtained at least one week, at least two weeks, at least three weeks, at least four weeks, at least five weeks, at least six weeks, at least eight weeks, or at least ten weeks after the treatment or procedure is administered. In at least some examples, the first sample and the second sample can be derived from at least one of a bodily fluid obtained from a subject or a tissue obtained from a subject.
[0250] In one or more instances, one or more samples can be obtained from a given subject. The sequencing data 212 generated from one or more samples can be analyzed by a computing system to determine the quantitative measure of more than one classification region. In one or more instances, the quantitative measure can correspond to the amount of sequence representations with at least a threshold amount of overlap with one or more classification regions. In one or more other instances, the quantitative measure can correspond to the sequence representation of at least a threshold amount of methylated cytosine in a CpG region with at least a threshold amount of CG content. In various instances, the indication of the presence of cancer in the subject can include a tumor score. In one or more other instances, the indication of the presence of cancer in the subject can include a mutant allele score. In at least some instances, relative to the total number of sequencing reads spanning more than one positive control region, the quantitative measure can correspond to the number of sequencing reads corresponding to a given classification region. In one or more other instances, the indicator of the presence of cancer can be used to determine the output corresponding to the presence or absence of cancer in a given individual in response to analyzing one or more indicators of the presence of cancer relative to one or more threshold values. In one or more illustrative examples, a tumor score determined from one or more samples obtained from a subject may be analyzed relative to one or more thresholds. In the case where the tumor score is greater than a threshold level, the computing system 214 may determine that the probability of the presence of cancer in the subject is at least 80%, at least 85%, at least 90%, at least 95%, or at least 99%. In addition, in the case where more than one sample is obtained from a subject, a first quantitative measure generated by a first sample obtained from a subject may be analyzed relative to a second quantitative measure generated by a second sample obtained from the subject. In at least some instances, the difference between the first quantitative measure and the second quantitative measure may be analyzed to determine an indication of a treatment response in the subject.
[0251] The quantitative measure for determining that there is an indication of cancer in a subject can be determined by analyzing the quantitative measure of the subset of the classification region. In at least some instances, the subset of the classification region can be different for different subjects. In one or more illustrative examples, the value of the quantitative measure of the number of classification regions can be analyzed relative to each other, and sorted according to the magnitude of the quantitative measure value. In various instances, the classification regions of a given sample can be sorted in descending order from one or more classification regions with the maximum quantitative measure value to one or more classification regions with the minimum quantitative measure value.
[0252] In various instances, after the quantitative measures of the classification regions are sorted, the quantitative measures corresponding to a group of classification regions can be removed before determining the indication of the presence of cancer in the subject. For example, the classification region group that is not used to determine the indication of the presence of cancer in the subject can include 1% of the classification regions with the maximum quantitative measure value, 2% of the classification regions with the maximum quantitative measure value, 3% of the classification regions with the maximum quantitative measure value, 4% of the classification regions with the maximum quantitative measure value, 5% of the classification regions with the maximum quantitative measure value, or 6% of the classification regions with the maximum quantitative measure value. In at least some instances, many classification regions with relatively high quantitative measure values can be excluded from the classification region group used to determine the indication of the presence of cancer in the subject, because in at least some cases, the classification regions corresponding to the quantitative measure values at or near the top of the sorted list can have non-tumor origins and / or be associated with sequencing artifacts. Therefore, by removing the quantitative measures corresponding to these classification regions from the analysis used to determine the indication of the presence of cancer in the subject, the accuracy of the indication of the presence of cancer in the subject can be increased.
[0253] In one or more instances, after determining a group of classification regions that are indicative of the presence of cancer in a subject, a subset of the group of classification regions can then be determined by identifying at least 10 classification regions of the group, at least 25 classification regions of the group, at least 50 classification regions of the group, at least 75 classification regions of the group, at least 100 classification regions of the group, at least 150 classification regions of the group, at least 200 classification regions of the group, at least 250 classification regions of the group, at least 300 classification regions of the group, at least 350 classification regions of the group, at least 400 classification regions of the group, at least 450 classification regions of the group, or at least 500 classification regions of the group having maximum values of corresponding quantitative measures.
[0254] In at least some instances, one or more statistical measures (such as at least one of the mean, median or mode) can be applied to the quantitative measure of the subset of the classification region of the group to generate an initial indication of the presence of cancer in the subject. In various instances, the initial indication of cancer can be modified according to a scaling factor. The scaling factor can be applied to the initial indication of the presence of cancer in the subject because, in at least some cases, the positive control region can have different amounts of methylated CpG. For example, at least a portion of the positive control region can have a fully methylated CpG, while other positive control regions may not be fully methylated. Additionally, in various cases, some classification regions may correspond to high indication values for the presence of cancer in the subject, such as 90% tumor scores, 95% tumor scores, 99% tumor scores, or 100% tumor scores, but the nucleic acid molecules corresponding to these classification regions may not be fully methylated. To illustrate these situations, a scaling factor can be applied to the initial indication of the presence of cancer in the subject to provide a more accurate determination of the indication. In one or more illustrative examples, the scaling factor can be determined by analyzing an indication of the presence of cancer in a subject determined using one or more techniques described herein, which is correlated with additional data corresponding to additional indications of the presence of cancer in the subject (such as validation data or other techniques that generate data orthogonal to the indication of the presence of a tumor in a subject described herein).
[0255] In various examples, the classification region for determining the quantitative measure can correspond to the classification region corresponding to one or more parts of the differentially methylated region. In one or more examples, the differentially methylated region may include a promoter region corresponding to one or more classifications of cancer. For example, the classification region can be determined by analyzing the number of sequencing representations across the differentially methylated region. In these cases, one or more parts of the differentially methylated region overlapping with at least a threshold number of sequencing representations may be included in the classification region. In one or more examples, the quantitative measure of one or more parts of the differentially methylated region can be determined based on the molecular count distribution of the differentially methylated region. For example, the quantitative measure can be determined based on the molecular counts within one or more peaks of the molecular distribution of the differentially methylated region. To illustrate, in various examples, the distribution of molecules across the differentially methylated region can indicate one or more peaks, wherein a larger amount of molecules overlaps with one or more sub-regions within the differentially methylated region. In various examples, one or more genomic regions corresponding to one or more sub-regions of the differentially methylated region corresponding to the highest amount of sequence representations of the sample can be defined as classification regions. In at least some instances, the distribution of sequence representations can have a peak corresponding to a sub-region of a differentially methylated region having a higher number of sequence representations than other sub-regions of the differentially methylated region. In these cases, the sub-region can be identified as a classification region. By determining sub-regions of at least a portion of a differentially methylated region for determining an indication of the presence of cancer in a subject, the amount of computational resources and memory resources used to determine an indication of the presence of cancer in a subject can be reduced.
[0256] To illustrate, a classification region may include one or more portions of a differentially methylated region in which at least 50% of the sequencing representations obtained from the sample overlap, at least 55% of the sequencing representations obtained from the sample overlap, at least 60% of the sequencing representations obtained from the sample overlap, at least 65% of the sequencing representations obtained from the sample overlap, at least 70% of the sequencing representations obtained from the sample overlap, at least 75% of the sequencing representations obtained from the sample overlap, at least 85% of the sequencing representations obtained from the sample overlap, at least 90% of the sequencing representations obtained from the sample overlap, at least 95% of the sequencing representations obtained from the sample overlap, or at least 99% of the sequencing representations obtained from the sample overlap. In one or more illustrative examples, one or more portions of a differentially methylated region including a classification region may be contiguous relative to a reference sequence.
[0257] Figure 3300 is a schematic diagram of an example framework 300 for training a computational model 302 to determine one or more tumor metrics about a sample according to one or more embodiments. The framework 300 may include a computing system 214. The computing system 214 may execute the computational model 302 to generate one or more model outputs 304. In one or more instances, the computational model 302 may be a machine learning model. The model output 304 may include an indication of the presence or absence of a tumor in a subject corresponding to the sample. In one or more illustrative instances, the model output 304 may include a tumor score. In one or more other illustrative instances, the model output 304 may include a probability of the presence of cancer in the subject. In one or more other illustrative instances, the model output may include an indication of the presence of cancer in the subject or an indication of the absence of cancer in the subject. In other illustrative instances, the model output 304 may indicate the methylation state of one or more regions of a nucleic acid molecule. For illustration, the computing system 214 may execute the computational model 302 for a quantitative measure corresponding to a promoter region to determine the amount of methylation of the promoter region. In other illustrative instances, the model output 304 may include a tumor tissue indication of the sample.
[0258] The framework 300 may also include a sequence representation 306. In one or more instances, the sequence representation 306 may be generated based on analyzing nucleic acid molecules derived from a sample provided by a subject. The sequence representation 306 may include a genomic region having a number of nucleotides corresponding to the number of regions of interest. For example, the sequence representation 306 may include a nucleotide sequence corresponding to a first classification region 308. In addition, the sequence representation 306 may include a nucleotide sequence corresponding to a second classification region 310. In addition, the sequence representation 306 may include a nucleotide sequence corresponding to a third classification region 312. In various instances, the first classification region 308, the second classification region 310, and the third classification region 312 of the sequence representation 306 may have different amounts of methylated cytosines included in the corresponding classification regions 308, 310, and 312. In one or more additional instances, the sequence representation 306 may include a nucleotide sequence corresponding to a positive control region 314 and a nucleotide sequence corresponding to a negative control region 316.
[0259] The computational model 302 may include many components corresponding to a single classification region. In one or more instances, the components of the computational model 302 may have corresponding values corresponding to the quantitative metrics of the corresponding classification regions. The quantitative metrics may indicate the number of sequence representations corresponding to the corresponding classification regions. In one or more instances, the computational model 302 may include many weights associated with the corresponding components of the computational model 302. For example, the computational model 302 may include a first model component 318 having a first weight 320. The first model component 318 may correspond to the first classification region 308. In addition, the computational model 302 may include a second model component 322 having a second weight 324. The second model component 322 may correspond to the second classification region 310. In various instances, at least one of the first weight 320, the second weight 324, or the third weight 328 may be different from at least another of the first weight 320, the second weight 324, or the third weight 328.
[0260] In one or more illustrative examples, the values of the first model component 318, the second model component 322, and the third model component 326 can be determined based on each sample. To illustrate, for different samples, the computational model 302 can determine different values of at least one of the first model component 318, the second model component 322, or the third model component 326. In various examples, the computing system 214 can determine the first quantitative measure of the first classification region 308 based on the sequencing data of the sample. The computing system 214 can execute the computational model 302 to determine the value of the first model component 318 based on the first quantitative measure. Additionally, the computing system 214 can determine the second quantitative measure of the second classification region 310 based on the sequencing data of the sample. The computing system 214 can execute the computational model 302 to determine the value of the second model component 322 based on the second quantitative measure. In addition, the computing system 214 can determine the third quantitative measure of the third classification region 312 based on the sequencing data of the sample. The computing system 214 can execute the computational model 302 to determine the value of the third model component 326 based on the third quantitative measure. The first quantitative measure, the second quantitative measure, and the third quantitative measure can be determined based on the number of sequence representations having at least a threshold amount of methylation in the CG regions corresponding to the first classification region 308, the second classification region 310, and the third classification region 312, respectively. In one or more additional illustrative examples, the value of the first weight 320, the value of the second weight 324, and the value of the third weight 328 can be determined on a per-sample basis. For example, the computational model 302 can determine different values for at least one of the first weight 320, the second weight 324, or the third weight 328 for different samples.
[0261] In one or more instances, the computing system 214 can perform a training process to generate the computing model 302. In various instances, the training process can determine one or more features associated with a classification region metric that can be used to determine the model output 304. Additionally, the training process can determine one or more parameters associated with the classification region metric that can be used to determine the model output 304. For example, the training process can be used to determine model components to include in the computing model 302 and corresponding weights of the model components.
[0262] exist Figure 3 In the illustrative example of, the training process can be performed using training data 330. The training data 330 can include information obtained about at least a first group of subjects 332 and information obtained about at least a second group of subjects 334. In one or more instances, the first group of subjects 332 can include subjects in which no tumors are detected, and the second group of subjects 334 can include subjects in which tumors are detected. In various instances, the training data 330 can include characteristics related to the amount of methylation of the classification regions of the reference sequences 306 of the first group of subjects 332 and the second group of subjects 334. For example, the training data 330 can indicate a quantitative measure of the number of sequence representations corresponding to the classification regions 308, 310, 312 of the first group of subjects 332 and the second group of subjects 334 having at least a threshold level of methylation. The training data 330 can also include weights of model components based on analysis of sequencing data of the first group of subjects 332 and the second group of subjects 334. In one or more illustrative examples, the training data 330 may include values of a first weight 320 , a second weight 324 , and a third weight 328 for a classification region metric determined based on sequencing data obtained from samples provided from a first group of subjects 332 and a second group of subjects 334 .
