Estimating tumor purity from single samples

Abstract

This disclosure provides a method for predicting tumor purity from tumor samples without using matched normal controls. A set of genomic regions is identified based on nucleic acid sequence data aligned to a reference genome. Each of these genomic regions comprises one or more nucleotide sequence variants relative to a corresponding genomic region of the reference genome. The B allele frequency distribution of the biological sample is determined based on the B allele frequency of each of the identified genomic regions. The B allele frequency distribution is processed using a trained machine learning model to predict a measure of tumor purity identified in the biological sample.

Description

[0001] Cross-references to related applications

[0002] This application claims priority to U.S. Provisional Patent Application No. 62 / 931,096, filed November 5, 2019, which is incorporated herein by reference in its entirety for all purposes. Technical Field

[0003] This disclosure generally relates to systems and methods for predicting tumor purity from a single sample. More specifically, but not in a limiting way, this disclosure relates to predicting tumor purity of a biological sample by using a trained machine learning model to process the distribution of B allele frequencies. Background Technology

[0004] Tumor cytogenetics, also known as "tumor purity," identifies the proportion of cancer cells in a sample. Accurate estimation of tumor purity in a biological sample can help improve the accuracy of detecting somatic mutation levels and / or copy number variations. This is because tumor purity indicates the allele frequency of somatic mutations present in the biological sample. The detection of somatic mutations and copy number variations can then be used to determine the cancer stage of a subject or to assess the effectiveness of a specific cancer treatment. Therefore, tumor purity can inform the determination of cancer stage and / or the assessment of treatment efficacy.

[0005] While tumor purity can be an effective indicator, it can also be a confounding variable in several bioinformatics analyses. For example, routine techniques for estimating tumor purity may require pathologists to perform a histopathological assessment by manually examining sample images. However, histopathological assessments, including manual examination of sample images, can be subjective and inaccurate. Other routine techniques for estimating tumor purity require comparing values ​​of nucleic acid sequence data from a given tumor sample (e.g., putative somatic variations) with other values ​​of nucleic acid sequencing data from a matched normal control sample. However, such a normal control sample may not be available.

[0006] For example, conventional techniques estimate the tumor purity of a sample as a function of the allelic portion of somatic mutations specific to the individual tumor. In the absence of a matching normal sample, the identification of these somatic mutations is less precise, and the accuracy of the estimated purity is significantly reduced. In some cases, a matching normal control may not be available if the sample provider has not collected or sequenced a normal control (e.g.).

[0007] Therefore, it is necessary to accurately estimate the tumor purity in samples to facilitate detection, without relying on subjective analysis (e.g., histopathological assessment) or the presence of normal control samples. Invention Overview

[0009] In some implementations, methods for estimating tumor purity are provided. These methods may include obtaining nucleic acid sequence data of multiple nucleic acid molecules representing a target tumor sample. The methods may also include aligning the nucleic acid sequence data with a reference genome. Furthermore, the methods may include identifying a set of genomic regions based on the aligned nucleic acid sequence data. In some cases, each of these genomic regions comprises one or more nucleotide sequence variants relative to a corresponding genomic region of the reference genome.

[0010] The method may further include determining the B allele frequency for each genomic region of the set of genomic regions. The method may further include determining the B allele frequency distribution of the biological sample based on the B allele frequencies of the set of genomic regions. The method may further include processing the B allele frequency distribution using a trained machine learning model to estimate a metric for identifying tumor purity in the biological sample. The method may further include an output metric.

[0011] In some implementations, a system is provided that includes one or more data processors and a non-transitory computer-readable storage medium containing instructions that, when executed on the one or more data processors, cause the one or more data processors to perform part or all of the methods disclosed herein.

[0012] In some implementations, a computer program product is provided, which is tangibly embodied in a non-transitory machine-readable storage medium and includes instructions configured to cause one or more data processors to perform part or all of the methods disclosed herein.

[0013] Some embodiments of this disclosure include a system comprising one or more data processors. In some embodiments, the system includes a non-transitory computer-readable storage medium containing instructions that, when executed on the one or more data processors, cause the one or more data processors to perform part or all of one or more methods disclosed herein, and / or part or all of one or more processes disclosed herein. Some embodiments of this disclosure include a computer program product tangibly embodied in a non-transitory machine-readable storage medium, including instructions configured to cause one or more data processors to perform part or all of one or more methods disclosed herein and / or part or all of one or more processes disclosed herein.

[0014] The terms and expressions used are descriptive rather than restrictive, and their use is not intended to exclude any equivalents of the features shown and described or any part thereof, but it should be recognized that various modifications may be made within the scope of the claimed invention. Therefore, it should be understood that although the claimed invention has been specifically disclosed by way of embodiments and optional features, modifications and variations of the concepts disclosed herein can be made by those skilled in the art, and such modifications and variations are considered to be within the scope of the invention as defined by the appended claims.

[0015] Brief description of the attached figures

[0016] The features, embodiments, and advantages of this disclosure will be better understood when the following detailed description is read in conjunction with the accompanying drawings.

[0017] Figure 1 A schematic diagram of a fully connected neural network for predicting tumor purity is shown according to some implementation schemes.

[0018] Figure 2 A schematic diagram of a one-dimensional convolutional neural network for predicting tumor purity is shown according to some implementation schemes.

[0019] Figure 3 A schematic diagram of a two-dimensional convolutional neural network for predicting tumor purity is shown according to some implementation schemes.

[0020] Figure 4A The loss of fully connected neural networks, one-dimensional convolutional neural networks, and two-dimensional convolutional neural networks trained on VCF-level datasets according to some implementation schemes is shown relative to epoch plots.

[0021] Figure 4B The data show the root mean square error (RSME) and mean absolute error (MAE) of fully connected neural networks, one-dimensional convolutional neural networks, and two-dimensional convolutional neural networks trained on VCF-level datasets according to some implementation schemes.

[0022] Figure 4C The root mean square error (RSME) and mean absolute error (MAE) data for testing FASTQ-level datasets are shown according to some implementation schemes.

[0023] Figure 5A Figure 500A shows a comparison of true tumor purity with tumor purity predicted by a trained fully connected neural network according to some implementation schemes.

[0024] Figure 5B Figure 500B shows a comparison of the true tumor purity with the tumor purity predicted by a trained two-dimensional convolutional neural network according to some implementation schemes.

[0025] Figure 5C Figure 500C shows a comparison of true tumor purity with tumor purity predicted by a one-dimensional convolutional neural network according to some implementation schemes.

[0026] Figure 6 An example of a schematic diagram illustrating the use of a trained machine learning model to estimate tumor purity according to some implementation schemes is shown.

[0027] Figures 7A to 7E Examples of BAF distributions plotted according to some implementation schemes are provided.

[0028] Figure 8 Figure 800 shows the first two main components of classifying biological samples using B allele frequency characteristics according to some implementation schemes.

[0029] Figure 9 The included flowchart illustrates examples of methods for estimating the tumor purity of biological samples according to certain implementation schemes.

[0030] Figure 10 Examples of computer systems for implementing some of the embodiments disclosed herein are shown. Invention Details

[0032] I. Overview

[0033] To at least address the aforementioned shortcomings of conventional systems, this technique can be used to predict tumor purity by processing the B allele frequency distribution of heterozygous sites in sequencing data using a trained machine learning model. The trained machine learning model can generate a predicted metric that can identify tumor purity in a biological sample, even in the absence of matching sequencing data from a normal control sample. Therefore, this technique can accurately predict tumor purity in biological samples without a matching normal control, which contributes to more accurate analysis of nucleic acid sequencing data from tumor-only samples.

[0034] Nucleic acid sequence data of multiple nucleic acid molecules representative of a tumor sample can be obtained. In some embodiments, the nucleic acid sequence data is whole-exome sequencing data. Nucleic acid sequence data can be whole-genome sequencing data. In some embodiments, sequencing data is derived from a tumor sample. The tumor sample can be derived from a human subject. Nucleic acid sequence data can be generated from shotgun sequencing. In some embodiments, nucleic acid sequence data can be generated by sequencing selected portions of the genome or exome.

[0035] Nucleic acid sequence data can be aligned with a reference genome. Based on the aligned nucleic acid sequence data, a set of genomic regions can be identified. In some cases, each of these genomic regions includes one or more nucleotide sequence variants relative to the corresponding genomic region of the reference genome. To identify nucleotide sequence variants, candidate variants can be identified, and the reference and substitution read depths of the candidate variants can be calculated.

[0036] The allele frequencies of B and B in each genomic region of this set of genomic regions can be determined. The B allele frequency is a normalized measure of the ratio of allele strength (BAF) of two alleles (A and B), such that a BAF of 1 or 0 indicates the complete absence of one of the two alleles (e.g., AA or BB), and a BAF of 0.5 indicates the equal presence of both alleles (e.g., AB). Based on the B allele frequencies of this set of genomic regions, the B allele frequency distribution of a biological sample can be determined. In some implementations, the B allele frequency distribution is normalized.

[0037] Trained machine learning models can be used to process B allele frequency distributions to predict metrics for identifying tumor purity in biological samples. As used herein, tumor purity or tumor cytogenetics refers to the proportion of cancer cells in a tumor sample. Tumor purity can be a metric used by various techniques to predict tumor characteristics corresponding to directly clinical relevance. For example, predicted tumor purity may directly alter the threshold of evidence required to identify copy number alterations, including those associated with approved companion diagnostics (e.g., EGFR exon 19 deletion in NSCLC). By providing prior information about the expected allele portion of such variants, tumor purity estimates can also be used to improve the quality (sensitivity, PPV) of somatic variant calls. For example, a 60% tumor purity value may indicate that a tumor sample comprises 60% tumor cells and 40% normal cells. Alternatively or optionally, tumor purity can indicate the number of cancer cells in a tumor sample. It should be noted that the term "tumor cytogenetics" is used interchangeably throughout this document whenever the term "tumor purity" is mentioned.

[0038] A trained machine learning model can be trained on a computer-generated dataset (e.g., nucleic acid sequence data generated via computer simulation). In some implementations, the trained machine learning model is trained on a training dataset generated from sequencing data derived from tumor cells diluted in normal cells. The trained machine learning model can achieve a mean absolute error of less than about 0.2. In some implementations, the trained machine learning model achieves a root mean square error of less than about 0.2.

[0039] In some implementations, the trained machine learning model includes a fully connected neural network. A fully connected network may include fully connected layers with rectified linear unit (ReLU) activation functions. In some implementations, the output activation function of the fully connected neural network is a sigmoid function. The loss function of the fully connected neural network may be configured to compute mean squared error (MSE). In some implementations, the fully connected neural network is tuned via hyperparameter search using random sampling, wherein it involves linear search on layers, linear search on size, logarithmic search on learning rate, or a combination thereof.

[0040] In some cases, the trained machine learning model includes a one-dimensional convolutional neural network. The one-dimensional convolutional neural network can be configured such that the B allele frequency distribution can be used as input, and can be encoded as an input of size 25 in height, 1 in width, and 100 in depth. In some implementations, each layer of the one-dimensional convolutional neural network performs an lx1 convolution followed by a ReLU activation function.

[0041] In some implementations, the trained machine learning model comprises a two-dimensional convolutional neural network (2D convolutional neural network). In some implementations, the B allele frequency distribution of the 2D convolutional neural network is used as input and encoded into an input size defined by a height of 25, a width of 100, and a depth of 1. In some implementations, each convolutional layer of the 2D convolutional neural network is followed by a ReLU layer. The output of the 2D convolutional neural network can be a densely connected layer with a sigmoid activation function. Furthermore, the hyperparameters of the 2D convolutional neural network can be tuned by adjusting the number of layers, filter size, number of filters, or a combination thereof.

[0042] An estimated metric for identifying tumor purity can be output. For example, a report including the estimated metric can be output. In some embodiments, the report includes information identifying the distribution of B allele frequencies. The report may also include information identifying at least one diagnostic biomarker and / or at least one prognostic biomarker. In some embodiments, the report includes information identifying predicted somatic variants. The report may also include treatment recommendations. For example, estimated tumor purity may indicate increased methylation levels of O-6-methylguanine-DNA methyltransferase, a prognostic biomarker for glioblastoma. In another instance, estimated tumor purity can be used to determine the mutational burden of cancer cells, which can then be used to determine certain types of immunotherapy. Depending on the amount of estimated tumor purity, it may be recommended whether treatment for glioblastoma should be initiated. In some embodiments, treatment recommendations include recommendations to administer treatment to human subjects. Treatment recommendations may include recommendations not to administer treatment to human subjects.