[0263] Training data 330 may also include information corresponding to additional characteristics of first group of subjects 332 and second group of subjects 334. To illustrate, training data 330 may include medical record information, medical history information, cancer treatment history information, demographic information, genomic information, one or more combinations thereof, and the like.
[0264] In one or more instances, computing system 214 can train computing model 302 to determine the indication related to one or more types of cancer present in an individual. Additionally, in various instances, computing model 302 can include more than one different model so that computing model 302 is a collection model. In these cases, computing system 214 can perform one or more training processes for a single model of the collection model. In one or more illustrative examples, computing model 302 can include many single models that correspond to the output of a model that determines a single genomic region (such as a gene or a specific genome). For example, computing model 302 can include many single models to generate the maximum MAF value of a single gene or a specific genome. In these cases, due to using genome information in the training process, the training of computing model 302 can have more restrictions than other models for determining the indication of cancer present in an individual. As a result, in the case of using genome information to train computing model 302, the accuracy of the output of computing model 302 can be increased. Furthermore, while training of computational model 302 may incorporate genomic information to determine maximum MAF values, using computational model 302 to determine an indication of cancer in a test subject following training may be performed without the use of genomic information and may be based on an input vector corresponding to a quantitative measure determined from sequencing data obtained from the test subject.
[0265] The computing system 214 may obtain a first training data set 336 up to an Nth training data set 338 to perform a training process to generate the computational model 302. In one or more instances, the first training data set 336 may include a first portion of the training data 330 corresponding to the first group of subjects 332 and the second group of subjects 334 for training the computational model 302, and the Nth training data set 338 may include a second portion of the training data 330 corresponding to the first group of subjects 332 and the second group of subjects 334 as part of the validation process of the computational model 302. In various instances, the computational model 302 may be updated over time and undergo more than one training process. In these cases, the first training data set 336 may include a portion of the training data 330 corresponding to the first group of subjects 332 and the second group of subjects 334 for a first time period, and the Nth training data set 338 may include a portion of the training data 330 corresponding to the first group of subjects 332 and the second group of subjects 334 for a second time period.
[0266] In one or more instances, computing system 214 may perform one or more optimization operations during the training process of computational model 302. In one or more illustrative instances, computing system 214 may identify one or more samples obtained from at least one of first group of subjects 332 or second group of subjects 334 as outliers relative to samples obtained from other subjects included in at least one of first group of subjects 332 or second group of subjects 334 during the training process of computational model 302. To illustrate, computing system 214 may determine that model output 304 generated for one or more subjects included in at least one of first group of subjects 332 or second group of subjects 334 differs by at least a threshold amount from model output 304 generated for one or more additional subjects included in at least one of first group of subjects 332 or second group of subjects 334. In one or more examples, the computing system 214 can identify at least one of the one or more first subjects 332 or the one or more second subjects 334 as having a model output 304 that is at least one standard deviation, at least 1.5 standard deviations, at least 2 standard deviations, at least 2.5 standard deviations, or at least 3 standard deviations different from the average model output 304 determined for the other group of at least one of the first group of subjects 332 or the second group of subjects 334. In various examples, the computing system 214 can apply a penalty to information generated from a sample corresponding to a subject that is an outlier relative to information generated from samples corresponding to the other subjects.
[0267] In one or more additional examples, one or more optimization processes implemented by computing system 214 in training computing model 302 can correspond to the number of training loops and / or the number of iterations of a single training loop performed during the training process.
[0268] In one or more illustrative examples, computing system 214 may perform at least 1000 iterations of a training process to generate computational model 302, at least 3000 iterations of a training process to generate computational model 302, at least 5000 iterations of a training process to generate computational model 302, at least 8000 iterations of a training process to generate computational model 302, at least 10,000 iterations of a training process to generate computational model 302, at least 12,000 iterations of a training process to generate computational model 302, or at least 15,000 iterations of a training process to generate computational model 302. In various examples, computing system 214 may end the training process before the loss function associated with computational model 302 converges. In one or more examples, the number of iterations of the training process to generate computational model 302 may correspond to the number of iterations of the training process performed before the training process stops and before the loss function converges.
[0269] In one or more examples, a first stage of the training process implemented by the computing system 214 to generate the computational model 302 can include determining samples that include somatic mutations indicative of one or more types of cancer in the training data 330 relative to samples that do not include somatic mutations indicative of one or more types of cancer in the training data 330. The computing system 214 can then perform a training process for the computational model 302 using samples of the training data 330 that include one or more somatic mutations indicative of one or more types of cancer and using a number of samples obtained from subjects in which no tumor was detected. In various examples, at least 100 iterations of the first stage of the training process can be performed.
[0270] In addition, the training process performed by the computing system 214 may include a second stage that includes predicting a value of a tumor metric for a sample that does not include somatic mutations for one or more types of cancer. The computing system 214 may perform at least 100 additional iterations of the second stage of the training process to generate the computational model 302. The second stage of the training process performed by the computing system 214 to generate the computational model 302 may also include training the computational model 302 using a portion of the training data 330 corresponding to samples having somatic mutations indicative of one or more types of cancer, using predicted values for samples that do not include somatic mutations indicative of one or more types of cancer, and a portion of the training data 330 corresponding to samples obtained from subjects in which no tumors were detected. In various examples, the second stage of the training process performed by the computing system 214 to generate the computational model 302 may be performed at least 2 additional times, at least 3 additional times, at least 4 additional times, at least 5 additional times, or at least 6 additional times. After the first phase of the training process and the second phase of the training process have been completed, computing system 214 may perform a validation process of computing model 302 using information obtained from different samples included in training data 330 .
[0271] In one or more illustrative examples, computing system 214 may perform a training process for more than one computing model 302. In these cases, a single computing model 302 trained by computing system 214 may correspond to a different tissue type as a source of genomic material obtained from a subject included in training data 330. In one or more examples, a single computing model 302 trained by computing system 214 may correspond to a different classification of cancer, such as colorectal cancer, lung cancer, pancreatic cancer, bladder cancer, breast cancer, liver cancer, skin cancer, or one or more additional classifications of cancer. In cases where computing system 214 trains more than one computing model 302 corresponding to different classifications of cancer, the outputs from the single computing models 302 may be aggregated and analyzed by computing system 214 to determine the subject's tissue of origin.
[0272] In various instances, a single computational model 302 corresponding to a given tissue from which genomic material included in a sample comes can have different model components. For example, a first computational model corresponding to a first tissue type generated by the computing system 214 can have a first model component corresponding to a first set of classification regions. In addition, a second computational model corresponding to a second tissue type generated by the computing system 214 can have a second model component corresponding to a second set of classification regions having at least one classification region different from the first set of classification regions. Additionally, the weights of individual components of computational models corresponding to different tissue types can be different. That is, in the case where the first set of classification regions of the first computational model and the second set of classification regions of the second computational model have at least one common classification region, the weights of the model components corresponding to at least one common classification region can be different relative to the first computational model and the second computational model.
[0273] Additionally, one or more additional normalization processes can be performed by the computing system when generating the computational model 302. For example, in at least some cases, molecules treated with MBD can be partitioned differently across different samples. In one or more instances, molecules can be partitioned differently across different samples due to differences in the composition of reagents used to treat the molecules with MBD. In one or more additional instances, molecules can be partitioned differently across different samples due to at least one of differences in equipment or process conditions used to treat the molecules with MBD.
[0274] To illustrate, for one or more first samples, treatment with MBD can result in a first molecule having a region containing a first CG content being separated into a first partition, and a second molecule having a region containing a second CG content being separated into a second partition. Additionally, for one or more second samples, treatment with MBD can result in a third molecule having a third CG content different from the first CG content being separated into the first partition, and a fourth molecule having a region containing a fourth CG content different from the second CG content being separated into the second partition. In various instances, the first molecule can be treated with MBD and separated into the first partition, and the second molecule can be treated with MBD and separated into the second partition across a first cutoff range of CG content. Additionally, the third molecule can be treated with MBD and separated into the first partition, and the fourth molecule can be treated with MBD and separated into the second partition across a second cutoff range of CG content different from the first cutoff range.
[0275] In one or more illustrative examples, the first cutoff range of CG content may include 3-10 CpGs with methylated cytosine, and the second cutoff range may include 6-14 CpGs with methylated cytosine. In one or more additional illustrative examples, the first cutoff range of CG content may include 4-9 CpGs with methylated cytosine, and the second cutoff range may include 7-13 CpGs with methylated cytosine. In one or more additional illustrative examples, the first cutoff range of CG content may include 5-8 CpGs with methylated cytosine, and the second cutoff range may include 8-12 CpGs. In other illustrative examples, the first cutoff range of CG content may include 4-7 CpGs, and the second cutoff range may include 6-10 CpGs. In various examples, the first cutoff range of CG content and the second cutoff range of CG content can be used to determine a threshold amount of methylated cytosine for determining at least one of the training sequencing reads or the test sequencing reads. In at least some examples, the threshold amount of methylated cytosines can include a cutoff number corresponding to a probability (such as at least about 80%, at least about 85%, at least about 90%, at least about 95%, or at least about 99%) that a single molecule treated with MBD is separated into a given partition. In one or more examples, the threshold amount of methylated cytosines can correspond to 5 methylated cytosines, 6 methylated cytosines, 7 methylated cytosines, 8 methylated cytosines, 9 methylated cytosines, 10 methylated cytosines, 11 methylated cytosines, 12 methylated cytosines, 13 methylated cytosines, or 14 methylated cytosines.
[0276] In one or more instances, the computing system 214 can generate a metric for a single classification region based on a quantitative metric determined by analyzing a first number of sequencing reads to identify a first number of nucleic acid molecules having a first amount of CG content and by analyzing a second number of sequencing reads to identify a second number of nucleic acid molecules having a second amount of CG content. In at least some instances, the second number of nucleic acid molecules can be used to modify the metric determined using the first number of nucleic acid molecules to account for variations in the separation of molecules treated using MBD for different samples. In various instances, for a single classification region, a first metric for a given sample can be determined by determining a first quantitative metric corresponding to the number of molecules having a threshold amount of methylated cytosines and having a first amount of cytosine-guanine content in one or more partitions (e.g., the second partition 130 and / or the third partition 134) corresponding to the single classification region. In some embodiments, the first amount of CG content can be at least 5, at least 6, at least 7, at least 8, at least 9, at least 10, at least 15, at least 20, at least 25, or at least 30 CpGs in the nucleic acid molecule. In some embodiments, the first amount of CG content can be between 5-10, 5-15, 5-20, 5-30, 10-15, 10-20, 10-30, 15-20, 15-30, or 20-40 CpGs in the nucleic acid molecule.
[0277] The first metric for a given sample can also be determined by determining a second quantitative metric corresponding to the number of molecules having a threshold amount of methylated cytosine and having a first amount of cytosine-guanine content in one or more partitions (e.g., second partition 130 and / or third partition 134) corresponding to more than one control region (e.g., a positive control region). To illustrate, for a single classification region, the first metric can be determined using a first quantitative metric for the single classification region and a second quantitative metric corresponding to more than one control region.
[0278] The normalization process can also include, for a given sample, determining a second metric for a given sample by determining one or more additional quantitative measurements based on the number of molecules having at least a threshold amount of methylated cytosine and a second amount of cytosine-guanine content in one or more partitions (e.g., the second partition 130 and / or the third partition 134), wherein the second amount of cytosine-guanine content is less than the first amount of cytosine-guanine content. In some embodiments, the second amount of CG content can be between 5-10, 5-15, 10-15, 10-20, or 15-20 CpGs in the nucleic acid molecule. In one or more instances, more than one control area can be a positive control area and / or a negative control area. In one or more instances, the second metric can be determined using an additional quantitative metric and a second quantitative metric. In at least some instances, the second metric of a given sample can be determined by determining the ratio of one or more additional quantitative measurements relative to the second quantitative measurement. In one or more additional examples, a second metric for a given sample can be determined by determining a logarithm (such as a logarithm according to base 10) of a ratio of one or more additional quantitative measures relative to the second quantitative measure.