[0043] Therefore, the embodiments of this disclosure offer a technical advantage over conventional systems by more accurately estimating tumor purity without relying on data derived from matched normal control samples. The estimated tumor purity can be used to improve the accuracy of sequencing data analysis and annotation from tumor samples. One or more reports can be generated describing the estimated tumor purity (e.g., diagnostic and / or prognostic reports). For example, the estimation of tumor purity can be used to improve the accuracy of diagnostic techniques to identify somatic mutations and / or copy number variations, and reports with detailed information on predicted somatic mutations and / or copy number variations can be generated.

[0044] Based on estimates of tumor purity and / or reports resulting from such estimates, patients may be given one or more treatments or not treated at all. For example, predicted somatic variants can be compared with one or more databases of known cancer mutations to diagnose or characterize cancer. Variants associated with responsiveness or non-responsiveness to certain cancer treatments can be identified, and treatment recommendations can be provided. Cancer can be treated based on these recommendations.

[0045] The following examples are provided to illustrate certain embodiments. In the following description, specific details are set forth for illustrative purposes to provide a thorough understanding of the embodiments of this disclosure. However, it will be apparent that various embodiments can be practiced without these specific details. For example, apparatus, systems, structures, components, methods, and other components may be shown as components in the form of block diagrams to avoid obscuring the embodiments with unnecessary details. In other cases, well-known apparatuses, processes, systems, structures, and techniques may be shown without the need for necessary details to avoid confusing the embodiments. The accompanying drawings and descriptions are not intended to be limiting. The terms and expressions used in this disclosure are used as descriptive terms rather than limiting terms, and the use of these terms and expressions is not intended to exclude any equivalents of the features shown and described or any part thereof. The term “example” is used herein to mean “as an example, illustration, or description.” Any embodiment or design described herein as an “example” is not necessarily to be construed as preferred or superior to other embodiments or designs.

[0046] II. Machine learning model for predicting tumor purity from a single sample

[0047] A. An exemplary machine learning model for predicting tumor purity in biological samples.

[0048] As discussed in this paper, tumor purity in biological samples can be estimated using trained machine learning models. These trained machine learning models can correspond to one of various machine learning models trained to estimate tumor purity in biological samples. In some implementations, the trained machine learning models include more than one model (e.g., 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 machine learning models). For example, one of three machine learning models can be trained to estimate tumor purity in nucleic acid sequencing data, including fully connected neural networks, one-dimensional convolutional neural networks, and two-dimensional convolutional neural networks. In some cases, the trained machine learning models include deep neural networks. Deep neural networks can be used to capture the internal structure of increasingly large and high-dimensional datasets, such as nucleic acid sequence data. Deep neural networks can identify high-level features, improve the performance of traditional statistical models, increase interpretability, and provide additional understanding of the structure of nucleic acid sequence data.

[0049] A trained machine learning model can include hyperparameters. Hyperparameters can be configurations external to the model, and their values ​​are not estimated from the data (e.g., training data, input data). In some cases, hyperparameters are tuned, for example, to solve a given predictive modeling problem.

[0050] In some cases, hyperparameters are used to help estimate model parameters. Hyperparameters can be specified by the user. In other cases, a set of heuristic algorithms can be used to determine the hyperparameters.

[0051] Figure 1 A schematic diagram 100 of a fully connected neural network for predicting tumor purity is shown according to some embodiments. The left side shows the input layer 105 of the fully connected neural network, followed by a set of hidden layers 110. The right side shows the output layer 115. For a fully connected network, input features may include the whole exome of nucleotide sequence variants identified in a biological sample, and the distribution of B allele frequencies (BAF). The fully connected neural network may consist of a series of fully connected layers with ReLU activation functions, and the output activation function may be a sigmoid function. In some cases, the loss function of the fully connected neural network is configured to generate mean squared error (MSE). The fully connected neural network can be optimized via a hyperparameter search using random sampling, including a linear search of layers and sizes, and a logarithmic search of the learning rate. Each output dimension of the fully connected neural network may depend on each input dimension. In some cases, the fully connected neural network may be a feedforward neural network.

[0052] Convolutional neural networks (CNNs) can be trained to predict tumor purity in biological samples. CNNs can rely on local connections and bound weights between units, followed by feature pooling (subsampling) to obtain translation-invariant descriptors. A basic CNN architecture may include a convolutional and pooling layer, optionally followed by a fully connected layer for supervised prediction. In some cases, CNNs consist of multiple (e.g., >10) convolutional and pooling layers to better model the input space. CNNs may require large datasets for good training. In some implementations, CNNs use fewer parameters than fully connected neural networks by computing convolutions over small regions of the input space and by sharing parameters between regions. CNNs can be one-dimensional. CNNs can be two-dimensional. In some implementations, CNNs include three or more dimensions.

[0053] Figure 2 A schematic diagram 200 of a one-dimensional convolutional neural network (CNN) for predicting tumor purity is shown according to some implementation schemes. For the CNN, input features might include a chromosome BAF distribution, which can be encoded as an input of defined height 25, width 1, and depth 100. Each layer of the CNN can perform an lx1 convolution followed by a ReLU activation function. In some cases, a deep network structure, known as a "network of networks," is used. Effectively, a one-dimensional convolution with a depth span per layer on a flattened input is used. In some cases, the hyperparameters corresponding to the depth and layer size are obtained from a fully connected neural network (see [link to documentation]). Figure 1 At this point, the hyperparameters corresponding to the learning rate of the one-dimensional convolutional neural network can be adjusted for optimization.

[0054] Figure 3 A schematic diagram 300 of a two-dimensional convolutional neural network (2D convolutional neural network) for predicting tumor purity is shown according to some embodiments. For the 2D convolutional neural network, the input features may include the chromosome BAF distribution, which can be encoded as an input of defined height 25, width 100, and depth 1. Each layer may correspond to a 2D convolutional layer (2D convolutional neural network layer) followed by a ReLU activation function. The output layer may correspond to a densely connected layer with a sigmoid activation function. In some embodiments, the hyperparameters for the 2D convolutional neural network are optimized by searching for the number of layers, filter size, and number of filters. Initial overfitting on the training set can be reduced by performing dropout regularization and / or L2 regularization.

[0055] Other types of machine learning models can be trained to estimate the tumor purity of biological samples. In some implementations, the machine learning model corresponds to one or more of gradient boosting decision trees (e.g., XGBoost framework, LightGBM framework), bagging procedures, boosting procedures, and / or random forest algorithms. For example, gradient boosting decision trees can be trained to predict the tumor purity of biological samples. Gradient boosting corresponds to a class of machine learning techniques that can be used for regression and classification problems and to generate predictive models, such as decision trees, that may include a set of weak predictive models. In some cases, gradient boosting decision trees may include, for example, the XGBoost framework or the LightGBM framework.

[0056] B. Training dataset for training a machine learning model to predict tumor purity in biological samples

[0057] Machine learning models used to predict tumor purity in biological samples can be trained using supervised training algorithms. Supervised training algorithms can be used to train machine learning models to generate outputs corresponding to linear and / or logistic regression values ​​for tumor purity. Machine learning models can be trained based on training datasets. In some cases, training datasets include sequencing datasets from pure tumor samples, pure normal samples, and tumor cell dilution series from normal cells. Sequencing datasets can be derived from biological samples of other subjects, including biological samples from subjects diagnosed with cancer. The biological samples in the training dataset can correspond to malignant tissue, benign tissue, or mixtures thereof. In some implementations, biological samples including cancerous tissue are obtained when no matching normal samples are available. Alternatively or optionally, matching normal samples are obtained for training and testing the machine learning model (e.g.).

[0058] In some implementations, the training dataset includes a training nucleotide sequence variant dataset. The training variant dataset may correspond to nucleic acid sequence data in which nucleotide sequence variants have been identified. Nucleotide sequence variants may include single nucleotide polymorphisms (SNPs), one or more single nucleotide variants, insertion-deletion mutations (indels), small insertions, small deletions, structural variant connections, variable-length tandem repeats, flanking sequences, and combinations of the above. In some cases, a biased “variant call format (VCF) level” dataset is generated using an artificial combination of VCF files generated from training samples (including tumor and normal samples). Therefore, the training dataset may include a computer-generated “dilution” of tumor sequences. VCF level datasets can be generated using the features listed in Table 1. To train a machine learning model, the machine learning model can be mounted on the “training” set, hyperparameters can be tuned based on performance on the “training” and “dev” sets, and the final performance can be evaluated on the “test” set.

[0059]

[0060] In some implementations, the training dataset includes training nucleic acid sequence data. This training nucleic acid sequence data can correspond to unbiased "FASTQ-level" datasets, which are generated by downsampling and combining reads from training samples. Training samples can include normal samples and tumor samples. FASTQ-level datasets can generate 99 unique samples and may be derived from in vitro diluted sequencing of tumor cells and matched normal cells. Compared to VCF-level datasets, FASTQ-level datasets can be at the read level rather than the variant level and are therefore closer to the original data. In some cases, FASTQ-level datasets are used to improve or optimize performance using more realistic data. Because generating large FASTQ-level datasets is computationally difficult, transfer learning can be used to train machine learning models on FASTQ-level datasets. Transfer learning is performed using machine learning models pre-trained on VCF-level datasets with 5-fold cross-validation. In some cases, each machine learning model can be trained on 4 / 5 of the training dataset and tested on 1 / 5 of the training dataset in each iteration.

[0061] C. Training and testing of machine learning models for predicting tumor purity in biological samples

[0062] Machine learning models (e.g., fully connected neural networks) can be trained on training datasets (e.g., VCF-level datasets). In some cases, machine learning models are first trained on VCF-level datasets. A machine learning model can be mounted on a "training" set, hyperparameters can be tuned based on performance on both the "training" and "dev" sets, and the final performance can be evaluated on a "test" set. Figure 4A Chart 400A shows the loss versus epoch for fully connected neural networks, one-dimensional convolutional neural networks, and two-dimensional convolutional neural networks trained on a VCF-level dataset according to some implementation schemes. Chart 400A shows that the performance level of each trained machine learning model increases with each epoch, reaching an error value (e.g., mean squared error) ranging from approximately 0.01 to 0.025.

[0063] Figure 4B The image shows 400B of root mean square error (RSME) and mean absolute error (MAE) data for fully connected neural networks, one-dimensional convolutional neural networks, and two-dimensional convolutional neural networks trained on a VCF-level dataset according to several implementation schemes. All models perform comparably on the VCF-level dataset, with RSME values ​​based on the training dataset ranging from 0.081 (two-dimensional convolutional neural network) to 0.088 (fully connected neural network). Figure 4A and Figure 4B As shown, the consistency performance among all machine learning models trained using the VCF-level training and test datasets indicates a good fit, possibly close to the optimal value for predicting the purity of the sample tumors.

[0064] In some implementations, FASTQ-level datasets are used to improve or optimize the performance of trained machine learning models. Since generating large FASTQ-level datasets is computationally difficult, transfer learning can be used to train machine learning models on them. Transfer learning can be performed using machine learning models pre-trained on VCF-level datasets. A 5-fold cross-validation strategy can be used to evaluate the performance of trained machine learning models. In some cases, each machine learning model can be trained on 4 / 5 of the training dataset and tested on 1 / 5 of the training dataset in each iteration. Figure 4C The RSME and MAE data 400C are shown for fully connected neural networks, one-dimensional convolutional neural networks, and two-dimensional convolutional neural networks trained on FASTQ-level datasets according to some implementations. The RSME and MAE data 400C indicate that the performance levels of the trained one-dimensional convolutional neural network (e.g., 0.057 RMSE) and the trained two-dimensional convolutional neural network (e.g., 0.067 RMSE) are comparable. Furthermore, the RSME and MAE data 400C show that the trained machine learning models achieve lower RSME and MAE after transfer learning on the FASTQ-level datasets.

[0065] The performance level of each trained machine learning model was evaluated. Figure 5A Figure 500A shows a comparison of the true tumor purity with the tumor purity predicted by a trained fully connected network according to some implementation schemes. Figure 5B Figure 500B shows a comparison of the true tumor purity with the tumor purity predicted by a trained two-dimensional convolutional neural network according to some implementation schemes. Figure 5C Figure 500C shows a comparison of true tumor purity with tumor purity predicted by a one-dimensional convolutional neural network according to some implementation schemes. (Source: [Original Source Name]) Figure 5A , Figure 5B and Figure 5C The results for each of these studies indicate that the trained machine learning model achieved a performance level comparable to existing tools that require matching normal control samples.

[0066] In some cases, convolutional neural network (CNN) models are prone to overfitting on the training dataset. For example, high RSME and MAE values ​​corresponding to CNNs may not be significantly improved by regularization and dropout. This tendency for CNNs to overfit may indicate that if these machine learning models were trained on a larger training dataset, they would be able to detect complex signals of tumor purity. Therefore, larger and more diverse training datasets can be used to further train machine learning models.

[0067] In some implementations, a trained machine learning model is trained using training and / or test data to achieve one or more predetermined performance levels for predicting tumor purity. The one or more predetermined performance levels include the following:

[0068] • Maximum MAE of 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, or 20%;

[0069] • RMSE of up to 1%, 2%, 3%, 4%, 5%, 6%, 7%, 8%, 9%, 10%, 11%, 12%, 13%, 14%, 15%, 16%, 17%, 18%, 19%, or 20%;

[0070] • At least approximately 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95 or higher accuracy. Alternatively or additionally, the trained machine learning model is trained to achieve an accuracy of approximately 0.2–1.0, 0.2–0.9, 0.2–0.8, 0.2–0.7, 0.2–0.6, 0.2–0.5, 0.2–0.4, 0.2–0.3, 0.3–1.0, 0.3–0.9, 0.3–0.8, 0.3–0.7, 0.3–0.6, 0.3–0.5, 0.3–0.4, 0.4–1.0, 0.4–0.9, 0. Predicting tumor purity with precision values ​​of 0.4-0.8, 0.4-0.7, 0.4-0.6, 0.4-0.5, 0.5-1.0, 0.5-0.9, 0.5-0.8, 0.5-0.7, 0.5-0.6, 0.6-1.0, 0.6-0.9, 0.6-0.8, 0.6-0.7, 0.7-1.0, 0.7-0.9, 0.7-0.8, 0.8-1.0, 0.8-0.9, or 0.9-1.0.

[0071] • At least approximately 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95 or higher for recall. Alternatively or additionally, a trained machine learning model is trained to recall at approximately 0.2–1.0, 0.2–0.9, 0.2–0.8, 0.2–0.7, 0.2–0.6, 0.2–0.5, 0.2–0.4, 0.2–0.3, 0.3–1.0, 0.3–0.9, 0.3–0.8, 0.3–0.7, 0.3–0.6, 0.3–0.5, 0.3–0.4, 0.4–1.0, 0.4–0.9, 0. The tumor purity can be estimated by re-calling values ​​of 4-0.8, 0.4-0.7, 0.4-0.6, 0.4-0.5, 0.5-1.0, 0.5-0.9, 0.5-0.8, 0.5-0.7, 0.5-0.6, 0.6-1.0, 0.6-0.9, 0.6-0.8, 0.6-0.7, -0.7-1.0, 0.7-0.9, 0.7-0.8, 0.8-1.0, 0.8-0.9, or 0.9-1.0.

[0072] • An F1 score of at least approximately 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.86, 0.87, 0.88, 0.89, 0.9, 0.91, 0.92, 0.93, 0.94, 0.95, 0.96, 0.97, 0.98, 0.99, 0.995 or higher (e.g., macro-average F1 classification score). Alternatively or optionally, the values ​​may be approximately 0.2-1.0, 0.2-0.99, 0.2-0.95, 0.2-0.9, 0.2-0.8, 0.2-0.7, 0.2-0.6, 0.2-0.5, 0.2-0.4, 0.2-0.3, 0.3-1.0, 0.3-0.99, 0.2-0.95, 0.3-0.9, 0.3-0.8, or 0.3. -0.7, 0.3-0.6, 0.3-0.5, 0.3-0.4, 0.4-1.0, 0.4-0.99, 0.4-0.95, 0.4-0.9, 0.4-0.8, 0.4-0.7, 0.4-0.6, 0.4-0.5, 0.5-1.0, 0.5-0.99, 0.5-0.95, 0.5-0.9, 0.5-0.8 0.5-0.7, 0.5-0.6, 0.6-1.0, 0.6-0.99, 0.6-0.95, 0.6-0.9, 0.6-0.8, 0.6-0.7, 0.7-1.0, 0.7-0.99, 0.7-0.98, 0.7-0.97, 0.7-0.96, 0.7-0.95, 0.7-0.9, 0.7-0.8, 0. The Fl scores of 8-1.0, 0.8-0.99, 0.8-0.98, 0.8-0.97, 0.8-0.96, 0.8-0.95, 0.8-0.9, 0.9-1.0, 0.9-0.99, 0.9-0.98, 0.9-0.97, 0.9-0.96, or 0.9-0.95 were used to train a trained machine learning model to predict tumor purity.

[0073] III. Estimated Tumor Purity

[0074] Figure 6 An example of schematic diagram 600 for estimating tumor purity using a trained machine learning model, according to some implementation schemes, is shown. Tumor purity from sequencing of unmatched biological samples can be estimated using trained machine learning models (e.g., regression models). In some cases, deep neural networks are used to predict tumor purity based on the allele fraction distribution (e.g., B allele frequency distribution) of nucleotide sequence variants in unmatched biological samples. Regression models can be used to predict tumor purity from B allele frequencies (BAF) in whole-exome sequencing data.

[0075] In box 605, mismatched tumor samples (i.e., no matching normal samples) can be obtained from cancer patients. In box 610, DNA is extracted from the tumor samples, processed, and subjected to whole-exome sequencing. In some cases, the sequencing reads undergo quality control processing (e.g., via FastQC) to provide a FASTQ file. The FASTQ file is aligned with a reference genome to generate a BAM file.

[0076] In box 615, GATK HaplotypeCaller can be used to call variants from BAM files and generate VCF files containing variant information. Heterozygous sites in the sample exome can be identified, and the VCF file can be filtered to obtain reference and substitution read depths for all heterozygous sites. This information is used to calculate B allele frequencies (BAF). Normalized BAF distributions are calculated for heterozygous sites across the exome.

[0077] In box 620, a trained machine learning model can be used to estimate tumor purity from a standardized BAF distribution. Despite the lack of matching normal samples, the trained machine learning model can thus generate an accurate estimate of tumor purity.

[0078] A. Objects and Samples

[0079] To estimate tumor purity, nucleic acid sequence data representing multiple nucleic acid molecules can be obtained from the subject's biological sample. The subject can be a human. The subject can be male or female. Subjects can be fetuses, infants, children, adolescents, teenagers, or adults. The subject can be a patient of any age. For example, a subject can be a patient under approximately 10 years of age. For example, a subject can be a patient at least approximately 0, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, or 100 years of age. The subject can be a patient or other individual receiving or evaluating a treatment regimen (e.g., cancer treatment). However, in some cases, the subject is not receiving a treatment regimen.

[0080] In some cases, the subject can be either a mammal or a non-mammal. In others, the subject is a mammal, such as a human, a non-human primate (e.g., ape, monkey, chimpanzee), a cat, a dog, a rabbit, a goat, a horse, a cow, a pig, a rodent, a mouse, a SCID mouse, a rat, a guinea pig, or a sheep. In some implementations, species variants or homologs of these genes are used in non-human animal models. Species variants can be genes from different species that have the greatest sequence identity and similarity to each other in terms of functional properties. Many such species variant human genes may be listed in the Swiss-Prot database.

[0081] Some implementations may include obtaining samples from a subject such as a human subject. In some cases, clinical samples may be obtained from a patient. For example, blood may be drawn from a patient. Some implementations may include specifically detecting, analyzing, or quantifying molecules (e.g., nucleic acids, DNA, RNA, etc.) within a biological sample.

[0082] The sample can be a tissue sample or a body fluid. In some cases, the sample is a tissue or organ sample, such as a biopsy. In some cases, the sample includes cancer cells. In some cases, the sample includes both cancer cells and normal cells. In some cases, the sample is a tumor biopsy. Body fluids can be sweat, saliva, tears, urine, blood, menstrual blood, semen, and / or cerebrospinal fluid. In some cases, the sample is a blood sample. The sample may include one or more peripheral blood lymphocytes. The sample can be a whole blood sample. The blood sample can be a peripheral blood sample. In some cases, the sample includes peripheral blood mononuclear cells (PBMCs); in some cases, the sample includes peripheral blood lymphocytes (PBLs). The sample can be a serum sample.

[0083] Samples can be obtained using any method that provides samples suitable for the analytical methods described herein. Samples can be obtained by non-invasive methods such as throat swabs, oral swabs, bronchoalveolar lavage, urine collection, skin or cervical scraping, cheek swabs, saliva collection, stool collection, menstrual blood collection, or semen collection. Samples can be obtained by minimally invasive methods such as blood draw. Samples can be obtained by venipuncture. In other cases, samples are obtained by invasive procedures, including but not limited to: biopsy, alveolar or lung lavage, or needle aspiration. Biopsy methods can include surgical biopsy, incisional biopsy, excisional biopsy, perforated biopsy, shaving biopsy, or skin biopsy. Samples can be formalin-fixed sections. Needle aspiration methods can also include fine-needle aspiration, core needle biopsy, vacuum-assisted biopsy, or large-core biopsy. In some cases, multiple samples can be obtained using the methods described herein to ensure an adequate amount of biological material. In some cases, samples are not obtained by biopsy. In some cases, samples are not kidney biopsies.

[0084] The methods disclosed herein can be used to estimate the tumor purity in a sample, including at least about 1%, 5%, 10%, 15%, 20%, 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or more tumor cells. In some embodiments, the methods disclosed herein can be used to estimate the tumor purity in a sample, including up to about 10%, 15%, 20%, 25%, 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, or more tumor cells.

[0085] B. Generate nucleic acid sequencing data

[0086] In some implementations, samples are processed to obtain nucleic acid sequence data. "Nucleic acid" or "nucleic acid molecule" can correspond to a polymer of nucleotides of any length, whether ribonucleotides, deoxyribonucleotides, or peptide nucleic acids (PNAs), which include purine and pyrimidine bases, or other natural, chemically or biochemically modified non-natural or derived nucleotide bases. The backbone of a polynucleotide may include sugar and phosphate groups, as commonly found in RNA or DNA, or modified or substituted sugar or phosphate groups. Polynucleotides may include modified nucleotides, such as methylated nucleotides and nucleotide analogs. Nucleotide sequences can be broken down by non-nucleotide components. Therefore, the terms nucleoside, nucleotide, deoxynucleoside, and deoxynucleotide generally include analogs, as described herein. These analogs are those molecules that share some common structural features with naturally occurring nucleosides or nucleotides, such that when incorporated into nucleic acid or oligonucleotide sequences, they allow hybridization with naturally occurring nucleic acid sequences in solution. Typically, these analogs are derived from naturally occurring nucleosides and nucleotides by substituting and / or modifying bases, ribose, or phosphodiester moieties. It can be customized to make the hybridization stable or unstable or to enhance the specificity of hybridization with complementary nucleic acid sequences as needed. The nucleic acid molecule can be a DNA molecule. The nucleic acid molecule can also be an RNA molecule.

[0087] Sample processing includes nucleic acid sample processing and subsequent nucleic acid sample sequencing. Some or all biological samples can be sequenced to provide nucleic acid sequence data, which can be stored or otherwise maintained in electronic, magnetic, or optical storage locations. Sequence information can be analyzed with the aid of a computer processor, and the analyzed sequence information can be stored in an electronic storage location. The electronic storage location can include a library or collection of sequence information generated from the nucleic acid samples and the analyzed sequence information. In some embodiments, the biological samples are taken from subjects who have or are suspected of having cancer.