[0279] In one or more illustrative examples, the second metric of a given sample may include a combination of values, wherein a single value corresponds to an additional quantitative measure and a second quantitative measure based on the number of molecules with at least a threshold amount of methylated cytosine and a given number of CpGs for more than one control region. For example, a first additional quantitative measure may be determined based on a first number of molecules with at least a threshold amount of methylated cytosine in a control region with a first number of CpGs (such as 6), and a second additional quantitative measure may be determined based on a second number of molecules with at least a threshold amount of methylated cytosine in a control region with a second number of CpGs (such as 7). In at least some instances, more additional quantitative measures may be determined based on additional numbers of molecules with a threshold amount of methylated cytosine in a control region with an additional number of CpGs (such as 8 CpGs, 9 CpGs, 10 CpGs, etc.) up to an upper threshold of CpGs (such as 12 CpGs, 13 CpGs, or 14 CpGs). A ratio determined using the further quantitative measure relative to the second quantitative measure may be determined and added to determine the second metric.
[0280] In various examples, correlation factors associated with different amounts of CpGs that can be used to determine a second metric can also be determined for a single classification region. In one or more examples, the correlation factor can modify a single additional quantitative metric, and the modified single additional quantitative metric can then be aggregated to determine the second metric. In one or more additional examples, the first metric and the second metric can be combined to determine a normalized metric corresponding to a given classification region. In one or more illustrative examples, the second metric can be subtracted from the first metric to determine the normalized metric.
[0281] In one or more additional illustrative examples, a correlation factor for a given classification region can be determined for each of more than one different amount of cytosine-guanine content, such as a first correlation factor for 6 CpGs, a second correlation factor for 7 CpGs, a third correlation factor for 8 CpGs, and so on, up to a threshold amount of CG content. In at least some examples, the correlation factor can be determined by analyzing the training data using one or more linear regression techniques. For example, the training data 330 can be fit to a linear regression model for a single classification region to determine the correlation factor. In various examples, fitting of at least a portion of the training data 330 to the linear regression model can be performed by aggregating additional quantitative measures for a given classification region across a range of CG content (such as 6 CpGs, 7 CpGs, up to a threshold number of CpGs) and determining an average quantitative measure.
[0282] In one or more instances, the normalized metric can reduce the variation of the quantitative measure determined for a single sample. In at least some instances, the reduction in variation can result in an increase in the accuracy of the model output 304 relative to at least some of the model outputs 304 determined without performing an additional normalization process to determine the normalized metric.
[0283] Figure 4It is a flowchart of an example method 400 for determining a tumor metric in a subject based on the methylation level of a classification region according to one or more embodiments. In operation 402, method 400 may include obtaining training sequence data including training sequencing reads of more than one sample derived from more than one subject. A single training sequencing read may include a nucleotide sequence corresponding to a nucleic acid fragment in a sample included in more than one sample. A single training sequencing read may have a threshold amount of molecules including methylated cytosine in a region with at least a threshold cytosine-guanine content in a nucleotide sequence. In one or more illustrative examples, more than one sample may include cell-free nucleic acid. In one or more examples, methylated cytosine may be determined using at least one of sodium bisulfite conversion and sequencing, Tet-assisted bisulfite sequencing (TAB-Seq), differential enzymatic cleavage, treatment with MSRE and / or MDRE, or MBD partitioning. In one or more additional examples, methylated cytosines can be determined using one or more single molecule sequencing methods, such as nanopore DNA sequencing or those described in Eid, J. et al. (2009) Real-time DNA sequencing from single polymerase molecules. Science, 323(5910), 133-138.
[0284] In one or more instances, the training process may include obtaining, by a computing system, test sequence data from another subject that is not included in more than one subject. The test sequence data may include sequencing reads of a sample derived from another subject. A single test sequencing read may include a nucleotide sequence corresponding to a nucleic acid fragment included in another sample. Additionally, a single test sequencing read may have at least a threshold amount of molecules with methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. Based on the additional sequence data, a model may be executed to determine an indication of the presence of cancer in another subject. The test sequencing read may then be analyzed to determine a first quantitative measure of a single classification region corresponding to more than one classification region derived from the test sequencing read. In addition, the test sequencing read may be analyzed to determine a second quantitative measure of a single control region corresponding to more than one control region derived from the test sequencing read; then the measure of a single classification region may be determined based on the first quantitative measure of the single classification region and the second quantitative measure of more than one control region. Subsequently, an input vector including the measure of a single classification region may be generated. The model may use the input vector to determine an indication of the presence of cancer in another subject.
[0285] In the case where the model is trained to determine an estimate of a tumor score, the training sequencing reads may constitute a first portion of the training sequence data, and the second portion of the training sequence data includes additional training sequencing reads that are different from the training sequencing reads. In these cases, at least one of the first portion of the training sequence data or the second portion of the training sequence data may be analyzed to determine a single frequency of more than one variant in a single sample present in more than one sample. With respect to a single sample, then a variant with the maximum frequency among more than one variants may be determined, the maximum frequency corresponding to a single frequency with a maximum value in a single frequency derived from a single sample. In one or more illustrative examples, the maximum mutant allele frequency of a single sample may be determined. In various examples, then a single measure of the tumor score of a single sample may be determined based on the maximum value of a single frequency derived from a single sample.
[0286] In at least some instances, the training process of the model can include one or more optimization operations. For example, the training process can include determining one or more additional weights for a single sample included in the training data based on the cancer indication of the single sample being within a threshold confidence level. In response to determining that the cancer indication of the single sample is outside the threshold confidence level, a penalty can be applied to the single sample during the training process.
[0287] One or more training optimization operations may also include using one or more machine learning algorithms to perform one or more first iterations of a training process for the model using a portion of the training data. In addition, the first output data of the model may be generated based on one or more first iterations of the training process. The first output data may correspond to one or more first additional indications of the presence of cancer in a first single subject in more than one subject, and the first single subject may correspond to a portion of the training data. In addition, the training process may include combining the first output data and the training data to generate additional training data, and performing one or more second iterations of the training process for the model using a portion of the additional training data. Then, second output data may be generated for the model based on one or more second iterations of the training process. The second output data may indicate one or more second additional indications of the presence of cancer in a second single subject in more than one subject, wherein the second single subject corresponds to a portion of the additional training data. In one or more illustrative examples, the weight of a single classification region in more than one classification region may be determined based on the first output data and the second output data.
[0288] In addition, the training process can include determining the number of indications of the presence of cancer that are determined during one or more iterations of the training process and have at least a threshold value for one or more samples included in the training data. In these cases, the modification of one or more weights of the model is not modified or is modified by a minimum amount. Additionally, a further number of indications of the presence of cancer that are determined during one or more iterations of the training process and are less than the threshold value for one or more additional samples included in the training data can be determined. In these cases, the modification of one or more additional weights of the model can be determined, and the one or more additional weights are modified by an amount exceeding the minimum amount.
[0289] In addition, at operation 404, process 400 may include analyzing the training sequencing reads to determine a first quantitative measure corresponding to a single classification region in more than one classification region derived from the training sequencing reads. In one or more instances, the first quantitative measure may be determined based on the number of training sequencing reads. In one or more additional instances, the first quantitative measure may be determined based on the number of polynucleotide molecules corresponding to the training sequencing reads. At least a portion of a single classification region in more than one classification region corresponds to a genomic region in which a threshold amount of molecules with methylated cytosine and at least a threshold cytosine-guanine content are detected in a subject of a reference genome. In various instances, more than one classification region may correspond to a genomic region in which at least one mutation occurs in a patient in which cancer is detected. Additionally, more than one classification region may correspond to a first more than one classification region of a first cancer type, and a model may be generated for a second cancer type based on a second more than one classification region different from the first more than one classification region.
[0290] In operation 406, process 400 may include analyzing the training sequencing reads to determine a second quantitative measure corresponding to more than one control region derived from the training sequencing reads. In one or more instances, the second quantitative measure may be determined based on the number of training sequencing reads. In one or more other instances, the second quantitative measure may be determined based on the number of polynucleotide molecules corresponding to the training sequencing reads. A single control region in more than one control region corresponds to another genomic region with at least a threshold cytosine-guanine content of a reference genome. Additionally, in a subject in which cancer is detected and another subject in which cancer is not detected, a single control region may have at least a threshold amount of molecules with methylated cytosine.
[0291] In addition, at operation 408, process 400 may include determining a metric for a single classification region of the more than one classification region based on the first quantitative metric for the single classification region and the second quantitative metric for the more than one control region. In one or more examples, the metric for the single classification region is determined based on a scaling factor and an error correction factor. In one or more illustrative examples, the scaling factor may include a logarithmic function and the error correction factor may include a pseudo count.
[0292] At operation 410, process 400 may include generating, by a computing device, training data including a measure for a single classification region of more than one classification region for the training sequence reads. In embodiments where the indication of cancer is a tumor score, the training data may include a single measure of the tumor score for a single sample of more than one sample, and the model may be performed relative to the single measure of the tumor score for the single sample of more than one sample.
[0293] Process 400 may also include, at operation 412, implementing one or more machine learning algorithms using the training data to generate a model to determine an indication of the presence of cancer in a subject based on the amount of molecules having methylated cytosine in at least a portion of more than one classification region. The model may determine the weight of a single classification region in more than one classification region, and at least a portion of the weights of the single classification regions may be different from each other. In various instances, the one or more machine learning algorithms may include one or more classification algorithms, and the indication of the presence of cancer corresponds to a probability of the presence of cancer in another subject. In one or more additional instances, the one or more machine learning algorithms include one or more regression algorithms, and the indicator corresponds to an estimate of the tumor score of another sample. In one or more illustrative examples, the limit of detection of a model for determining a tumor score for a sample may be no greater than 0.01% at 95% sensitivity, no greater than 0.05% at 95% sensitivity, greater than 0.1% at 95% sensitivity, no greater than 0.15% at 95% sensitivity, no greater than 0.2% at 95% sensitivity, no greater than 0.25% at 95% sensitivity, or no greater than 0.3% at 95% sensitivity.
[0294] In various examples, sequence reads provided to the model during or after the training process have at least a threshold amount of methylated cytosine in the classification region. Sequence reads that meet the methylation level can be generated at least in part using one or more molecular separation methods. The molecular separation method may include combining more than one nucleic acid derived from at least one of the blood or tissue of the subject with a solution containing a certain amount of methyl binding domain (MBD) protein to produce a nucleic acid-MBD protein solution. The nucleic acid-MBD protein solution can then be washed more than once with a salt solution to produce a number of nucleic acid fractions. A single nucleic acid fraction can have a threshold number of molecules with methylated cytosine in a region with at least a threshold cytosine-guanine content in more than one nucleic acid. In one or more illustrative examples, washing in more than one wash can be performed with a solution having a certain concentration of sodium chloride (NaCl), and a nucleic acid fraction having a certain binding strength range with the MBD protein in many nucleic acid fractions can be produced.
[0295] In one or more instances, a first nucleic acid fraction can be determined to be associated with a first partition in more than one nucleic acid partition. The first partition corresponds to a first range of binding strengths to an MBD protein. In addition, a first molecular barcode can be attached to nucleic acid of the first nucleic acid fraction. The first molecular barcode can be associated with the first partition. In addition, a second nucleic acid fraction can be determined to be associated with a second partition in more than one nucleic acid partition. The second partition can correspond to a second range of binding strengths to an MBD protein that is different from the first range of binding strengths to an MBD protein. A second molecular barcode can be attached to nucleic acid of the second nucleic acid fraction. The second molecular barcode is associated with the second partition.
[0296] In one or more additional examples, at least a portion of a number of nucleic acid fractions can be combined with an amount of a restriction endonuclease that cleaves molecules with one or more unmethylated cytosines to produce at least a portion of more than one sample for producing sequencing reads. In these cases, the threshold amount of molecules with methylated cytosines corresponds to the minimum frequency of molecules with methylated cytosines in a region with at least a threshold cytosine-guanine content. In one or more additional examples, at least a portion of a number of nucleic acid fractions can be combined with an amount of a restriction endonuclease that cleaves molecules with methylated cytosines to produce at least a portion of more than one sample for producing sequencing reads. In these cases, the threshold amount of molecules with methylated cytosines corresponds to the maximum frequency of molecules with methylated cytosines in a region with at least a threshold cytosine-guanine content.