[0088] In some implementations, nucleic acid sequencing data are generated from pure tumor and pure normal samples. Paired cell lines can be obtained from another source (e.g., the U.S. Type Culture Collection). Each pair may include a tumor cell line and a normal cell line from the same subject. Cell lines can be cultured and expanded in vitro to obtain an appropriate number of cells for DNA extraction. DNA is extracted, processed, and sequenced using whole-exome or whole-genome sequencing. Sequence reads can undergo quality control processing (e.g., via FastQC) to provide FASTQ files.

[0089] In some cases, nucleic acid sequence data is generated using whole-genome sequencing. In some cases, whole-genome sequencing is used to identify variations in a person. In some cases, sequencing may include deep sequencing of a small portion of the genome. For example, the fraction of the genome may be at least approximately 50; 75; 100; 125; 150; 175; 200; 225; 250; 275; 300; 350; 400; 450; 500; 550; 600; 650; 700; 750; 800; 850; 900; 950; 1,000; 1100; 1200; 1300; 1400; 1500; 1600; 170 0; 1800; 1900; 2,000; 3,000; 4,000; 5,000; 6,000; 7,000; 8,000; 9,000; 10,000; 15,000; 20,000; 30,000; 40,000; 50,000; 60,000; 70,000; 80,000; 90,000; 100,000 or more bases or base pairs. In some cases, the genome can be sequenced to more than 1 million, 2 million, 3 million, 4 million, 5 million, 6 million, 7 million, 8 million, 9 million, 10 million or more bases or base pairs. In some cases, the genome can be sequenced along the entire exome (e.g., whole exome sequencing). In some cases, deep sequencing may involve obtaining multiple reads from a portion of the genome. For example, obtaining multiple reads may include at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 10,000, or more than 10,000 reads from a portion of the genome.

[0090] In some cases, generating nucleic acid sequence data involves detecting low-allelic portions via deep sequencing. In others, deep sequencing is performed using next-generation sequencing. In still others, deep sequencing is performed by avoiding error-prone regions. Error-prone regions may include those near sequence repeats, regions with abnormally high or low GC percentages, regions near homopolymers, dinucleotides, and trinucleotides, and regions near other short repeats. In some cases, error-prone regions may include areas that lead to DNA sequencing errors (e.g., polymerase slip in homopolymer sequences).

[0091] In some cases, generating nucleic acid sequence data involves performing one or more sequencing reactions on one or more nucleic acid molecules in a sample. Certain embodiments may include performing one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, fifteen or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, one hundred or more, two hundred or more, three hundred or more, four hundred or more, five hundred or more, six hundred or more, seven hundred or more, eight hundred or more, nine hundred or more, or one thousand or more. Sequencing reactions may be performed simultaneously, sequentially, or in combination. Sequencing reactions may include whole-genome sequencing or exome sequencing. Sequencing reactions may include Maxim-Gilbert, chain termination, or high-throughput systems. Optionally or additionally, sequencing reactions may include Helioscope™ single-molecule sequencing, nanopore DNA sequencing, Lynx Therapeutics' massively parallel characterization (MPSS), 454 pyrosequencing, single-molecule real-time (RNAP) sequencing, Illumina (Solexa) sequencing, SOLiD sequencing, IonTorrent™, ion semiconductor sequencing, single-molecule SMRT™ sequencing, Polony sequencing, DNA nanosphere sequencing, VisiGenBiotechnologies methods, or combinations thereof. Optionally or additionally, sequencing reactions may include one or more sequencing platforms, including but not limited to Illumina's Genome Analyzer IIx, HiSeq, and MiSeq; single-molecule real-time (SMRT™) technologies, such as the PacBio RS system provided by Pacific Biosciences (California) and Solexa Sequencer; and true single-molecule sequencing (tSMS™) technologies, such as the HeliScope™ Sequencer provided by Helicos Inc. (Cambridge, MA). Sequencing reactions may also include electron microscopy or chemically sensitive field-effect transistor (chemFET) arrays. In some aspects of this disclosure, sequencing reactions include capillary sequencing, next-generation sequencing, Sanger sequencing, sequencing synthesis, ligation sequencing, hybridization sequencing, single-molecule sequencing, or combinations thereof. Sequencing synthesis may include reversible terminator sequencing, progressive single-molecule sequencing, sequential flow sequencing, or combinations thereof. Sequential flow sequencing may include pyrosequencing, pH-mediated sequencing, semiconductor sequencing, or combinations thereof.

[0092] In some cases, generating nucleic acid sequence data involves performing at least one long-read sequencing reaction and at least one short-read sequencing reaction. Long-read and / or short-read sequencing reactions can be performed on at least a subset of nucleic acid molecules. Long-read and / or short-read sequencing reactions can be performed on at least a subset of two or more nucleic acid molecule subgroups. Long-read and short-read sequencing reactions can be performed on at least a subset of one or more nucleic acid molecule subgroups.

[0093] Sequencing of one or more nucleic acid molecules or their subgroups may include at least about 5; 10; 15; 20; 25; 30; 35; 40; 45; 50; 60; 70; 80; 90; 100; 200; 300; 400; 500; 600; 700; 800; 900; 1,000; 1500; 2,000; 2500; 3,000; 3500; 4000; 4500; 5,000; 5500; 6,000; 6500; 7,000; 7500; 8, 000; 8500; 9,000; 10,000; 25,000; 50,000; 75,000; 100,000; 250,000; 500,000; 750,000; 10,000,000; 25,000,000; 50,000,000; 100,000,000; 250,000,000; 500,000,000; 750,000,000; 1,000,000,000 or more sequencing reads.

[0094] Sequencing reactions may include sequencing one or more nucleic acid molecules at a rate of at least approximately 50; 60; 70; 80; 90; 100; 110; 120; 130; 140; 150; 160; 170; 180; 190; 200; 210; 220; 230; 240; 250; 260; 270; 280; 290; 300; 325; 350; 375; 400; 425; 450; 475; 500; 600; 700; 800; 900; 1,0 00; 1500; 2,000; 2500; 3,000; 3500; 4,000; 4500; 5,000; 5500; 6,000; 6500; 7,000; 7500; 8,000; 8500; 9,000; 10,000; 20,000; 30,000; 40,000; 50,000; 60,000; 70,000; 80,000; 90,000; 100,000 or more bases or base pairs. Sequencing reactions may include sequencing one or more nucleic acid molecules at a rate of at least approximately 50; 60; 70; 80; 90; 100; 110; 120; 130; 140; 150; 160; 170; 180; 190; 200; 210; 220; 230; 240; 250; 260; 270; 280; 290; 300; 325; 350; 375; 400; 425; 450; 475; 500; 600; 700; 800; 900; 1,0 00; 1500; 2,000; 2500; 3,000; 3500; 4,000; 4500; 5,000; 5500; 6,000; 6500; 7,000; 7500; 8,000; 8500; 9,000; 10,000; 20,000; 30,000; 40,000; 50,000; 60,000; 70,000; 80,000; 90,000; 100,000 or more consecutive bases or base pairs.

[0095] In some cases, sequencing technology produces at least 100 reads per run, at least 200 reads per run, at least 300 reads per run, at least 400 reads per run, at least 500 reads per run, at least 600 reads per run, at least 700 reads per run, at least 800 reads per run, at least 900 reads per run, at least 1000 reads per run, at least 5000 reads per run, at least 10000 reads per run, at least 500000 reads per run, or at least 1,000,000 reads per run. Optionally, the sequencing technology generates at least 1,500,000 reads per run, at least 2,000,000 reads per run, at least 2,500,000 reads per run, at least 3,000,000 reads per run, at least 3,500,000 reads per run, at least 4,000,000 reads per run, at least 4,500,000 reads per run, or at least 5,000,000 reads per run.

[0096] In some cases, sequencing technologies produce at least approximately 30 base pairs, at least approximately 40 base pairs, at least approximately 50 base pairs, at least approximately 60 base pairs, at least approximately 70 base pairs, at least approximately 80 base pairs, at least approximately 90 base pairs, at least approximately 100 base pairs, at least approximately 110, at least approximately 120 base pairs / read, at least approximately 150 base pairs, at least approximately 200 base pairs, at least approximately 250 base pairs, at least approximately 300 base pairs, at least approximately 350 base pairs, at least approximately 400 base pairs, at least approximately 450 base pairs, at least approximately 500 base pairs, at least approximately 550 base pairs, at least approximately 600 base pairs, at least approximately 700 base pairs, at least approximately 800 base pairs, at least approximately 900 base pairs, or at least approximately 1,000 base pairs / read. Alternatively or concurrently, sequencing technologies can produce long sequencing reads. In some cases, sequencing technologies can produce at least approximately 1,200 base pairs / read, at least approximately 1,500 base pairs / read, at least approximately 1,800 base pairs / read, at least approximately 2,000 base pairs / read, at least approximately 2,500 base pairs / read, at least approximately 3,000 base pairs / read, at least approximately 3,500 base pairs / read, at least approximately 4,000 base pairs / read, at least approximately 4,500 base pairs / read, at least approximately 5,000 base pairs / read, at least approximately 6,000 base pairs / read, and so on. At least 7,000 base pairs / read, at least 8,000 base pairs / read, at least 9,000 base pairs / read, at least 10,000 base pairs / read, 20,000 base pairs / read, 30,000 base pairs / read, 40,000 base pairs / read, 50,000 base pairs / read, 60,000 base pairs / read, 70,000 base pairs / read, 80,000 base pairs / read, 90,000 base pairs / read, or 100,000 base pairs / read.

[0097] High-throughput sequencing systems allow for the detection of sequenced nucleotides immediately after or during incorporation into the growth chain, i.e., real-time or near-real-time sequence detection. In some cases, high-throughput sequencing generates at least 1,000, 5,000, 10,000, 20,000, 30,000, 40,000, 50,000, 100,000, or 500,000 sequence reads per hour; each read contains at least 50, 60, 70, 80, 90, 100, 120, 150, 200, 250, 300, 350, 400, 450, or 500 bases. Nucleic acids described herein, such as genomic DNA, cDNA derived from RNA transcripts, or RNA, can be used as templates for sequencing.

[0098] C. Identification of nucleotide sequence variants

[0099] To predict the tumor purity of a biological sample, the B allele frequency (BAF) of the biological sample can be determined. To determine the BAF, a set of genomic regions can be identified, where each set of genomic regions may include nucleotide sequence variants of a corresponding genomic region relative to a reference genome. To identify a set of genomic regions, nucleic acid sequence data can be aligned to a reference genome. For example, a FASTQ file corresponding to nucleic acid sequence data can be aligned to a reference genome to generate one or more BAM files. One or more BAM files can be processed by another module (e.g., GATK HaplotypeCaller) to identify a set of genomic regions. In some cases, a VCF file containing a set of genomic regions is generated. Alternatively or optionally, a VCF file can be obtained for a matching cell line. The VCF file can identify a set of genomic regions with nucleotide sequence variations based on the comparison of nucleic acid sequence data and a reference genome.

[0100] Some implementations may include nucleic acid molecules containing one or more genomic regions. Some implementations may include nucleic acid molecules containing one or more sets of genomic regions. One or more genomic regions may include one or more genomic region features. Genomic region features may include the entire genome or a portion thereof. Genomic region features may include the entire exome or a portion thereof. Genomic region features may include one or more sets of genes. Genomic region features may include one or more genes. Genomic region features may include one or more sets of regulatory elements. Genomic region features may include one or more regulatory elements.

[0101] Genome region features may include a set of polymorphisms. Genome region features may include one or more polymorphisms. Genome region features may be related to the GC content, complexity, and / or mappability of one or more nucleic acid molecules. Genome region features may include one or more simple tandem repeats (STRs), unstable extended repeats, segmental repeats, single and paired read degeneracy mapping fractions, GRCh37 patches, or combinations thereof. Genome region features may include one or more low-average coverage regions from whole-genome sequencing (WGS), zero-average coverage regions from WGS, validated compression, or combinations thereof. Genome region features may include one or more substitution or non-reference sequences. Genome region features may include one or more gene phasing and recombinant genes. In some aspects of this disclosure, one or more genome region features are not mutually exclusive. For example, a genome region feature comprising the entire genome or a portion thereof may overlap with other genome region features, such as the entire exome or a portion thereof, one or more genes, one or more regulatory elements, etc. Optionally, one or more genome region features are mutually exclusive. For example, a genome region comprising the non-coding portion of the entire genome will not overlap with genome region features, such as the exome or a portion thereof, or the coding portion of a gene. Optionally or additionally, one or more genomic region features may be partially exclusive or partially inclusive. For example, a genomic region comprising an entire exon or a portion thereof may partially overlap with a genomic region comprising the exon portion of a gene. However, a genomic region comprising an entire exon or a portion thereof will not overlap with a genomic region comprising the intron portion of a gene. Therefore, a genomic region feature comprising a gene or a portion thereof may partially exclude and / or partially include a genomic region feature comprising an entire exon or a portion thereof.

[0102] Some embodiments may include a nucleic acid sample or molecule comprising one or more genomic regions, wherein at least one of the one or more genomic regions includes a genomic region feature that includes the entire genome or a portion thereof. The entire genome or a portion thereof may include one or more coding portions of the genome, one or more non-coding portions of the genome, or a combination thereof. The coding portions of the genome may include one or more coding portions of genes encoding one or more proteins. The one or more coding portions of the genome may include the entire exome or a portion thereof. Optionally or additionally, the one or more coding portions of the genome may include one or more exons.

[0103] One or more non-coding regions of the genome may include one or more non-coding molecules or a portion thereof. Non-coding molecules may include one or more non-coding RNAs, one or more regulatory elements, one or more introns, one or more pseudogenes, one or more repetitive sequences, one or more transposons, one or more viral elements, one or more telomeres, a portion thereof, or a combination thereof. Non-coding RNAs may be functional RNA molecules that are not translated into proteins. Examples of non-coding RNAs include, but are not limited to, ribosomal RNA, transfer RNA, piwi-interacting RNA, microRNA, siRNA, shRNA, snoRNA, sncRNA, and lncRNA. Pseudogenes may be associated with known genes and are generally no longer expressed. Repetitive sequences may include one or more tandem repeats, one or more scattered repeats, or a combination thereof. Tandem repeats may include one or more satellite DNAs, one or more microsatellites, one or more microsatellites, or a combination thereof.

[0104] Scattered repeats can include one or more transposons. Transposons can be mobile genetic elements. Mobile genetic elements are generally able to change their location in the genome. Transposons can be classified into class I transposons (class I TEs) or class II transposons (class II TEs). Class I TEs (e.g., retrotransposons) typically undergo self-replication in two stages, first by transcription from DNA to RNA, and then by reverse transcription from RNA back to DNA. The DNA copy can then be inserted into a new location in the genome. Class I TEs can include one or more long terminal repeats (LTRs), one or more long scattered nuclear elements (LINEs), one or more short scattered nuclear elements (SINEs), or combinations thereof. Examples of LTRs include, but are not limited to, human endogenous retroviruses (HERVs), medium reiterated repeats 4 (MER4), and retrotransposons. Examples of LINEs include, but are not limited to, LINE1 and LINE2. SINEs can include one or more Alu sequences, one or more mammalian-wide scattered repeats (MIRs), or combinations thereof. Class II TEs (e.g., DNA transposons) typically do not involve RNA intermediates. DNA transposons typically cleave from one site and insert into another site in the genome. Optionally, DNA transposons are replicated and inserted into a new location in the genome. Examples of DNA transposons include, but are not limited to, MER1, MER2, and mariners. Viral elements may include one or more endogenous retroviral sequences. Telomeres are typically repetitive DNA regions located at the ends of chromosomes.

[0105] Some implementations may include a subset of nucleic acid samples or nucleic acid molecules comprising one or more genomic regions, wherein at least one of the one or more genomic regions includes a genomic region feature comprising the entire exome or a portion thereof. The exome is typically a portion of the genome formed by exons. The exome may be formed by untranslated regions (UTRs), splice sites, and / or intron regions. The entire exome or a portion thereof may include one or more exons of a protein-coding gene. The entire exome or a portion thereof may include one or more untranslated regions (UTRs), splice sites, and introns.

[0106] Some implementations may include a nucleic acid sample or molecule comprising one or more genomic regions, wherein at least one of the genomic regions includes a genomic region feature comprising a gene or a portion thereof. Typically, a gene comprises a nucleic acid fragment encoding a polypeptide or functional RNA. A gene may include one or more exons, one or more introns, one or more untranslated regions (UTRs), or a combination thereof. Exons are typically the coding portion of a gene, transcribed into a precursor mRNA sequence, and are present in the gene's final mature RNA product. Introns are typically the non-coding portion of a gene, transcribed into a precursor mRNA sequence, and removed by RNA splicing. A UTR may refer to a portion on each side of the coding sequence on the mRNA strand. A UTR located on the 5' side of the coding sequence may be called a 5' UTR (or leader sequence). A UTR located on the 3' side of the coding sequence may be called a 3' UTR (or tail sequence). A UTR may contain one or more elements for controlling gene expression. Elements, such as regulatory elements, may be located in the 5' UTR. Regulatory sequences, such as polyadenylation signals, protein binding sites, and miRNA binding sites, may be located in the 3' UTR. Protein binding sites located in the 3'UTR may include, but are not limited to, selenocysteine ​​insertion sequence (SECIS) elements and AU-rich elements (AREs). SECIS elements can instruct ribosomes to translate the codon UGA to selenocysteine ​​instead of a stop codon. AREs are typically fragments composed primarily of adenine and uracil nucleotides, which may affect mRNA stability.

[0107] Some implementations may include nucleic acid samples or subsets of nucleic acid molecules comprising one or more genomic regions, wherein at least one of the one or more genomic regions includes a genomic region feature containing a set of genes. This set of genes may include, but is not limited to, Mendelian DB genes, Human Genetic Mutation Database (HGMD) genes, Cancer Gene Screening genes, Online Human Mendelian Inheritance (OMIM) Mendelian genes, HGMD Mendelian genes, and Human Leukocyte Antigen (HLA) genes. This set of genes may possess one or more known Mendelian traits, one or more known disease traits, one or more known drug traits, one or more known biomedically interpretable variants, or a combination thereof. Mendelian traits may be controlled by a single locus and may exhibit a Mendelian inheritance pattern. A set of genes with known Mendelian traits may include one or more genes encoding Mendelian traits, including but not limited to the ability to taste phenylthiourea (dominant), the ability to smell (almond-like) hydrogen cyanide (recessive), albinism (recessive), short fingers (short fingers and toes), and wet (dominant) or dry (recessive) earwax. Disease traits result in or increase disease risk and may be inherited in a Mendelian or complex pattern. A group of genes with known disease characteristics may include one or more genes encoding disease characteristics, including but not limited to cystic fibrosis, hemophilia, and Lynch syndrome.

[0108] Drug properties may alter the metabolism, optimal dosage, adverse reactions, and side effects of one or more drugs or drug families. A group of genes with known drug properties may include one or more genes encoding drug properties, including but not limited to CYP2D6, UGT1A1, and ADRB1. Biomedically explainable variants may be polymorphisms in genes associated with a disease or indication. A group of genes with known biomedically explainable variants may include one or more genes encoding biomedically explainable variants, including but not limited to cystic fibrosis (CF) mutations, muscular dystrophy mutations, p53 mutations, Rb mutations, cell cycle regulators, receptors, and kinases. Optionally or additionally, a group of genes with known biomedically explainable variants may include one or more genes associated with Huntington's disease, cancer, cystic fibrosis, or muscular dystrophy (e.g., Duchenne muscular dystrophy).

[0109] Some implementations may include a nucleic acid sample or molecule comprising one or more genomic regions, wherein at least one of the genomic regions includes a genomic region feature that includes a regulatory element or a portion thereof. The regulatory element may be a cis-regulatory element or a trans-regulatory element. A cis-regulatory element may be a sequence that controls the transcription of a nearby gene. A cis-regulatory element may be located in a 5' or 3' untranslated region (UTR) or within an intron. A trans-regulatory element may control the transcription of a distant gene. The regulatory element may include one or more promoters, one or more enhancers, or a combination thereof. Promoters can promote the transcription of a specific gene and may be found upstream of the coding region. Enhancers may have a distant effect on the transcriptional level of a gene.

[0110] Some implementations may include nucleic acid samples or subsets of nucleic acid molecules comprising one or more genomic regions, wherein at least one of the one or more genomic regions includes a genomic region feature that includes polymorphism or a portion thereof. Generally, polymorphism refers to a mutation in the genotype. Polymorphism can be a germline variant or a somatic variant. Polymorphism can include one or more base changes, insertions, duplications, or deletions of one or more bases. Copy number variations (CNVs), transversions, and other rearrangements are also forms of genetic variation. Polymorphic markers include restriction fragment length polymorphism, variable number tandem repeats (VNTRs), hypervariable regions, microsatellites, dinucleotide repeats, trinucleotide repeats, tetranucleotide repeats, simple sequence repeats, and insertion elements such as Alu. The allelic form most frequently occurring in a selected population is sometimes referred to as the wild-type form. Diploid organisms can be homozygous or heterozygous for allelic forms. There are two forms of diallelic polymorphism. There are three forms of triallelic polymorphism. Single nucleotide polymorphism (SNP) is a form of polymorphism. In some aspects of this disclosure, one or more polymorphisms include one or more single nucleotide variants, insertions / deletions (inDel), small insertions, small deletions, structural variant connections, variable-length tandem repeats, flanking sequences, or combinations thereof. One or more polymorphisms may be located within coding and / or non-coding regions. One or more polymorphisms may be located within, around, or near genes, exons, introns, splice sites, untranslated regions, or combinations thereof. One or more polymorphisms may span at least a portion of genes, exons, introns, or untranslated regions.

[0111] Some embodiments may include a nucleic acid sample or molecule comprising one or more genomic regions, wherein at least one of the one or more genomic regions includes a genomic region feature, the genomic region feature including one or more simple tandem repeats (STRs), unstable extended repeats, segmental repeats, single and paired read degeneracy mapping fractions, GRCh37 patches, or a combination thereof. One or more STRs may include one or more homopolymers, one or more dinucleotide repeats, one or more trinucleotide repeats, or a combination thereof. One or more homopolymers may be about 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, or more bases or base pairs. Dinucleotide repeats and / or trinucleotide repeats may be about 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 35, 40, 45, 50, or 50 bases or base pairs. The degeneracy mapping scores for single and paired reads can be based on or derived from the 100-mer alignment of GEMs from ENCODE / CRG (Guigo), the 75-mer alignment of GEMs from ENCODE / CRG (Guigo), the mean of 100-base-pair box car for signal mapping capability, the maximum value of loci and possible pairs of paired read scores, or a combination of the above.

[0112] Genome region features may include one or more low-mean-coverage regions from whole-genome sequencing (WGS), zero-mean-coverage regions from WGS, validated compression, or a combination thereof. Low-mean-coverage regions from WGS may include regions chemically generated from Illumina v3, regions below the first percentile of a Poisson distribution based on mean coverage, or a combination thereof. Zero-mean-coverage regions from WGS may include regions chemically generated from Illumina v3. Validated compression may include regions with high mapping depth, regions with two or more observed haplotypes, regions expected to have deleted repeats in the reference, or a combination thereof. Genome region features may include one or more substitution or non-reference sequences. One or more substitution or non-reference sequences may include known structural variant connections, known insertions, known deletions, substitutional haplotypes, or a combination thereof. Genome region features may include one or more phasing and recombinant genes. Examples of phasing and recombinant genes include, but are not limited to, one or more major histocompatibility complexes, blood group and amylase gene families. One or more major histocompatibility complexes may include one or more HLA class I, HLA class II, or a combination thereof. One or more HLA class I blood types may include HLA-A, HLA-B, HLA-C, or combinations thereof. One or more HLA class II blood types may include HLA-DP, HLA-DM, HLA-DOA, HLA-DOB, HLA-DQ, HLA-DR, or combinations thereof. Blood group genes may include ABO, RHD, RHCE, or combinations thereof.