[0297] Example Method
[0298] A. Determining Indications of Cancer in a Sample
[0299] In some embodiments, the method disclosed herein includes sequencing the cfDNA from the sample and determining the methylation level of more than one target region including the DNA sequence as the differentially methylated region and the control region. In some embodiments, the method disclosed herein includes capturing at least one epigenetic target region group from cfDNA or its subsample, including contacting the cfDNA or its subsample with a target-specific probe specific to at least one epigenetic target region group, determining the methylation level of the target region, and determining whether there is an indication of cancer in the sample obtained from the subject. In any of these embodiments, the methylation level / state of the nucleic acid molecule can be determined using one or more methods, including using a binding agent that recognizes modified cytosine to partition the molecule, MSRE / MDRE digestion, methylation-sensitive conversion such as bisulfite conversion, direct detection during sequencing, or any other suitable method. Various methods are described herein.
[0300] In some embodiments, the methods disclosed herein include the steps of partitioning a sample comprising DNA by contacting the DNA with an agent that recognizes modified cytosine in the DNA, sequencing the DNA, and determining a quantitative measure of nucleic acid in more than one region.
[0301] In one or more aspects, the method includes obtaining training sequence data by a computing system having one or more hardware processors and memory, the training sequence data including training sequencing reads from more than one sample derived from more than one subject, a single training sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in a sample in the more than one sample, and the single training sequencing read corresponds to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The method also includes analyzing the training sequencing reads by the computing system to determine a first quantitative measure corresponding to a single classification region in more than one classification region derived from the training sequencing reads, at least a portion of the single classification region in more than one classification region corresponding to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer was detected. The method also includes analyzing the training sequencing reads by a computing system to determine a second quantitative measure corresponding to more than one control region derived from the training sequencing reads, wherein a single control region in the more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and having at least a threshold amount of methylated cytosine in the subject in which the cancer was detected and another subject in which the cancer was not detected. The method also includes determining, by the computing system, a measure of a single classification region in more than one classification region based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region. The method also includes generating training data by a computing device, the training data including measures of single classification regions in more than one classification region of the training sequence reads of the sample of the training subject. The method also includes implementing one or more machine learning algorithms by the computing system and using the training data to generate a model to determine an indication of the presence of cancer in the subject based on the amount of methylated cytosine in at least a portion of the more than one classification region, the model including weights of single classification regions in more than one classification region, and at least a portion of the weights of the single classification regions are different from each other.
[0302] In one or more aspects, a computing system includes: one or more hardware processors; and one or more non-transitory computer-readable storage media including computer-readable instructions, which, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations including: obtaining training sequence data, the training sequence data including training sequencing reads from more than one sample derived from more than one subject, including a single training sequencing read corresponding to a nucleotide sequence of a nucleic acid fragment included in a sample of the more than one sample and a single training sequencing read corresponding to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The operations also include analyzing the training sequencing reads to determine a first quantitative measure corresponding to a single classification region in more than one classification region derived from the training sequencing reads, at least a portion of the single classification region in more than one classification region corresponding to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer is detected. The operation also includes analyzing the training sequencing reads to determine a second quantitative measure corresponding to more than one control region derived from the training sequencing reads, wherein a single control region in the more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and at least a threshold amount of methylated cytosine in the subject in which the cancer was detected and another subject in which the cancer was not detected. The operation also includes determining a measure of a single classification region in more than one classification region based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region. The operation also includes generating training data, which includes measures of single classification regions in more than one classification region of the training sequence reads of the sample of the training subject. The operation also includes implementing one or more machine learning algorithms using the training data to generate a model to determine an indication of the presence of cancer in the subject based on the amount of methylated cytosine in at least a portion of the more than one classification region, the model including weights of single classification regions in more than one classification region, and at least a portion of the weights of the single classification regions are different from each other.
[0303] In one or more aspects, one or more computer-readable storage media include computer-readable instructions that, when executed by one or more processors of a computing system, cause the computing system to perform operations including obtaining training sequence data, the training sequence data including training sequencing reads from more than one sample derived from more than one subject, a single training sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in a sample in the more than one sample, and the single training sequencing read corresponds to a molecule having a threshold amount of methylated cytosine in a region of the nucleotide sequence having at least a threshold cytosine-guanine content. The operations also include analyzing the training sequencing reads to determine a first quantitative measure corresponding to a single classification region in the more than one classification region derived from the training sequencing reads, at least a portion of the single classification region in the more than one classification region corresponding to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer was detected. The operation also includes analyzing the training sequencing reads to determine a second quantitative measure corresponding to more than one control region derived from the training sequencing reads, wherein a single control region in the more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and at least a threshold amount of methylated cytosine in the subject in which the cancer was detected and another subject in which the cancer was not detected. The operation also includes determining a measure of a single classification region in more than one classification region based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region. The operation also includes generating training data, which includes measures of single classification regions in more than one classification region of the training sequence reads of the sample of the training subject. The operation also includes implementing one or more machine learning algorithms using the training data to generate a model to determine an indication of the presence of cancer in the subject based on the amount of methylated cytosine in at least a portion of the more than one classification region, the model including weights of single classification regions in more than one classification region, and at least a portion of the weights of the single classification regions are different from each other.
[0304] In one or more aspects, the method includes obtaining, by a computing system having one or more hardware processors and memory, sequencing reads derived from a sample obtained from a subject, wherein a single sequencing read includes a nucleotide sequence corresponding to a nucleic acid fragment included in the sample and corresponds to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The method may also include determining, by the computing system, a first quantitative measure corresponding to a single classification region in more than one classification region derived from the sequencing read, wherein at least a portion of the single classification region in more than one classification region corresponds to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer is detected. The method may also include analyzing the sequencing reads by the computing system to determine a second quantitative measure corresponding to more than one control region derived from the sequencing read, wherein a single control region in more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and having at least a threshold amount of methylated cytosine in a subject in which cancer is detected and another subject in which cancer is not detected. The method may also include determining, by the computing system, more than one metric based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region, wherein a single metric in the more than one metric corresponds to a single classification region in the more than one classification region. The method may also include determining, by the computing system, an indication of the presence of cancer in the subject based on at least a portion of the more than one metric.
[0305] In one or more aspects, a computing system includes: a processor; and a memory storing instructions, which, when executed by the processor, configure the computing system to: obtain sequencing reads derived from a sample obtained from a subject, a single sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in the sample, and corresponding to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The computing system can also determine a first quantitative measure corresponding to a single classification region in more than one classification region derived from the sequencing read, at least a portion of the single classification region in more than one classification region corresponds to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer is detected. The computing system can also analyze the sequencing reads to determine a second quantitative measure corresponding to more than one control region derived from the sequencing reads, wherein a single control region in more than one control region corresponds to another genomic region in the reference genome having at least a threshold cytosine-guanine content and having at least a threshold amount of methylated cytosine in a subject in which cancer is detected and another subject in which cancer is not detected. The computing system may also determine more than one metric based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region, wherein a single metric in the more than one metrics corresponds to a single classification region in the more than one classification region. The computing system may also determine an indication of the presence of cancer in the subject based on at least a portion of the more than one metrics.
[0306] The computing system includes: a processor; and a memory storing instructions, which, when executed by the processor, configure the system to: obtain test sequence data from a subject, the test sequence data including test sequencing reads of a sample derived from the subject, a single test sequencing read including a nucleotide sequence corresponding to a nucleic acid fragment included in another sample, and the single test sequencing read corresponds to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content in the nucleotide sequence. The computing system can also analyze the test sequencing reads to determine a first quantitative measure corresponding to a single classification region of more than one classification region derived from the test sequencing reads, at least a portion of the single classification region of more than one classification region corresponding to a genomic region of a reference genome having a threshold amount of methylated cytosine and having at least a threshold cytosine-guanine content in a subject in which cancer is detected. The computing system can also analyze the test sequencing reads to determine a second quantitative measure corresponding to a single control region in more than one control region derived from the test sequencing reads, wherein the single control region in more than one control region corresponds to another genomic region in the reference genome in which the subject in which the cancer was detected and another subject in which the cancer was not detected having at least a threshold cytosine-guanine content and having at least a threshold amount of methylated cytosine. The computing system can also determine the measure of the single classification region based on the first quantitative measure of the single classification region and the second quantitative measure of more than one control region. The computing system can also generate an input vector including the measure of the single classification region. The computing system can also determine the indication of the presence of cancer in the subject by providing an input vector to a model that implements one or more machine learning techniques to generate an indication of the presence of cancer in the subject, the model including weights of single classification regions in more than one classification region, and at least a portion of the weights of the single classification regions are different from each other.
[0307] B. Partitioning a sample into more than one subsample
[0308] In some embodiments described herein, different forms of DNA (e.g., hypermethylated and hypomethylated DNA) are physically partitioned based on one or more features of the DNA. The method can be used to determine, for example, whether certain sites or regions are hypermethylated or hypomethylated. Partitioning can be performed before the adapter is attached to the DNA molecules in the sample, for example, so as to facilitate including a partition label in the adapter. The partition label can be used to identify which partition a molecule is present in. After partitioning (and attaching adapters, if applicable), further steps such as amplification, target capture, and sequencing can be performed.
[0309] Methylation spectrum analysis can include determining the methylation pattern throughout the different regions of the genome. For example, based on the methylation degree (for example, the relative number of the methylated core base of each molecule) by molecular partition and as discussed above comprising the further step of sequencing, the molecular sequence in the different partitions can be mapped to the reference genome. This can show that the methylation is higher or the methylation is not too high region compared with other regions in the genome. In this way, in contrast to a single molecule, a genomic region can be different in its methylation degree.
[0310] The nucleic acid molecules in the sample can be partitioned, for example, by enriching in a rare nucleic acid molecule that is more common in a partition of the sample to increase rare signals. For example, genetic variations present in hypermethylated DNA but less (or not) present in hypomethylated DNA can be more easily detected by partitioning the sample into hypermethylated and hypomethylated nucleic acid molecules. By analyzing more than one partition of the sample, multidimensional analysis of a single molecule can be performed, and therefore higher sensitivity can be achieved. Partitions can include physically partitioning nucleic acid molecules into partitions or subsamples based on the presence or absence of one or more methylated core bases. Samples can be partitioned into partitions or subsamples based on the features indicating differential gene expression or disease states. Samples can be partitioned based on the features or feature combinations that provide signal differences between normal states and diseased states during nucleic acid analysis, such as cell-free DNA (cfDNA), non-cfDNA, tumor DNA, circulating tumor DNA (ctDNA) and cell-free nucleic acid (cfNA).
[0311] In some embodiments, hypermethylated and / or hypomethylated variable epigenetic target regions are analyzed to determine whether they display a differential methylation signature of a specific immune cell type, such as a rare immune cell type, a tumor cell, or a cell type that does not normally contribute to the DNA sample being analyzed (such as cfDNA).
[0312] In some cases, the heterogeneous DNA in the sample is partitioned into two or more partitions (e.g., at least 3, 4, 5, 6, or 7 partitions). In some embodiments, each partition is differentially labeled. The labeled partitions can then be pooled together for collective sample preparation and / or sequencing. The partition-labeling-pooling step can occur more than once, with each round of partitioning occurring based on a different feature (examples provided herein) and labeled using a differential label that is distinct from other partitions and partitioning means. In other cases, the differentially labeled partitions are sequenced separately.
[0313] In some embodiments, sequence reads from differentially labeled and pooled DNA are obtained and analyzed in silico. Labels are used to sort reads from different partitions. Analysis can be performed at the level of each partition and at the level of the entire nucleic acid population to detect genetic variation. For example, analysis can include determining genetic variation, such as CNV, SNV, insertion / deletion, fusion in the nucleic acid of each partition by computer analysis. In some cases, analysis can include determining chromatin structure by computer simulation. For example, the coverage of sequence reads can be used to determine the positioning of nucleosomes in chromatin. Higher coverage can be associated with higher nucleosome occupancy in a genomic region, while lower coverage can be associated with lower nucleosome occupancy or nucleosome depleted region (NDR).