[0113] Some embodiments may include a nucleic acid sample or molecule comprising one or more genomic regions, wherein at least one of the one or more genomic regions includes genomic region features associated with the GC content of the one or more nucleic acid molecules. GC content may refer to the GC content of the nucleic acid molecule. Optionally, GC content may refer to the GC content of one or more nucleic acid molecules and may be referred to as average GC content. As used herein, the terms “GC content” and “average GC content” are used interchangeably. The GC content of a genomic region may be a high GC content. Typically, a high GC content means a GC content greater than or equal to about 65%, 70%, 75%, 80%, 85%, 90%, 95%, 97% or more. In some aspects of this disclosure, a high GC content may refer to a GC content greater than or equal to about 70%. The GC content of a genomic region may be a low GC content. Typically, a low GC content means a GC content less than or equal to about 65%, 60%, 55%, 50%, 45%, 40%, 35%, 30%, 25%, 20%, 15%, 10%, 5%, 2% or less.

[0114] Some implementations may include a nucleic acid sample or molecule containing one or more genomic regions, wherein at least one of the genomic regions includes genomic region features associated with the complexity of the one or more nucleic acid molecules. The complexity of a nucleic acid molecule may refer to the randomness of its nucleotide sequence. Low complexity may refer to patterns, repetitions, and / or depletion of one or more types of nucleotides in the sequence.

[0115] Some implementations may include a nucleic acid sample or molecule comprising one or more genomic regions, wherein at least one of the genomic regions includes genomic region features associated with the mappability of the one or more nucleic acid molecules. The mappability of a nucleic acid molecule may refer to its uniqueness in alignment with a reference sequence. Nucleic acid molecules with low mappability may have poor alignment with a reference sequence.

[0116] Calculation of D.BAF distribution

[0117] For each genomic region in a set of genomic regions, the B allele frequency (BAF) can be determined. BAF is a normalized measure of the ratio of allele strengths of the two alleles (A and B). A BAF of 100 or 0 indicates the complete absence of one of the two alleles (e.g., AA or BB). For example, when both chromosomes of a sample contain nucleotide sequence variants relative to a reference genome, the BAF for a given genomic region could be 100. Therefore, variant alleles should be present in approximately 100% of the reads in that sample. In some cases, the BAF distribution is normalized to produce a density map such that the sum of the areas under the curves is 1. Specifically, a BAF histogram for all variants is generated, the area of ​​that histogram is calculated, and each bin of the histogram is divided by the calculated area.

[0118] For heterozygous alleles in a sample, variants will appear in a subset of sequence reads. Therefore, a BAF of 50 indicates the presence of two identical alleles (e.g., AB). In the absence of copy number variations, the BAF of a normal sample might be 0, 50, or 100. However, in tumor samples, mutations can alter the BAF value. For example, a copy number event involving the replication of the B allele can result in a B allele frequency of -67%. In another instance, when only one chromosome contains a variant, the variant allele should be present in approximately 50% of the reads in that sample. The inherent randomness in the data generation process introduces noise, causing observed BAF measurements to deviate from the ideal value for a given copy number in both tumor and normal samples.

[0119] In some cases, subsets of genomic regions corresponding to heterozygous sites in a biological sample (e.g., an exome sample) are identified, and the allele frequency of each subset of genomic regions is quantified. Therefore, VCF files can be filtered to obtain reference and substitution read depths for all heterozygous sites with nucleotide sequence variations. This information can be used to calculate the BAF (Body Accumulation Factor) for each genomic region within the subset of heterozygous genomic regions.

[0120] Based on the BAF determined for each group of genomic regions, a normalized BAF distribution can be calculated. The BAF distribution can indicate the absolute amount, percentage, and / or normalized amount of the BAF corresponding to each genomic region in that group of genomic regions. In some cases, the normalized BAF distribution is determined from the BAF corresponding to a subset of genomic regions (e.g., heterozygous sites between exons). Because BAF values ​​are typically between 0 and 100, a normalized BAF distribution can include up to 101 BAF values, each indicating a corresponding normalized frequency. Alternatively or additionally, the normalized BAF distribution can be modified such that the corresponding BAF values ​​can be filled into a predetermined number of bins. For example, each bin can correspond to a non-overlapping range of BAF values ​​(e.g., 1–9, 10–19), and BAF values ​​can be assigned to the corresponding bins. The BAF distribution can then be determined based on the values ​​corresponding to each predetermined bin.

[0121] Figures 7A-7E Examples of BAF distributions plotted according to some implementation schemes are provided. For Figure 7A and 7B The x-axis represents the BAF value, ranging from 0% (homozygous A allele) to 100% (homozygous B allele). As explained above, a BAF of 50 indicates heterozygosity, corresponding to the equal presence of both alleles (e.g., AB). The expected BAF for a normal sample might be 0, 50, or 100. However, in tumor samples, mutations can alter the BAF value. For example, a copy number event of the B allele duplication can result in a B allele frequency of -67%. The y-axis represents the normalized frequency value corresponding to each BAF value on the x-axis. Normalized frequency values ​​can identify multiple genomic regions corresponding to a specific BAF value. For example, a normalized frequency of 50% BAF could be a value greater than 4.

[0122] Figure 7A This is a histogram of the frequency of the B allele at heterozygous loci in chromosomes from pure normal samples. The histogram exhibits a normal distribution, with most observed B allele frequencies decreasing to close to 50%. Figure 7B This is a histogram of the B allele frequencies at heterozygous sites in pure tumor samples. The histogram shows a changing distribution, with more observations of B allele frequencies decreasing further from 50%. Figure 7BAs shown, an increase in the number of somatic mutations may lead to a change in the BAF frequency from 50%.

[0123] Figure 7C This is a heatmap where each row represents the BAF distribution of chromosomes from a purely normal sample. The heatmap exhibits a normal distribution, with most observed B allele frequencies decreasing to close to 50%. Figure 7E This is a heatmap where each row represents the BAF distribution of chromosomes from a pure normal sample. The heatmap shows a modified distribution, with more observations of the B allele frequency decreasing further from 50%. Figure 7D This is a thermogram of a sample consisting of 50% normal and 50% tumor. The thermogram shows the amount of material falling into the sample. Figure 7C and Figure 7E The intermediate distributions between the distributions shown in the figure.

[0124] E. Using BAF to classify pure tumor and pure normal samples.

[0125] In some cases, the BAF (Biological Absorption Frame) features of biological samples are used to classify them as normal or tumor-bearing. For example, for multiple pure tumor and pure normal samples, a normalized BAF distribution can be computed at heterozygous sites across exomes. Logistic regression with whole-exome BAF features can be used to classify pure tumor and pure normal samples.

[0126] Figure 8 Figure 800 shows the first two principal components (PC1 and PC2) for classifying biological samples using B allele frequency characteristics according to some implementation schemes. Figure 8 In the results, the first two principal components show near-linear separability between tumors and normal samples, suggesting that BAF features may be suitable for estimating the tumor content of samples. Furthermore, it can be demonstrated that logistic regression with whole-exome BAF features can be used to classify pure tumor and pure normal samples.

[0127] F. Tumor purity prediction based on BAF distribution

[0128] A trained machine learning model can be used to process the BAF distribution of biological samples to predict indicators of tumor purity in the biological samples. In some embodiments, the trained machine learning model includes a fully connected neural network. The fully connected neural network may include fully connected layers with rectified linear unit (ReLU) activation functions. In some embodiments, the output activation function of the fully connected neural network is a sigmoid function. The loss function of the fully connected neural network can be configured to generate mean squared error (MSE). In some embodiments, the fully connected neural network is tuned via hyperparameter search using random sampling, wherein it involves linear search on layers, linear search on size, logarithmic search on learning rate, or a combination thereof.

[0129] In some cases, the trained machine learning model includes a one-dimensional convolutional neural network. The one-dimensional convolutional neural network can be configured such that the B allele frequency distribution can be used as input, and encoded as an input of size 25 in height, 1 in width, and 100 in depth. In some implementations, each layer of the one-dimensional convolutional neural network performs an lx1 convolution followed by a ReLU activation function.

[0130] In some implementations, the trained machine learning model comprises a two-dimensional convolutional neural network (2D convolutional neural network). In some implementations, the B allele frequency distribution of the 2D convolutional neural network is used as input and encoded into an input size defined by a height of 25, a width of 100, and a depth of 1. In some implementations, each convolutional layer of the 2D convolutional neural network is followed by a ReLU layer. The output of the 2D convolutional neural network can be a densely connected layer with a sigmoid activation function. Furthermore, the hyperparameters of the 2D convolutional neural network can be tuned by adjusting the number of layers, filter size, number of filters, or a combination thereof.

[0131] The system can output an estimated metric for identifying tumor purity. For example, it can output a report that includes the estimated metric. In some embodiments, the report includes information identifying the distribution of B allele frequencies. The report may also include information identifying at least one diagnostic biomarker and / or at least one prognostic biomarker. In some embodiments, the report includes information identifying predicted somatic variants. The report may also include treatment recommendations. In some embodiments, treatment recommendations include recommendations to administer treatment to human subjects. Treatment recommendations may include recommendations not to administer treatment to human subjects.

[0132] IV. An exemplary procedure for estimating tumor purity from a single sample

[0133] Figure 9 The included flowchart 900 illustrates an example of a method for predicting the tumor purity of a biological sample according to certain embodiments. The operations described in flowchart 900 can be performed by, for example, a computer system implementing a trained machine learning model, such as a trained one-dimensional or two-dimensional convolutional neural network. Although flowchart 900 may describe the operations as a sequential process, in various embodiments, many operations can be performed in parallel or simultaneously. Furthermore, the order of operations can be rearranged. Operations may have additional steps not shown in the figure. Moreover, embodiments of the method can be implemented using hardware, software, firmware, middleware, microcode, hardware description languages, or any combination thereof. When implemented in software, firmware, middleware, or microcode, the program code or code portions used to perform the relevant tasks can be stored in a computer-readable medium such as a storage medium.

[0134] At operation 910, the computer system acquires nucleic acid sequence data for multiple nucleic acid molecules representing a biological sample of the object. Nucleic acid sequence data can be generated by sequencing multiple nucleic acid molecules from a tumor sample. In some cases, multiple nucleic acid molecules are isolated prior to sequencing. The nucleic acid sequence data may correspond to whole exome sequence data. Optionally or additionally, the nucleic acid sequence data is whole genome sequencing data.

[0135] At step 920, the computer system aligns the nucleic acid sequence data with a reference genome. For example, a FASTQ file corresponding to the nucleic acid sequence data can be aligned with a reference genome to generate one or more BAM files.

[0136] At operation 930, the computer system identifies a set of genomic regions based on aligned nucleic acid sequence data. Each of these genomic regions may include one or more nucleotide sequence variants relative to a corresponding genomic region of a reference genome. In some cases, the computer system identifies one or more candidate nucleotide sequence variants in the nucleic acid sequence data and calculates the reference and substitution read depths for each of the one or more candidate nucleotide sequence variants.

[0137] At operation 940, the computer system determines the BAF (Body Affinity) for each genomic region within the set of genomic regions. BAF is a standardized measure of the ratio of allele strength between two alleles (A and B). In some cases, a BAF of 100 or 0 indicates the complete absence of one of the two alleles (e.g., AA or BB), and a BAF of 50 indicates the equal presence of both alleles (e.g., AB). The expected BAF for a normal sample is likely to be 0, 50, or 100. However, in tumor samples, mutations can alter the BAF value.

[0138] At operation 950, the computer system determines the BAF distribution of the biological sample based on the BAF of this set of genomic regions. In some cases, the B allele frequency is normalized.

[0139] At operation 960, the computer system uses a trained machine learning model to process the B allele frequency distribution to predict a metric for identifying the tumor purity of a biological sample. In some cases, the trained machine learning model is trained on a training dataset generated from nucleic acid sequence data derived from one or more tumor cells diluted to normal cells. The trained machine learning model may have a mean absolute error of less than approximately 0.2.

[0140] At operation 970, the computer system outputs a metric. In some cases, the computer system output includes a report that includes a predictive metric identifying tumor purity. In addition to the predicted metric, the report may also include the determined distribution of B allele frequencies. The report may also include other types of information, including but not limited to at least one biomarker, at least one prognostic marker, predicted somatic variants, and treatment recommendations (e.g., whether treatment should be given to the subject).

[0141] V. Other considerations

[0142] A. Detection technology

[0143] Some implementations may include one or more markers. One or more markers may be attached to one or more capture probes, nucleic acid molecules, beads, primers, or combinations thereof. Examples of markers include, but are not limited to, detectable markers such as radioisotopes, fluorophores, chemiluminescent compounds, chromophores, luminescent compounds, enzymes, colloidal particles and fluorescent microparticles, quantum dots, and antigens, antibodies, haptens, avidin / streptoavidin, biotin, haptens, enzyme cofactors / substrates, one or more members of quenching systems, chromogens, haptens, magnetic particles, materials exhibiting nonlinear optics, semiconductor nanocrystals, metal nanoparticles, enzymes, aptamers, and one or more members of binding pairs.

[0144] Some implementations may include one or more capture probes, multiple capture probes, or one or more sets of capture probes. Typically, a capture probe contains a nucleic acid binding site. A capture probe may also contain one or more adapters. A capture probe may also contain one or more tags. One or more adapters can attach one or more tags to the nucleic acid binding site.

[0145] Capture probes can hybridize with one or more nucleic acid molecules in a sample. Capture probes can hybridize with one or more genomic regions. Capture probes can hybridize within, around, near, or across one or more combinations of one or more genes, exons, introns, UTRs, or UTRs. Capture probes can hybridize across one or more combinations of one or more genes, exons, introns, UTRs, or UTRs. Capture probes can hybridize with one or more known inDel molecules. Capture probes can hybridize with one or more known structural variants.

[0146] Some implementations may include one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, one hundred or more, 125 or more, 150 or more, 175 or more, 200 or more, 250 or more, 300 or more, 350 or more, 400 or more, 500 or more, 600 or more, 700 or more, 800 or more, 900 or more, or one thousand or more. One or more capture probes or sets of capture probes may be different, similar, identical, or combinations thereof.

[0147] One or more capture probes may contain nucleic acid binding sites that hybridize to at least a portion of one or more nucleic acid molecules or their variants or derivatives in a subset of the sample or nucleic acid molecules. Capture probes may contain nucleic acid binding sites that hybridize to one or more genomic regions. Capture probes may hybridize to different, similar, and / or identical genomic regions. One or more capture probes may have complementarity with one or more nucleic acid molecules or their variants or derivatives of at least about 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, 97%, 99%, or more.

[0148] A capture probe may contain one or more nucleotides. A capture probe may contain one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, one hundred or more, one hundred and twenty-five or more, one hundred and twenty-five, one hundred and twenty-five, one hundred and twenty-five, one hundred and twenty-five, one hundred and twenty-five, one hundred and twenty-five, one hundred and twenty-five, one hundred and twenty-five, one hundred and twenty-five, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, one hundred and twenty-six, or one hundred and twenty-six. A capture probe may contain approximately 100 nucleotides. The capture probe may comprise about 10 to about 500 nucleotides, about 20 to about 450 nucleotides, about 30 to about 400 nucleotides, about 40 to about 350 nucleotides, about 50 to about 300 nucleotides, about 60 to about 250 nucleotides, about 70 to about 200 nucleotides, or about 80 to about 150 nucleotides. In some aspects of this disclosure, the capture probe comprises about 80 nucleotides to about 100 nucleotides.

[0149] Multiple capture probes or sets of capture probes may contain two or more capture probes having the same, similar, and / or different nucleic acid binding site sequences, adapters, and / or labels. For example, two or more capture probes contain the same nucleic acid binding site. In another example, two or more capture probes contain similar nucleic acid binding sites. In yet another example, two or more capture probes contain different nucleic acid binding sites. Two or more capture probes may further contain one or more adapters. Two or more capture probes may further contain different adapters. Two or more capture probes may further contain similar adapters. Two or more capture probes may further contain the same adapters. Two or more capture probes may further contain one or more labels. Two or more capture probes may further contain different labels. Two or more capture probes may further contain similar labels. Two or more capture probes may further contain the same label.

[0150] B. Assay and Amplification Techniques

[0151] Some implementations may include performing one or more assays on a sample containing one or more nucleic acid molecules. Generating two or more nucleic acid molecule subgroups may include performing one or more assays. Nucleic acid molecule subgroups from a sample may be assayed. One or more nucleic acid molecules from a sample may be assayed. At least a portion of the nucleic acid molecule subgroup may be assayed. Assays may include one or more techniques, reagents, capture probes, primers, labels, and / or components for detecting, quantifying, and / or analyzing one or more nucleic acid molecules.

[0152] The assay may include, but is not limited to, sequencing, amplification, hybridization, enrichment, separation, elution, fragmentation, detection, and quantification of one or more nucleic acid molecules. The assay may include methods for preparing one or more nucleic acid molecules.

[0153] Some implementations may include one or more amplification reactions on one or more nucleic acid molecules in a sample. The term “amplification” refers to any process that produces at least one copy of a nucleic acid molecule. The terms “amplifier” and “amplified nucleic acid molecule” refer to copies of nucleic acid molecules and are used interchangeably. Amplification reactions may include PCR-based methods, non-PCR-based methods, or combinations thereof. Examples of non-PCR-based methods include, but are not limited to, multiple displacement amplification (MDA), transcription-mediated amplification (TMA), nucleic acid sequence-based amplification (NASBA), strand displacement amplification (SDA), real-time SDA, rolling circle amplification, or cycle-to-cycle amplification. PCR-based methods may include, but are not limited to, PCR, HD-PCR, next-generation PCR, digital RTA, or any combination thereof. Other PCR methods include, but are not limited to, linear amplification, allele-specific PCR, Alu PCR, assembly PCR, asymmetric PCR, droplet PCR, emulsion PCR, helicase-dependent amplification (HDA), hot-start PCR, inverse PCR, linear post-exponential (LATE)-PCR, long PCR, multiplex PCR, nested PCR, semi-nested PCR, quantitative PCR, RT-PCR, real-time PCR, single-cell PCR, and falling PCR.

[0154] Some implementations may include one or more hybridization reactions on one or more nucleic acid molecules in a sample. Hybridization reactions may include hybridization of one or more capture probes with one or more nucleic acid molecules in a sample or a subset of nucleic acid molecules. Hybridization reactions may include hybridization of one or more sets of capture probes with one or more nucleic acid molecules in a sample or a subset of nucleic acid molecules. Hybridization reactions may include one or more hybridization arrays, multiplex hybridization reactions, hybridization chain reactions, isothermal hybridization reactions, nucleic acid hybridization reactions, or combinations thereof. One or more hybridization arrays may include hybridization array genotyping, hybridization array proportional sensing, DNA hybridization arrays, macroarrays, microarrays, high-density oligonucleotide arrays, genomic hybridization arrays, comparative hybridization arrays, or combinations thereof. Hybridization reactions may include one or more capture probes, one or more beads, one or more labels, one or more nucleic acid molecule subgroups, one or more nucleic acid samples, one or more reagents, one or more washing buffers, one or more elution buffers, one or more hybridization buffers, one or more hybridization chambers, one or more incubators, one or more separators, or combinations thereof.

[0155] Some implementations may include one or more enrichment reactions on one or more nucleic acid molecules in a sample. The enrichment reaction may include contacting the sample with one or more beads or groups of beads. The enrichment reaction may include differential amplification of two or more subgroups of nucleic acid molecules based on the characteristics of one or more genomic regions. For example, the enrichment reaction includes differential amplification of two or more subgroups of nucleic acid molecules based on GC content. Optionally or additionally, the enrichment reaction includes differential amplification of two or more subgroups of nucleic acid molecules based on methylation status. The enrichment reaction may include one or more hybridization reactions. The enrichment reaction may further include isolating and / or purifying one or more hybridized nucleic acid molecules, one or more bead-bound nucleic acid molecules, one or more free nucleic acid molecules (e.g., capture probe free nucleic acid molecules, bead free nucleic acid molecules), one or more labeled nucleic acid molecules, one or more unlabeled nucleic acid molecules, one or more amplicons, one or more unamplified nucleic acid molecules, or a combination of the above. Optionally or additionally, the enrichment reaction may include enriching one or more cell types in the sample. One or more cell types can be enriched by flow cytometry.

[0156] One or more enrichment reactions can produce one or more enriched nucleic acid molecules. Enriched nucleic acid molecules can include nucleic acid molecules or their variants or derivatives. For example, enriched nucleic acid molecules may include one or more hybrid nucleic acid molecules, one or more bead-bound nucleic acid molecules, one or more free nucleic acid molecules (e.g., capture probe free nucleic acid molecules, bead-bound free nucleic acid molecules), one or more labeled nucleic acid molecules, one or more unlabeled nucleic acid molecules, one or more amplicons, one or more unamplified nucleic acid molecules, or combinations thereof. Enriched nucleic acid molecules can be distinguished from non-enriched nucleic acid molecules by GC content, molecular size, genomic region, genomic region characteristics, or combinations thereof. Enriched nucleic acid molecules can be derived from one or more assays, supernatants, elution buffers, or combinations thereof. Enriched nucleic acid molecules can be distinguished from non-enriched nucleic acid molecules by average size, average GC content, genomic region, or combinations thereof.

[0157] Some implementations may include one or more isolation or purification reactions on one or more nucleic acid molecules in a sample. Isolation or purification reactions may include contacting the sample with one or more beads or groups of beads. Isolation or purification reactions may include one or more hybridization reactions, enrichment reactions, amplification reactions, sequencing reactions, or combinations thereof. Isolation or purification reactions may include using one or more separators. One or more separators may include magnetic separators. Isolation or purification reactions may include separating bead-bound nucleic acid molecules from bead-free nucleic acid molecules. Isolation or purification reactions may include separating nucleic acid molecules hybridized to capture probes from capture probe-free nucleic acid molecules. Isolation or purification reactions may include separating a first subset of nucleic acid molecules from a second subset of nucleic acid molecules, wherein the first subset of nucleic acid molecules and the second subset of nucleic acid molecules differ in average size, average GC content, genomic region, or a combination thereof.

[0158] Some implementations may include performing one or more elution reactions on one or more nucleic acid molecules in a sample. Elution reactions may include contacting the sample with one or more beads or groups of beads. Elution reactions may include separating bead-bound nucleic acid molecules from bead-free nucleic acid molecules. Elution reactions may include separating capture probe-hybridized nucleic acid molecules from capture probe-free nucleic acid molecules. Elution reactions may include separating a first subset of nucleic acid molecules from a second subset of nucleic acid molecules, wherein the first subset and the second subset of nucleic acid molecules differ in average size, average GC content, genomic region, or a combination thereof.

[0159] Some implementations may include one or more fragmentation reactions. Fragmentation reactions may include fragmenting one or more nucleic acid molecules in a sample or nucleic acid molecule subgroup to produce one or more fragmented nucleic acid molecules. One or more nucleic acid molecules may be fragmented by sonication, needle shearing, nebulization, shearing (e.g., acoustic shearing, mechanical shearing, spot shearing), by French pressure chamber, or by enzymatic digestion. Enzymatic digestion may be performed by nuclease digestion (e.g., micrococcal nuclease digestion, endonuclease, exonuclease, RNase H, or DNase I). Fragmentation of one or more nucleic acid molecules may result in fragment sizes of about 100 base pairs to about 2000 base pairs, about 200 base pairs to about 1500 base pairs, about 200 base pairs to about 1000 base pairs, about 200 base pairs to about 500 base pairs, about 500 base pairs to about 1500 base pairs, and about 500 base pairs to about 1000 base pairs. One or more fragmentation reactions can result in fragments of about 50 base pairs to about 1000 base pairs. One or more fragmentation reactions can produce fragments of about 100 base pairs, 150 base pairs, 200 base pairs, 250 base pairs, 300 base pairs, 350 base pairs, 400 base pairs, 450 base pairs, 500 base pairs, 550 base pairs, 600 base pairs, 650 base pairs, 700 base pairs, 750 base pairs, 800 base pairs, 850 base pairs, 900 base pairs, 950 base pairs, 1000 base pairs, or more.

[0160] Fragmentation of one or more nucleic acid molecules may include mechanically shearing one or more nucleic acid molecules in a sample for a period of time. The fragmentation reaction may occur for at least about 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 125, 150, 175, 200, 225, 250, 275, 300, 325, 350, 375, 400, 425, 450, 475, 500 seconds or more.