[0314] In some embodiments, partition is based on one or more features, such as methylation.Can be based on other features, such as sequence length, nucleosome binding, sequence mismatch, immunoprecipitation and / or protein combined with DNA, use appropriate technology as a part of data analysis or partition when applicable, molecule is sorted.The partition of gained can include one or more of the following nucleic acid forms: single-stranded DNA (ssDNA), double-stranded DNA (dsDNA), shorter DNA fragments and longer DNA fragments.In some embodiments, partition based on cytosine modification (for example, cytosine methylation) or methylation is usually carried out, and optionally combined with at least one other partition step, this step can be based on any aforementioned features or form of DNA.In some embodiments, heterogeneous nucleic acid colony is partitioned into nucleic acids with one or more epigenetic modifications and without described one or more epigenetic modifications. The example of epigenetic modification includes the presence or absence of methylation, methylation level, methylation type (for example, 5-methylcytosine and other types of methylation, such as adenine methylation and / or cytosine hydroxymethylation) and the correlation and correlation level with one or more proteins (such as histone).Alternatively or additionally, the heterogeneous nucleic acid population can be partitioned into nucleic acid molecules associated with nucleosomes and nucleic acid molecules without nucleosomes.Alternatively or additionally, the heterogeneous nucleic acid population can be partitioned into single-stranded DNA (ssDNA) and double-stranded DNA (dsDNA).Alternatively or additionally, the heterogeneous nucleic acid population can be partitioned based on nucleic acid length (for example, molecules of maximum 160bp and molecules with a length greater than 160bp).
[0315] The agent for partitioning the nucleic acid population within the sample can be an affinity agent, such as an antibody with desired specificity, a natural binding partner or a variant thereof (Bock et al., Nat Biotech 28: 1106-1114 (2010); Song et al., Nat Biotech 29: 68-72 (2011)), or an artificial peptide with specificity for a specific target selected by phage display, for example. In some embodiments, the agent for partitioning is an agent that recognizes modified nucleobases. In some embodiments, the modified nucleobase recognized by the agent is a modified cytosine, such as methylcytosine (e.g., 5-methylcytosine). In some embodiments, the modified nucleobase recognized by the agent is the product of a program that differently affects the first nucleobase in the DNA of the sample and the second nucleobase in the DNA. In some embodiments, the modified nucleobase can be a "converted nucleobase", meaning that its base pairing specificity is changed by a program. For example, some procedures convert unmethylated or unmodified cytosine into dihydrouracil, or more generally, at least one modified or unmodified form of cytosine undergoes deamination to produce uracil (considered to be a modified nucleobase in the context of DNA) or a further modified form of uracil. Examples of partitioning agents include antibodies, such as antibodies that recognize modified nucleobases, which can be modified cytosines, such as methylcytosine (e.g., 5-methylcytosine). In some embodiments, partitioning agents are antibodies that recognize modified cytosines such as 5-carboxycytosine (5caC) other than 5-methylcytosine. Alternative partitioning agents include methyl binding domains (MBDs) and methyl binding proteins (MBPs) as described herein, including proteins such as MeCP2.
[0316] Additional non-limiting examples of partitioning agents are histone binding proteins, which can separate nucleic acids bound to histones from free or unbound nucleic acids.Examples of histone binding proteins that can be used in the methods disclosed herein include RBBP4, RbAp48, and SANT domain peptides.
[0317] Partition agent and specific nucleic acid combination and nucleic acid partition are subsamples that can occur to a certain extent or can occur in a binary mode basically.In some cases, the nucleic acid that comprises a certain modification of a larger ratio is combined with agent to a greater extent than the nucleic acid that comprises a modification of a smaller ratio.Similarly, partition can produce a subsample that comprises a nucleic acid that comprises a certain modification of a larger and smaller ratio.Alternatively, partition can produce a subsample that comprises the nucleic acid that comprises all modification or does not have a nucleic acid that comprises modification basically.In all cases, the modification of various levels of sequential wash-out can be eluted from partition agent.
[0318] In some embodiments, partitioning can include binary partitioning and partitioning based on degree of modification / level.For example, methylated fragments can be partitioned by methylated DNA immunoprecipitation (MeDIP), or all methylated fragments and unmethylated fragments can be partitioned using methyl binding domain proteins (e.g., MethylMinder methylated DNA enrichment kit (ThermoFisher Scientific)). Subsequently, other partitioning can include eluting fragments with different methylation levels by adjusting the salt concentration of a solution containing a methyl binding domain and a binding fragment.As salt concentration increases, fragments with greater methylation levels are eluted.
[0319] In some cases, the final partition is enriched with nucleic acids with varying degrees of modification (overrepresentative or underrepresentative modifications). Overrepresentation and underrepresentation can be defined by the number of modifications carried by the nucleic acid relative to the median of the modification of each chain in the population. For example, if the median of 5-methylcytosine residues in the nucleic acid in the sample is 2, the modification of nucleic acids containing more than two 5-methylcytosine residues is overrepresentative, while nucleic acids with 1 or 0 5-methylcytosine residues are underrepresentative. The effect of affinity separation is to enrich nucleic acids whose modifications are overrepresented in the binding phase and nucleic acids whose modifications are not fully represented in the non-binding phase (i.e., in solution). The nucleic acids in the binding phase can be eluted before subsequent treatment.
[0320] When using MeDIP or During the methylation DNA enrichment kit (ThermoFisher Scientific), sequential elution can be used to separate the methylation partitions of different levels. For example, low methylation partitions (no methylation) can be separated from methylation partitions by contacting the nucleic acid colony with the MBD attached to the magnetic beads from the kit. Pearls are used to separate methylated nucleic acids from unmethylated nucleic acids. Subsequently, one or more elution steps are sequentially carried out to elute the nucleic acids with various methylation levels. For example, the first group of methylated nucleic acids can be eluted at a salt concentration of 160mM or higher, such as at least 150mM, at least 200mM, 300mM, 400mM, 500mM, 600mM, 700mM, 800mM, 900mM, 1000mM or 2000mM. After such methylated nucleic acids are eluted, magnetic separation is used again to separate the methylated nucleic acids of higher levels from the nucleic acids with lower methylation levels. The elution and magnetic separation steps can be repeated to produce various partitions, such as a hypomethylated partition (enriched in nucleic acids containing no methylation), a methylated partition (enriched in nucleic acids containing low levels of methylation), and a hypermethylated partition (enriched in nucleic acids containing high levels of methylation).
[0321] In some methods, the nucleic acid bound to the agent for affinity separation is subjected to a washing step based on the partition. The washing step washes away the nucleic acid weakly bound to the affinity agent. Such nucleic acid can be enriched with a nucleic acid that is close to the mean value or median (i.e., the intermediate value between the nucleic acid bound to the solid phase and the nucleic acid not bound to the solid phase when the sample is initially contacted with the agent) degree of modification.
[0322] Affinity separation results in at least two and sometimes three or more partitions of nucleic acids with different degrees of modification. When the partitions are still separated, the nucleic acids of at least one partition and usually two or three (or more) partitions are connected to nucleic acid tags, which are usually provided as components of adapters, and the nucleic acids in different partitions receive different tags that distinguish members of one partition from members of another partition. The tags connected to the nucleic acid molecules of the same partition can be the same or different from each other. But if they are different from each other, a portion of the tag encoding can be shared so that the molecules to which they are attached are identified as belonging to a specific partition.
[0323] For more details on partitioning nucleic acid samples based on features such as methylation, see WO2018 / 119452, which is incorporated herein by reference.
[0324] In some embodiments, nucleic acid molecules can be fractionated into different partitions based on nucleic acid molecules that bind to a specific protein or fragment thereof and nucleic acid molecules that do not bind to the specific protein or fragment thereof.
[0325] Nucleic acid molecules can be fractionated based on DNA-protein binding. Protein-DNA complexes can be fractionated based on the specific properties of protein. Examples of such properties include various epitopes, modifications (e.g., histone methylation or acetylation) or enzymatic activity. Examples of proteins that can be used in conjunction with DNA and as the basis for fractionation can include but are not limited to protein A and protein G. Any suitable method can be used to fractionate nucleic acid molecules based on protein binding regions. Examples for methods for fractionating nucleic acid molecules based on protein binding regions include but are not limited to SDS-PAGE, chromatin immunoprecipitation (ChIP), heparin chromatography and asymmetric field flow fractionation (AF4).
[0326] In some embodiments, partitioning a sample into more than one subsample is performed by contacting the nucleic acid with an antibody that recognizes a modified nucleobase in the DNA, which can be a modified cytosine or a product of a procedure that differently affects a first nucleobase in the DNA of the sample and a second nucleobase in the DNA. In some embodiments, the modified nucleobase is 5mC. In some embodiments, the modified nucleobase is 5caC. In some embodiments, the modified nucleobase is dihydrouracil (DHU). In some embodiments, antibodies that recognize modified nucleobases in DNA are used to partition single-stranded DNA.
[0327] In some embodiments, partitioning is performed by contacting the nucleic acid with a methyl binding domain ("MBD") of a methyl binding protein ("MBP"). In some such embodiments, the nucleic acid is contacted with an intact MBP. In some embodiments, the MBD binds to 5-methylcytosine (5mC), and the MBP comprises the MBD and is referred to interchangeably herein as a methyl binding protein or a methyl binding domain protein. In some embodiments, the MBD binds to 5mC and 5hmC. In some embodiments, the MBD is attached to a paramagnetic bead (such as a M-280 streptavidin) coupled. Partitioning into fractions with different degrees of methylation can be performed by eluting the fractions with increasing NaCl concentrations.
[0328] In some embodiments, bound DNA is eluted by contacting the antibody or MBD with a protease such as proteinase K. This may be performed instead of or in addition to an elution step using NaCl as discussed above.
[0329] Examples of agents that recognize modified nucleobases contemplated herein include, but are not limited to:
[0330] (a) The protein MeCP2 preferentially binds to 5-methyl-cytosine compared to unmodified cytosine.
[0331] (b) RPL26, PRP8, and the DNA mismatch repair protein MHS6 preferentially bind to 5-hydroxymethyl-cytosine compared to unmodified cytosine;
[0332] (c) FOXK1, FOXK2, FOXP1, FOXP4, and FOXI3 preferentially bind to 5-formylcytosine compared to unmodified cytosine (Iurlaro et al., Genome Biol. 14:R119 (2013)).
[0333] (d) Antibodies specific for one or more methylated or modified nucleobases or their conversion products such as 5mC, 5caC or DHU.
[0334] Typically, elution varies with the number of modifications such as the number of methylation sites per molecule, wherein the molecule with more methylation is eluted under increased salt concentration. In order to elute DNA into different populations based on the degree of methylation, people can use a series of elution buffers that increase the NaCl concentration. The salt concentration can be in the range of about 100nM to about 2500mM NaCl. In an embodiment, the process results in three (3) partitions. The molecule is contacted with a solution of a first salt concentration, and the solution comprises a molecule of an agent containing a modified nucleobase, which can be attached to a capture portion, such as streptavidin. At a first salt concentration, a molecular population will be combined with an agent, and a population will remain unbound. Unbound populations can be separated into "hypomethylation" populations. For example, the first partition of DNA enriched in hypomethylated form is a partition that remains unbound at low salt concentrations (e.g., 100mM or 160mM). The second partition of the enriched moderately methylated DNA is eluted using a medium salt concentration (e.g., a concentration between 100mM and 2000mM). This is also separated from the sample. The DNA-enriched third partition enriched for hypermethylated forms is eluted using a high salt concentration (eg, at least about 2000 mM).
[0335] In some embodiments, monoclonal antibodies raised against 5-methylcytidine (5mC) are used to purify methylated DNA. For example, at 95°C, the DNA is denatured to produce single-stranded DNA fragments. Protein G coupled to a standard or magnetic beads and washes after incubation with an anti-5mC antibody are used to immunoprecipitate DNA bound to the antibody. Such DNA can then be eluted. Partitions can include unprecipitated DNA and one or more partitions eluted from the beads.
[0336] In some embodiments, the sample DNA (e.g., between 5ng and 200ng) is mixed with a methyl binding domain (MBD) buffer and magnetic beads conjugated to the MBD protein and incubated overnight. During this incubation, the methylated DNA (highly methylated DNA) binds to the MBD protein on the magnetic beads. Unmethylated (lowly methylated DNA) or less methylated (medium methylated) DNA is washed off the beads with a buffer containing increasing salt concentrations. For example, one, two or more fractions containing unmethylated DNA, lowly methylated DNA and / or medium methylated DNA can be obtained from such washings. Finally, the heavily methylated DNA (highly methylated DNA) is eluted from the MBD protein using a high salt buffer. In some embodiments, these washes produce three partitions of DNA with increasing methylation levels (lowly methylated partitions, medium methylated fractions, and high methylated partitions).
[0337] In some embodiments, the partitioning procedure can result in incomplete sorting of DNA molecules in the subsamples. For example, a small portion of molecules in an unmethylated or hypomethylated subsample may be highly modified (e.g., hypermethylated), and / or a small portion of molecules in a hypermethylated subsample may be unmodified or substantially unmodified (e.g., unmethylated or substantially unmethylated). Such molecules are considered to be nonspecifically partitioned.
[0338] In some embodiments, methylation-dependent nuclease is used, such as methylation-dependent restriction endonuclease (MDRE), digestion / cracking wherein restriction endonuclease (RE) recognition site contains DNA of methylated nucleotides, but does not crack wherein restriction endonuclease (RE) recognition site contains DNA of unmethylated nucleotides, removes the molecule of non-specific partition.In some embodiments, methylation-sensitive nuclease is used, such as methylation-sensitive restriction endonuclease (MSRE), digestion / cracking wherein restriction endonuclease (RE) recognition site contains DNA of unmethylated nucleotides, but does not crack wherein restriction endonuclease (RE) recognition site contains DNA of methylated nucleotides, removes the molecule of non-specific partition.For example, in some embodiments, low methylation subsample is contacted with methylation-dependent nuclease (such as methylation-dependent restriction endonuclease), so as to degrade the DNA (such as methylated DNA) of non-specific partition in subsample.Alternatively or additionally, high methylation subsample is contacted with methylation-sensitive endonuclease (such as methylation-sensitive restriction endonuclease), so as to degrade the DNA of non-specific partition in subsample.
[0339] The degradation of the DNA of non-specific partitions in the subsample of one or more partitions can improve the performance of the method for the accurate partition of the DNA modified based on cytosine.For example, such degradation can provide the sensitivity improved and / or simplify downstream analysis.In some embodiments, based on modification (such as methylation) by DNA partition, then using MDRE and / or MSRE as described herein to remove the DNA of non-specific partitions provides efficiency and / or cost improved compared to DNA analysis method, and the DNA analysis method comprises the program that differently affects the first core base and the second core base, such as bisulfite sequencing or bisulfite conversion.
[0340] In some embodiments, one or more nucleases are used to degrade the DNA molecules of non-specific partitions. In some embodiments, subsamples are contacted with more than one nuclease. Subsamples can be contacted with nucleases sequentially or simultaneously. When nucleases are active under similar conditions (e.g., buffer compositions), it may be advantageous to use nucleases simultaneously to avoid unnecessary sample manipulations. Subsamples are contacted with more than one methylation-dependent restriction endonucleases to more completely degrade the high methylated DNA of non-specific partitions. Subsamples are contacted with more than one methylation-sensitive restriction endonucleases to more completely degrade the low methylated and / or unmethylated DNA of non-specific partitions.
[0341] In some embodiments, the methylation-dependent nuclease comprises one or more of MspJI, LpnPI, FspEI, or McrBC. In some embodiments, at least two methylation-dependent nucleases are used. In some embodiments, at least three methylation-dependent nucleases are used.
[0342] In some embodiments, the methylation-sensitive nuclease includes one or more of the following: AatII, AccII, AciI, Aor13HI, Aor15HI, BspT104I, BssHII, BstUI, Cfr10I, ClaI, CpoI, Eco52I, HaeII, HapII, HhaI, Hin6I, HpaII, HpyCH4IV, MluI, MspI, NaeI, NotI, NruI, NsbI, PmaCI, Psp1406I, PvuI, SacII, SalI, SmaI and SnaBI. In some embodiments, at least two methylation-sensitive nucleases are used. In some embodiments, at least three methylation-sensitive nucleases are used. In some embodiments, the methylation-sensitive nuclease includes BstUI and HpaII. In some embodiments, two methylation-sensitive nucleases include HhaI and AccII. In some embodiments, the methylation-sensitive nucleases include BstUI, HpaII, and Hin6I.
[0343] In some embodiments, partitions of DNA are desalted and concentrated in preparation for enzymatic steps of library preparation.
[0344] C. Adapter ligation
[0345] In some embodiments, adapters are added to DNA. This can be performed simultaneously with the amplification procedure, for example, by providing adapters in the 5' portion of the primer (in the case of using PCR, this can be referred to as library preparation PCR or LP-PCR). In some embodiments, adapters are added by other methods such as connection. In some such methods, before partitioning or before capture, the first adapter is added to the nucleic acid by being connected to the 3' end of the nucleic acid, which can include connection to single-stranded DNA. The adapter can be used as a priming site for the synthesis of the second chain, for example, using a universal primer and DNA polymerase. The second adapter can then be connected to at least the 3' end of the second chain of the now double-stranded molecule. In some embodiments, the first adapter comprises an affinity tag, such as biotin, and the nucleic acid connected to the first adapter is bound to a solid support (e.g., beads), which can comprise a binding partner of the affinity tag, such as streptavidin. For further discussion of related procedures, see Gansauge et al., Nature Protocols 8: 737-748 (2013). Commercial kits are available for preparing sequencing libraries compatible with single-stranded nucleic acids, such as the Accel- Methyl-Seq DNA Library Kit. In some embodiments, after adaptor ligation, the nucleic acid is amplified.
[0346] Preferably, the adapters comprise a sufficient number of different tags so that the number of tag combinations results in a high probability, such as 95%, 99% or 99.9%, that two nucleic acids with the same start and end points receive different tag combinations. Adapters with the same or different tags may comprise the same or different primer binding sites, but preferably the adapters comprise the same primer binding sites.
[0347] In some embodiments, after attachment of the adaptor, the nucleic acid is subjected to amplification. Amplification can use, for example, a universal primer that recognizes a primer binding site in the adaptor.
[0348] In some embodiments, after the attachment of the adapter, the DNA is partitioned, including contacting the DNA with an agent that preferentially binds to the nucleic acid carrying the epigenetic modification. The nucleic acid is partitioned into at least two subsamples, the difference being the degree of binding of the modified nucleic acid to the agent. For example, if the agent has an affinity for the modified nucleic acid, the modification is preferentially combined with the agent by the over-represented nucleic acid (compared with the median representation in the population), while the modification is not fully represented. The nucleic acid is not bound to the agent or is more easily eluted from the agent. The nucleic acid can then be amplified by a primer that binds to the primer binding site in the adapter. On the contrary, the partition can be performed before the adapter is attached, in which case the adapter can include a differential tag, and the differential tag includes a component that identifies the molecule in which partition it appears. In some embodiments, the nucleic acid is connected to a Y-shaped adapter comprising a primer binding site and a tag at both ends. Amplification molecule.
[0349] D. Add labels
[0350] "Tag" a DNA molecule is a process in which a tag is attached to or associated with a DNA molecule. A tag can be a molecule, such as a nucleic acid, that contains information indicating the characteristics of the molecule associated with the tag. For example, a molecule can carry a sample tag (which distinguishes molecules in one sample from molecules in different samples) or a molecular tag / molecular barcode / barcode (which distinguishes different molecules from each other (in both the case of unique and non-unique tagging). For methods involving a partitioning step, a partitioning tag (which distinguishes molecules in one partition from molecules in different partitions) can be included. In some embodiments, the adapter added to the DNA molecule comprises a tag. In certain embodiments, a tag can include a combination of one barcode or more barcodes. As used herein, the term "barcode" refers to a nucleic acid molecule having a specific nucleotide sequence, or to the nucleotide sequence itself, depending on the context. A barcode can have, for example, between 10 and 100 nucleotides. As desired for a particular purpose, a collection of barcodes can have a degenerate sequence or can have a sequence with a certain Hamming distance (hamming distance). distance). Thus, for example, a molecular barcode may comprise a barcode or a combination of two barcodes, each barcode being attached to a different end of the molecule. Additionally or alternatively, different sets of molecular barcodes or molecular tags may be used for different partitions and / or samples, such that the barcodes are used as molecular tags by their individual sequences and also for identifying the partitions and / or samples to which they correspond based on the group of which they are members.
[0351] In some embodiments, two or more partitions (e.g., each partition) are labeled for differences. Labels can be used to mark a single polynucleotide population partition so that a label (or more than one label) is associated with a specific partition. Alternatively, labels can be used in embodiments where a partitioning step is not used. In some embodiments, a single label can be used to mark a specific partition. In some embodiments, more than one different label can be used to mark a specific partition. In embodiments where more than one different label is used to mark a specific partition, the set of labels used to mark a partition can be easily distinguished from the set of labels used to mark other partitions. In some embodiments, the tag can have additional functions, such as the tag can be used to index the source of the sample or used as a unique molecule identifier (which can be used to improve the quality of sequencing data by distinguishing sequencing errors from mutations, for example, as in Kinde et al., Proc Nat'l Acad Sci USA 108:9530-9535 (2011); Kou et al., PLoS ONE, 11:e0146638 (2016)) or as a non-unique molecule identifier, for example, as described in U.S. Patent No. 9,598,731. Similarly, in some embodiments, the tag can have additional functions, such as the tag can be used to index the source of the sample or used as a non-unique molecule identifier (which can be used to improve the quality of sequencing data by distinguishing sequencing errors from mutations). In one embodiment, labeling the partitions includes labeling the molecules in each partition with the partition tags. After the partitions are reassembled (e.g., to reduce the number of sequencing runs required and avoid unnecessary costs) and the molecules are sequenced, the partition tags identify the source partitions. In some embodiments, the partition labels can be used as identifiers of source partitions and molecules, i.e., different partitions are labeled with different sets of molecular labels (e.g., comprising a pair of barcodes). In this way, one or more molecular barcodes attached to a molecule indicate the source partition and can be used to distinguish molecules within a partition. For example, a first set of 35 barcodes can be used to label molecules in a first partition, while a second set of 35 barcodes can be used to label molecules in a second partition.
[0352] In some embodiments, after partitioning and labeling with partition labels, the molecules can be pooled for sequencing in a single run. In some embodiments, sample labels are added to the molecules in a step, such as after adding partition labels and pooling. Sample labels can help pool material generated from more than one sample for sequencing in a single sequencing run.
[0353] Optionally, in some embodiments, partition labels can be associated with samples and partitions. As a simple example, a first label can indicate a first partition of a first sample; a second label can indicate a second partition of the first sample; a third label can indicate a first partition of a second sample; and a fourth label can indicate a second partition of the second sample....
Claims
1. A method, comprising: obtaining, by a computing system having one or more hardware processors and a memory, training sequence data, the training sequence data comprising training sequencing reads from more than one sample derived from more than one subject, a single training sequencing read comprising a nucleotide sequence corresponding to a nucleic acid fragment included in a sample of the more than one sample, and a single training sequencing read corresponding to a molecule having a threshold amount of methylated cytosine in a region of the nucleotide sequence having at least a threshold cytosine-guanine content; analyzing, by the computing system, the training sequenced reads to determine a first quantitative measure corresponding to a single one of more than one classification regions derived from the training sequenced reads, at least a portion of the single one of more than one classification regions corresponding to a genomic region of a reference genome having the threshold amount of methylated cytosine and having at least the threshold cytosine-guanine content in a subject in which cancer was detected; analyzing, by the computing system, the training sequenced reads to determine a second quantitative measure corresponding to more than one control region derived from the training sequenced reads, a single control region of the more than one control region corresponding to an additional genomic region of the reference genome having at least the threshold cytosine-guanine content and having at least the threshold amount of methylated cytosine in the subject in which cancer was detected and in the additional subject in which cancer was not detected; determining, by the computing system, a metric for the single classification region among the more than one classification regions based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control regions; generating, by the computing system, training data comprising the metric for the single classification region of the more than one classification regions of the training sequence reads of a sample from a training subject; as well as One or more machine learning algorithms are implemented by the computing system and using the training data to generate a model to determine an indication of the presence of cancer in a subject based on the amount of methylated cytosine in at least a portion of the more than one classification regions, the model comprising weights for individual classification regions of the more than one classification regions, and at least a portion of the weights for the individual classification regions are different from each other.
2. The method according to claim 1, comprising: obtaining, by the computing system, test sequence data from an additional subject not included in the more than one subject, the test sequence data comprising test sequencing reads of a sample derived from the additional subject, a single test sequencing read comprising a nucleotide sequence corresponding to a nucleic acid fragment included in the additional sample, and the single test sequencing read corresponding to a molecule having a threshold amount of methylated cytosine included in a region of the nucleotide sequence having at least the threshold cytosine-guanine content; and Using the model and the additional sequence data, determining an indication of the presence of cancer in the additional subject.
3. The method according to claim 2, comprising: analyzing, by the computing system, the test sequencing reads to determine a first quantitative measure derived from the test sequencing reads corresponding to the single one of the more than one classification regions; analyzing, by the computing system, the test sequencing reads to determine a second quantitative measure derived from the test sequencing reads corresponding to a single control region of the more than one control regions; determining, by the computing system, the metric for the single classified region based on the first quantitative measure for the single classified region and the second quantitative measures for the more than one control regions; as well as generating, by the computing system, an input vector comprising a metric for the single classification region; Wherein the model uses the input vector to determine an indication of the presence of cancer in the additional subject.
4. The method according to claim 2 or 3, wherein: The one or more machine learning algorithms include one or more classification algorithms; and The indication of the presence of cancer corresponds to a probability of the presence of cancer in the further subject.
5. The method of claim 2 or 3, wherein the one or more machine learning algorithms include one or more regression algorithms; and The indication corresponds to an estimate of the tumor fraction of the further sample.
6. The method of claim 5, wherein the training sequencing reads constitute a first portion of the training sequence data and further training sequencing reads constitute a second portion of the training sequence data, wherein the further training sequencing reads are different from the training sequencing reads; and The method comprises: analyzing, by the computing system, at least one of the first portion of the training sequence data or the second portion of the training sequence data to determine a single frequency of more than one variant present in a single sample of the more than one samples; determining, by the computing system and for the single sample, a variant among the more than one variants having a maximum frequency, the maximum frequency corresponding to a single frequency having a maximum value among the single frequencies derived from the single sample; and A single measure of tumor fraction for a single sample is determined by the computing system based on a maximum value of a single frequency derived from the single sample.
7. The method according to claim 6, wherein: The training data comprises a single measure of tumor fraction for a single sample of the more than one samples; and The model is generated based on a single measure of tumor fraction for a single sample of the more than one samples.
8. The method according to any one of claims 1-7, wherein the metric of the single classification region is determined based on a scaling factor and an error correction factor.
9. The method of any one of claims 1-8, wherein the more than one classification regions individually correspond to genomic regions wherein the cytosine methylation rate in a nucleic acid genomic region derived from a cell obtained from a subject in which cancer is present is different from the cytosine methylation rate in a nucleic acid genomic region derived from a cell obtained from a subject in which cancer is not present.
10. The method of claim 9, wherein the more than one classification regions correspond to first more than one classification regions for a first cancer type, and the model is capable of being generated for a second cancer type based on second more than one classification regions different from the first more than one classification regions.
11. The method of any one of claims 1-10, wherein the more than one sample and the additional sample comprise cell-free nucleic acid.
12. The method according to any one of claims 1 to 11, comprising: The computing system performs a training process using the training data to generate the model, wherein the training process includes: One or more additional weights are determined, by the computing system, for a single sample based on the indication of cancer for the single sample included in the training data being within a threshold confidence level.
13. The method of claim 12, wherein the indication of cancer for a single sample is outside the threshold confidence level, and the method comprises: A penalty is applied by the computing system to the weight of the single sample during the training process.
14. The method according to claim 12 or 13, comprising: performing, by the computing system and using the one or more machine learning algorithms, one or more first iterations of the training process for the model using a portion of the training data; as well as First output data of the model is generated by the computing system based on the one or more first iterations of the training process, the first output data corresponding to one or more first additional indications of the presence of cancer in a first single subject among the more than one subjects, the first single subject corresponding to the portion of the training data.
15. The method according to claim 14, comprising: combining, by the computing system, the first output data and the training data to generate additional training data; performing, by the computing system, one or more second iterations of the training process for the model using a portion of the additional training data; as well as Second output data of the model is generated by the computing system based on the one or more second iterations of the training process, the second output data indicating one or more second additional indications of the presence of cancer in a second single subject among the more than one subjects, the second single subject corresponding to the portion of the additional training data. 16 . The method of claim 15 , wherein the weight of a single classification region of the more than one classification regions is determined based on the first output data and the second output data.
17. The method according to any one of claims 12 to 16, comprising: determining, by the computing system, that a number of indications of the presence of cancer determined during one or more iterations of the training process is at least a threshold for one or more samples included in the training data; as well as A determination by the computing system that modification of one or more weights of the model is not modified or is modified by a minimal amount.
18. The method according to claim 17, comprising: determining, by the computing system, that an additional number of indications of the presence of cancer determined during the one or more iterations of the training process is less than the threshold for one or more additional samples included in the training data; as well as A determination is made by the computing system that modification of one or more additional weights of the model is modified by an amount that exceeds the minimum amount.
19. The method according to any one of claims 1 to 18, comprising: combining more than one nucleic acid derived from at least one of the blood or tissue of the subject with a solution comprising an amount of a methyl binding domain (MBD) protein to produce a nucleic acid-MBD protein solution; as well as The nucleic acid-MBD protein solution is washed more than once with a salt solution to produce a plurality of nucleic acid fractions, a single nucleic acid fraction having a threshold number of methylated cytosines in a region of the more than one nucleic acids having at least the threshold cytosine-guanine content.
20. The method of claim 19, wherein the washing in the more than one washing is performed with a solution having a certain concentration of sodium chloride (NaCl) and produces nucleic acid fractions having a certain range of binding strengths to the MBD protein among the plurality of nucleic acid fractions.
21. The method according to claim 19 or 20, comprising: determining that a first nucleic acid fraction is associated with a first partition of the more than one nucleic acid partitions, the first partition corresponding to a first range of binding intensities to an MBD protein; attaching a first molecular barcode to nucleic acid of the first nucleic acid fraction, the first molecular barcode being included in a first set of molecular barcodes associated with the first partition; determining that a second nucleic acid fraction is associated with a second partition of the more than one nucleic acid partitions, the second partition corresponding to a second range of binding energies to the MBD protein that is different than the first range of binding strengths to the MBD protein; as well as A second molecular barcode is attached to the nucleic acid of the second nucleic acid fraction, the second molecular barcode being included in a second set of molecular barcodes associated with the second partition.
22. The method according to any one of claims 19 to 21, comprising: combining at least a portion of the plurality of nucleic acid fractions with an amount of a restriction enzyme that cleaves molecules having one or more unmethylated cytosines to produce at least a portion of the more than one samples used to generate the sequencing reads; wherein the threshold amount of methylated cytosine corresponds to a minimum frequency of methylated cytosine in a region having at least the threshold cytosine-guanine content.
23. The method according to any one of claims 19 to 21, comprising: combining at least a portion of the plurality of nucleic acid fractions with an amount of a restriction enzyme that cleaves molecules having one or more methylated cytosines to produce at least a portion of the more than one samples used to generate the sequencing reads; wherein the threshold amount of unmethylated cytosines corresponds to a maximum frequency of methylated cytosines that are not cleaved within a region having at least the threshold cytosine-guanine content.
24. The method of any one of claims 1-23, wherein the limit of detection of the model used to determine the tumor fraction of a sample is no greater than 0.05%.
25. A computing system, the computing system comprising: one or more hardware processors; and One or more non-transitory computer-readable storage media including computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations including: obtaining training sequence data, the training sequence data comprising training sequencing reads from more than one sample derived from more than one subject, a single training sequencing read comprising a nucleotide sequence corresponding to a nucleic acid fragment included in a sample of the more than one sample, and a single training sequencing read corresponding to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content of the nucleotide sequence; analyzing the training sequenced reads to determine a first quantitative measure corresponding to a single one of more than one classification regions derived from the training sequenced reads, at least a portion of the single one of more than one classification regions corresponding to a genomic region of a reference genome having the threshold amount of methylated cytosine and having at least the threshold cytosine-guanine content in a subject in which cancer was detected; analyzing the training sequenced reads to determine a second quantitative measure corresponding to more than one control region derived from the training sequenced reads, a single control region of the more than one control region corresponding to an additional genomic region of the reference genome having at least the threshold cytosine-guanine content and having at least the threshold amount of methylated cytosine in the subject in which cancer was detected and in the additional subject in which cancer was not detected; determining a metric for the single classification region among the more than one classification regions based on the first quantitative measure for the single classification region and the second quantitative measure for the more than one control regions; generating training data comprising the metric for the single one of the more than one classification regions of the training sequence reads from a sample of a training subject; One or more machine learning algorithms are implemented using the training data to generate a model to determine an indication of the presence of cancer in a subject based on the amount of methylated cytosine in at least a portion of the more than one classification regions, the model comprising weights for individual classification regions of the more than one classification regions, and at least a portion of the weights for the individual classification regions are different from each other.
26. The computing system of claim 25, wherein the one or more non-transitory computer-readable storage media store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: obtaining test sequence data from an additional subject not included in the more than one subjects, the test sequence data comprising test sequencing reads of a sample derived from the additional subject, a single test sequencing read comprising a nucleotide sequence corresponding to a nucleic acid fragment included in the additional sample, and the single test sequencing read corresponding to a molecule having a threshold amount of methylated cytosine in a region of the nucleotide sequence having at least the threshold cytosine-guanine content; and Using the model and the additional sequence data, determining an indication of the presence of cancer in the additional subject.
27. The computing system of claim 26, wherein the one or more non-transitory computer-readable storage media store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: analyzing the test sequencing reads to determine a first quantitative measure derived from the test sequencing reads corresponding to a single classification region of the more than one classification regions; analyzing the test sequencing reads to determine a second quantitative measure derived from the test sequencing reads corresponding to a single control region of the more than one control regions; determining a metric for the single classified region based on the first quantitative measure for the single classified region and the second quantitative measures for the more than one control regions; as well as generating an input vector comprising a measure of the single classification region; Wherein the model uses the input vector to determine an indication of the presence of cancer in the additional subject.
28. A computing system according to claim 26 or 27, wherein: The one or more machine learning algorithms include one or more classification algorithms; and The indication of the presence of cancer corresponds to a probability of the presence of cancer in the further subject.
29. The computing system of claim 26 or 27, wherein the one or more machine learning algorithms include one or more regression algorithms; and The indicator corresponds to an estimate of the tumor fraction of the further sample.
30. The computing system of claim 29, wherein the training sequencing reads constitute a first portion of the training sequence data, and a second portion of the training sequence data comprises additional training sequencing reads different from the training sequencing reads; and wherein the one or more non-transitory computer-readable storage media store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations including: analyzing at least one of the first portion of the training sequence data or the second portion of the training sequence data to determine a single frequency of more than one variant present in a single sample of the more than one samples; determining, for the single sample, a variant having a maximum frequency among the more than one variants, the maximum frequency corresponding to a single frequency having a maximum value among the single frequencies derived from the single sample; as well as A single measure of tumor fraction for a single sample is determined based on the maximum of the single frequencies derived from the single sample.
31. The computing system of claim 30, wherein: The training data comprises a single measure of tumor fraction for a single sample of the more than one samples; and The model is generated based on a single measure of tumor fraction for a single sample of the more than one samples.
32. The computing system of any one of claims 25-31, wherein the metric of the single classification region is determined based on a scaling factor and an error correction factor.
33. A computing system according to any one of claims 25-32, wherein the more than one classification regions individually correspond to such genomic regions, wherein the cytosine methylation rate in the nucleic acid genomic region derived from cells obtained from a subject in which cancer is present is different from the cytosine methylation rate in the nucleic acid genomic region derived from cells obtained from a subject in which cancer is not present.
34. The computing system of claim 33, wherein the one or more classification regions correspond to first one or more classification regions for a first cancer type, and the model is generable for a second cancer type based on a second one or more classification regions that are different from the first one or more classification regions.
35. The computing system of any one of claims 25-34, wherein the more than one sample and the additional sample comprise cell-free nucleic acid.
36. The computing system of any of claims 25-35, wherein the one or more non-transitory computer-readable storage media store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: Performing a training process using the training data to generate the model, wherein the training process comprises: One or more additional weights for the single sample included in the training data are determined based on the indication of cancer for the single sample being within a threshold confidence level.
37. The computing system of claim 36, wherein the indication of cancer for a single sample is outside the threshold confidence level; and wherein the one or more non-transitory computer-readable storage media store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations including: Penalties are applied to the weights of the individual samples during the training process.
38. The computing system of claim 26 or 37, wherein the one or more non-transitory computer-readable storage media store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: Using the one or more machine learning algorithms, performing one or more first iterations of the training process for the model using a portion of the training data; and First output data of the model is generated based on the one or more first iterations of the training process, the first output data corresponding to one or more first additional indications of the presence of cancer in a first single subject among the more than one subjects, the first single subject corresponding to the portion of the training data.
39. The computing system of claim 38, wherein the one or more non-transitory computer-readable storage media store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: combining the first output data and the training data to generate further training data; performing one or more second iterations of the training process for the model using a portion of the additional training data; and Generating second output data of the model based on the one or more second iterations of the training process, the second output data indicating one or more second additional indications of the presence of cancer in a second single subject among the more than one subjects, the second single subject corresponding to the portion of the additional training data.
40. The computing system of claim 39, wherein the weight of a single classification region of the more than one classification regions is determined based on the first output data and the second output data.
41. The computing system of any of claims 36-40, wherein the one or more non-transitory computer-readable storage media store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: determining that the number of indications of the presence of cancer determined during one or more iterations of the training process is at least a threshold for one or more samples included in the training data; and A determination is made that modification of one or more weights of the model is not modified or is modified by a minimal amount.
42. The computing system of claim 41, comprising: determining that an additional number of indications of the presence of cancer determined during the one or more iterations of the training process is less than the threshold for one or more additional samples included in the training data; as well as A determination is made that modification of one or more additional weights of the model is modified by an amount exceeding the minimum amount.
43. A method comprising: obtaining, by a computing system having one or more hardware processors and memory, sequencing reads derived from a sample obtained from a subject, a single sequencing read comprising a nucleotide sequence corresponding to a nucleic acid fragment included in the sample and corresponding to a molecule having a threshold amount of methylated cytosine in a region having at least a threshold cytosine-guanine content included in the nucleotide sequence; determining, by the computing system, a first quantitative measure corresponding to a single one of the more than one classification regions derived from the sequencing reads, at least a portion of the single one of the more than one classification regions corresponding to a genomic region of a reference genome having the threshold amount of methylated cytosine and having at least the threshold cytosine-guanine content in a subject in which cancer was detected; analyzing, by the computing system, the sequencing reads to determine a second quantitative measure derived from the sequencing reads corresponding to more than one control region, a single control region of the more than one control region corresponding to an additional genomic region of the reference genome having at least the threshold cytosine-guanine content and having at least the threshold amount of methylated cytosine in the subject in which cancer was detected and in the additional subject in which cancer was not detected; determining, by the computing system, more than one metric based on the first quantitative measure of the single classified region and the second quantitative measure of the more than one control regions, wherein a single metric of the more than one metric corresponds to a single classified region of the more than one classified regions; as well as An indication of the presence of cancer in the subject is determined by the computing system based on at least a portion of the more than one metrics.
44. The method of claim 43, comprising: determining, by the computing system and using the sequencing data, a distribution of sequence representations of differentially methylated regions; determining, by the computing system, that at least a threshold amount of the sequence representations included in the distribution overlap with a sub-region of the differentially methylated region; as well as The sub-region of the differentially methylated region is determined by the computing system to be a classified region among the more than one classified regions.
45. The method of claim 44, wherein the threshold amount of sequence representations is at least about 70% of the sequence representations included in the distribution.
46. The method according to any one of claims 43 to 45, comprising: determining, by the computing system, an order of the values of the more than one metrics; as well as determining, by the computing system, a subset of classification regions from among the more than one classification regions based on the order; Wherein the portion of the subset of the more than one metrics corresponding to the classification region is used to determine an indication of the presence of cancer in the subject.
47. The method of any one of claims 43-46, wherein the indication that cancer is present in the subject is an initial indication that cancer is present in the subject, and the method comprises: A scaling factor is applied by the computing system to the initial indication of the presence of cancer in the subject to determine a modified indication of the presence of cancer in the subject.
48. The method of any one of claims 43-47, wherein the indication of the presence of cancer in the subject corresponds to a tumor score.
49. The method of any one of claims 43-48, wherein the sample is a first sample collected at least one of before or at the start of cancer treatment, and the method comprises: obtaining, by the computing system having one or more hardware processors and memory, additional sequencing reads derived from a second sample obtained from the subject, the individual additional sequencing reads comprising additional nucleotide sequences corresponding to nucleic acid fragments included in the second sample and corresponding to additional molecules having the threshold amount of methylated cytosine included in a region having at least the threshold cytosine-guanine content in the additional nucleotide sequences; determining, by the computing system, an additional first quantitative measure corresponding to a single one of the more than one classification regions derived from the additional sequencing reads, at least a portion of the single one of the more than one classification regions corresponding to a genomic region of the reference genome having the threshold amount of methylated cytosine and having at least the threshold cytosine-guanine content in a subject in which cancer was detected; analyzing, by the computing system, the additional sequencing reads to determine an additional second quantitative measure corresponding to more than one control region derived from the additional sequencing reads, a single control region of the more than one control region corresponding to an additional genomic region of the reference genome having at least the threshold cytosine-guanine content and having at least the threshold amount of methylated cytosine in the subject in which cancer was detected and in additional subjects in which cancer was not detected; determining, by the computing system, more than one additional metric based on the additional first quantitative measure of the single classified region and the additional second quantitative measure of the more than one control region, wherein a single additional metric of the more than one additional metrics corresponds to a single classified region of the more than one classified regions; as well as An additional indication of the presence of cancer in the subject is determined by the computing system based on at least a portion of the more than one additional metrics.
50. The method of claim 49, wherein the second sample is obtained at least one week after the cancer therapy is administered to the subject.
51. The method of claim 49 or 50, comprising: The indication of the presence of cancer in the subject is analyzed by the computing system relative to the additional indication of the presence of cancer in the subject to determine the subject's response to the treatment.
52. A method comprising: obtaining, by a computing system having one or more hardware processors and memory, test sequence data from a subject, the test sequence data comprising test sequencing reads of a sample derived from the subject, a single test sequencing read comprising a nucleotide sequence corresponding to a nucleic acid fragment included in the additional sample, and a single test sequencing read corresponding to a molecule having a threshold amount of methylated cytosine in a region of the nucleotide sequence having at least the threshold cytosine-guanine content; analyzing, by the computing system, the test sequencing reads to determine a first quantitative measure derived from the test sequencing reads corresponding to a single one of more than one classification regions, at least a portion of the single one of the more than one classification regions corresponding to a genomic region of a reference genome having the threshold amount of methylated cytosine and having at least the threshold cytosine-guanine content in a subject in which cancer was detected; analyzing, by the computing system, the test sequencing reads to determine a second quantitative measure derived from the test sequencing reads corresponding to a single control region of more than one control region, the single control region of the more than one control region corresponding to an additional genomic region of the reference genome having at least the threshold cytosine-guanine content and having at least the threshold amount of methylated cytosine in the subject in which cancer was detected and in the additional subject in which cancer was not detected; determining, by the computing system, a metric for the single classified region based on the first quantitative measure for the single classified region and the second quantitative measures for the more than one control regions; as well as generating, by the computing system, an input vector comprising a metric for the single classification region; The indication of the presence of cancer in the subject is determined by the computing system by providing the input vector to a model that implements one or more machine learning techniques to generate an indication of the presence of cancer in the subject, the model including weights for individual classification regions of the more than one classification regions, and at least a portion of the weights for the individual classification regions are different from each other.
53. The method of claim 52, comprising: obtaining, by the computing system having one or more hardware processors and memory, training sequence data, the training sequence data comprising training sequencing reads from more than one sample derived from more than one training subject, a single training sequencing read comprising a nucleotide sequence corresponding to a nucleic acid fragment included in a sample of the more than one sample, and a single training sequencing read corresponding to a molecule having a threshold amount of methylated cytosine in a region of the nucleotide sequence having at least a threshold cytosine-guanine content; analyzing, by the computing system, the training sequence reads to determine an additional first quantitative measure derived from the training sequence reads corresponding to a single classification region of the more than one classification regions; analyzing, by the computing system, the training sequence reads to determine additional second quantitative measures corresponding to more than one control region derived from the training sequence reads; determining, by the computing system, an additional metric for the single classification region of the more than one classification regions based on the additional first quantitative measure of the single classification region and the additional second quantitative measure of the more than one control regions; generating, by the computing device, training data comprising the additional metric for the single classification region of the more than one classification regions of the training sequence reads from samples of the more than one training subjects; One or more machine learning algorithms are implemented by the computing system and using the training data to generate the model to determine an indication of the presence of cancer in the subject based on the amount of methylated cytosine in at least a portion of the more than one classification regions.
54. The method of claim 52 or 53, wherein: The one or more machine learning algorithms include one or more classification algorithms; and The indication of the presence of cancer corresponds to a probability of the presence of cancer in the subject.
55. The method of claim 52 or 53, wherein the one or more machine learning algorithms include one or more regression algorithms; and The indication corresponds to an estimate of the tumor fraction of the sample.
56. The method of claim 55, wherein the training sequencing reads constitute a first portion of the training sequence data and further training sequencing reads constitute a second portion of the training sequence data, wherein the further training sequencing reads are different from the training sequencing reads; and The method comprises: analyzing, by the computing system, at least one of the first portion of the training sequence data or the second portion of the training sequence data to determine a single frequency of more than one variant present in a single sample of the more than one samples; determining, by the computing system and for the single sample, a variant among the more than one variants having a maximum frequency, the maximum frequency corresponding to a single frequency having a maximum value among the single frequencies derived from the single sample; and A single measure of tumor fraction for a single sample is determined by the computing system based on a maximum value of a single frequency derived from the single sample.
57. The method of claim 56, wherein: The training data comprises a single measure of tumor fraction for a single sample of the more than one samples; and The model is generated based on a single measure of tumor fraction for a single sample of the more than one samples.
58. The method of any one of claims 52-57, wherein the sample of the subject and the more than one samples of the more than one training subjects comprise cell-free nucleic acid.
59. A method comprising: obtaining, by a computing system having one or more hardware processors and a memory, sequencing data from more than one subject, the sequencing data comprising sequencing reads from more than one sample derived from the more than one subject, a single sequencing read comprising a nucleotide sequence corresponding to a nucleic acid fragment included in the additional sample, and a single sequencing read corresponding to a molecule having a threshold amount of methylated cytosine in a promoter region having at least a threshold cytosine-guanine content included in the nucleotide sequence; analyzing the sequencing reads by the computing system to determine a first quantitative measure corresponding to the promoter region derived from the sequencing reads; analyzing, by the computing system, the sequencing reads to determine a second quantitative measure derived from the sequencing reads corresponding to a single control region of more than one control region, the single control region of the more than one control region corresponding to an additional genomic region of the reference genome having at least the threshold cytosine-guanine content and having at least the threshold amount of methylated cytosine in the subject in which cancer was detected and in the additional subject in which cancer was not detected; determining, by the computing system, a metric for the promoter region based on the first quantitative measure of the single classified region and the second quantitative measure of the more than one control regions; An indication of a methylation state of the promoter region is generated by the computing system based on the metric having at least a threshold value.
60. A computing system, the computing system comprising: one or more hardware processors; and One or more non-transitory computer-readable storage media including computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations including: obtaining sequencing data from more than one subject, the sequencing data comprising sequencing reads from more than one sample derived from the more than one subject, a single sequencing read comprising a nucleotide sequence corresponding to a nucleic acid fragment included in the additional sample, and a single sequencing read corresponding to a molecule having a threshold amount of methylated cytosine in a promoter region having at least a threshold cytosine-guanine content included in the nucleotide sequence; analyzing the sequenced reads to determine a first quantitative measure corresponding to the promoter region derived from the sequenced reads; analyzing the sequencing reads to determine a second quantitative measure derived from the sequencing reads corresponding to a single control region of more than one control region, the single control region of the more than one control region corresponding to an additional genomic region of the reference genome having at least the threshold cytosine-guanine content and having at least the threshold amount of methylated cytosine in the subject in which cancer was detected and in the additional subjects in which cancer was not detected; determining a metric for the promoter region based on the first quantitative measure of the single classification region and the second quantitative measure of the more than one control region; An indication of a methylation status of the promoter region is generated based on the metric having at least a threshold value.