[0161] Fragmenting one or more nucleic acid molecules may include contacting a nucleic acid sample with one or more beads. Fragmenting one or more nucleic acid molecules may also include contacting a nucleic acid sample with a plurality of beads, wherein the volume ratio of the plurality of beads to the volume of the nucleic acid sample is about 0.10, 0.20, 0.30, 0.40, 0.50, 0.60, 0.70, 0.80, 0.90, 1.00, 1.10, 1.20, 1.30, 1.40, 1.50, 1.60, 1.70, 1.80, 1.90, 2.00 or more. Fragmenting one or more nucleic acid molecules may include contacting a nucleic acid sample with a plurality of beads, wherein the volume ratio of the plurality of beads to the volume of the nucleic acid sample is about 2.00, 1.90, 1.80, 1.70, 1.60, 1.50, 1.40, 1.30, 1.20, 1.10, 1.00, 0.90, 0.80, 0.70, 0.60, 0.50, 0.40, 0.30, 0.20, 0.10, 0.05, 0.04, 0.03, 0.02, 0.01 or less.

[0162] Some implementations may include performing one or more detection reactions on one or more nucleic acid molecules in a sample. The detection reaction may include one or more sequencing reactions. Optionally, performing the detection reaction may include optical sensing, electrical sensing, or a combination thereof. Optical sensing may include photoluminescence photon emission, fluorescence photon emission, pyrophosphate photon emission, chemiluminescence photon emission, or a combination thereof. Electrical sensing may include sensing based on ion concentration, ion current modulation, nucleotide electric field, nucleotide tunneling current, or a combination thereof.

[0163] Some implementation schemes may include performing one or more quantification reactions on one or more nucleic acid molecules in a sample. Quantification reactions may include sequencing, PCR, qPCR, digital PCR, or a combination of the above.

[0164] Some implementations may include one or more samples. Some implementations may include 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100 or more samples. Samples may be derived from an object. Two or more samples may be derived from a single object. Two or more samples may be derived from 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100 or more different objects. Objects may be mammals, reptiles, amphibians, birds, and fish. Mammals can be humans, apes, orangutans, monkeys, chimpanzees, cows, pigs, horses, rodents, birds, reptiles, dogs, cats, or other animals. Reptiles may be lizards, snakes, alligators, turtles, crocodiles, and tortoises. Amphibians may be toads, frogs, salamanders, and fire salamanders. Examples of birds include, but are not limited to, ducks, geese, penguins, ostriches, and owls. Examples of fish include, but are not limited to, catfish, eels, sharks, and swordfish. Preferably, the object is a human being. The object may suffer from a disease or condition (e.g., cancer).

[0165] Two or more samples can be collected at 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 15, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000 or more time points. Time points can occur within time intervals of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 30, 35, 40, 45, 50, 55, 60 or more hours. The time point can occur within a time period of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 30, 35, 40, 45, 50, 55, 60 or more days. The time point can also occur within a time period of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 30, 35, 40, 45, 50, 55, 60 or more weeks. The time point can occur within a time period of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 30, 35, 40, 45, 50, 55, 60 or more months. The time point can also occur within a time period of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 30, 35, 40, 45, 50, 55, 60 or more years.

[0166] Samples can come from bodily fluids, cells, skin, tissues, organs, or a combination of the above. Samples can be blood, plasma, blood fractions, saliva, sputum, urine, semen, vaginal fluid, cerebrospinal fluid, stool, cells, or tissue biopsies. Samples can come from the adrenal glands, appendix, bladder, brain, ear, esophagus, eye, gallbladder, heart, kidneys, large intestine, liver, lungs, oral cavity, muscle, nose, pancreas, parathyroid glands, pineal gland, pituitary gland, skin, small intestine, spleen, stomach, thymus, thyroid gland, trachea, uterus, appendix, cornea, skin, heart valves, arteries, or veins.

[0167] The sample may contain one or more nucleic acid molecules. Nucleic acid molecules can be DNA molecules, RNA molecules (e.g., mRNA, cRNA, or miRNA), and DNA / RNA hybrids. Examples of DNA molecules include, but are not limited to, double-stranded DNA, single-stranded DNA, single-stranded DNA hairpins, cDNA, and genomic DNA. Nucleic acid molecules can be RNA molecules, such as double-stranded RNA, single-stranded RNA, ncRNA, RNA hairpins, and mRNA. Examples of ncRNAs include, but are not limited to, siRNA, miRNA, snoRNA, piRNA, tiRNA, PASR, TASR, aTASR, TSSa-RNA, snRNA, RE-RNA, uaRNA, x-ncRNA, hY RNA, usRNA, snRNA, and vtRNA.

[0168] Some embodiments may include one or more containers. Certain embodiments may include one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, one hundred or more, 125 or more, 150 or more, 175 or more, 200 or more, 250 or more, 300 or more, 350 or more, 400 or more, 500 or more, 600 or more, 700 or more, 800 or more, 900 or more, or one thousand or more containers. The one or more containers may be different, similar, identical, or a combination of more than one. Examples of containers include, but are not limited to, plates, microplates, PCR plates, wells, microwells, tubes, Eppendorf tubes, vials, arrays, microarrays, and chips.

[0169] Some embodiments may include one or more reagents. Certain embodiments may include one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, ten or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, or ten thousand or more reagents. One or more reagents may be different, similar, identical, or combinations thereof. Reagents may improve the efficiency of one or more assays. Reagents can enhance the stability of nucleic acid molecules or their variants or derivatives. Reagents can include, but are not limited to, enzymes, proteases, nucleases, molecules, polymerases, reverse transcriptases, ligases, and compounds. Some embodiments may include performing an assay containing one or more antioxidants. Typically, an antioxidant is a molecule that inhibits the oxidation of another molecule. Examples of antioxidants include, but are not limited to, ascorbic acid (e.g., vitamin C), glutathione, lipoic acid, uric acid, carotene, α-tocopherol (e.g., vitamin E), panthenol (e.g., coenzyme Q), and vitamin A.

[0170] Some embodiments may include one or more buffers or solutions. These buffers or solutions may be different, similar, identical, or combinations thereof. Buffers or solutions may improve the efficiency of one or more assays. Buffers or solutions may improve the stability of nucleic acid molecules or their variants or derivatives. Buffers or solutions may include, but are not limited to, wash buffers, elution buffers, and hybridization buffers.

[0171] Some embodiments may include one or more beads, multiple beads, or groups of one or more beads. Some embodiments may include one or more nucleic acid molecules in a sample in one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, fifteen or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, one hundred or more, two hundred or more, three hundred or more, four hundred or more, five hundred or more, six hundred or more, seven hundred or more, eight hundred or more, nine hundred or more, or one thousand or more, or one thousand or more. The one or more beads or groups of beads may be different, similar, identical, or combinations thereof. The beads can be magnetic, antibody-coated, protein A cross-linked, protein G cross-linked, streptavidin-coated, oligonucleotide-conjugated, silica-coated, or a combination thereof. Examples of beads include, but are not limited to: Ampure beads, AMPure XP beads, streptavidin beads, agarose beads, magnetic beads, etc. Microbeads, antibody-conjugated beads (e.g., anti-immunoglobulin microbeads), protein A-conjugated beads, protein G-conjugated beads, protein A / G-conjugated beads, protein L-conjugated beads, oligomeric dT-conjugated beads, silica beads, silica-like beads, anti-biotin microbeads, anti-fluorescent dye microbeads, and BcMag™ carboxyl-terminated magnetic beads. In some aspects of this disclosure, one or more beads comprise one or more Ampure beads. Optionally or additionally, one or more beads comprise AMPure XP beads.

[0172] Some implementations may include one or more primers, multiple primers, or sets of one or more primers. Primers may also contain one or more adapters. Primers may also contain more or more markers. Primers may be used for one or more assays. For example, primers may be used for one or more sequencing reactions, amplification reactions, or combinations thereof. Some implementations may include one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, ten or more, twenty or more, thirty or more, forty ... Primers may contain about 10 to about 500 nucleotides, about 20 to about 450 nucleotides, about 30 to about 400 nucleotides, about 40 to about 350 nucleotides, about 50 to about 300 nucleotides, about 60 to about 250 nucleotides, about 70 to about 200 nucleotides, or about 80 to about 150 nucleotides. In some aspects of this disclosure, primers contain about 80 nucleotides to about 100 nucleotides. One or more primers or primer sets may be different, similar, identical, or combinations of more than one.

[0173] Primers can hybridize with at least a portion of one or more nucleic acid molecules or their variants or derivatives in a subset of samples or nucleic acid molecules. Primers can hybridize with one or more genomic regions. Primers can hybridize with different, similar, and / or identical genomic regions. One or more primers can have at least about 50%, 55%, 60%, 65%, 70%, 75%, 80%, 85%, 90%, 95%, 97%, 99%, or more complementarity with one or more nucleic acid molecules or their variants or derivatives.

[0174] Primers may contain one or more nucleotides. Primers can contain one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, one hundred or more, 125 or more, 150 or more, 175 or more, two hundred or more, two hundred or more, two hundred or more, three hundred or more, three hundred or more, four hundred or more, five hundred or more, six hundred or more, seven hundred or more, eight hundred or more, nine hundred or more, or one hundred or more. Primers may contain approximately 100 nucleotides. Primers may contain about 10 to about 500 nucleotides, about 20 to about 450 nucleotides, about 30 to about 400 nucleotides, about 40 to about 350 nucleotides, about 50 to about 300 nucleotides, about 60 to about 250 nucleotides, about 70 to about 200 nucleotides, or about 80 to about 150 nucleotides. In some aspects of this disclosure, primers contain about 80 nucleotides to about 100 nucleotides.

[0175] Multiple primers or primer sets may contain two or more primers having the same, similar, and / or different sequences, adapters, and / or markers. For example, two or more primers contain the same sequence. In another instance, two or more primers contain similar sequences. In yet another instance, two or more primers contain different sequences. Two or more primers may further contain one or more adapters. Two or more primers may further contain different adapters. Two or more primers may further contain similar adapters. Two or more primers may further contain the same adapter. Two or more primers may further contain one or more markers. Two or more primers may further contain different markers. Two or more primers may further contain similar markers. Two or more primers may further contain the same marker.

[0176] Capture probes, primers, tags, and / or beads may contain one or more nucleotides. One or more nucleotides may contain RNA, DNA, a mixture of DNA and RNA residues or analogs of modifications thereof, such as 2'-OMe or 2'-fluoro (2'-F), locked nucleic acid (LNA), or a base-free site.

[0177] Some embodiments may include one or more marks. Certain embodiments may include one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, one hundred or more, 125 or more, 150 or more, 175 or more, 200 or more, 250 or more, 300 or more, 350 or more, 400 or more, 500 or more, 600 or more, 700 or more, 800 or more, 900 or more, or one thousand or more marks. One or more marks may be different, similar, identical, or combinations thereof.

[0178] Examples of labels include, but are not limited to, chemical, biochemical, biological, colorimetric, enzymatic, fluorescent, and luminescent labels well known in the art. Labels include dyes, photocrosslinkers, cytotoxic compounds, drugs, affinity labels, photoaffinity labels, reactive compounds, antibodies or antibody fragments, biomaterials, nanoparticles, spin labels, fluorophores, metal-containing moieties, radioactive moieties, novel functional groups, groups that interact covalently or non-covalently with other molecules, photocage moieties, photoelectrochemically excitable moieties, ligands, photoisomerization moieties, biotin, biotin analogs, moieties doped with heavy atoms, chemically cleavable groups, photocleavable groups, redox activators, isotope-labeled moieties, biophysical probes, phosphorescent groups, chemiluminescent groups, electron-dense groups, magnetic groups, intercalation groups, chromophores, energy transfer agents, bioactive agents, detectable labels, or combinations thereof.

[0179] The label can be a chemical label. Examples of chemical labels may include, but are not limited to, biotin and radioactive subtypes (e.g., iodine, carbon, phosphate, hydrogen).

[0180] The methods, kits, and compositions disclosed herein may include biomarkers. Biomarkers may include metabolic markers, including but not limited to bioorthogonally modified amino acids, sugars, and other compounds.

[0181] The methods, kits, and compositions disclosed herein may include enzyme labeling. Enzyme labeling may include, but is not limited to, horseradish peroxidase (HRP), alkaline phosphatase (AP), glucose oxidase, and O-galactosidase. Enzyme labeling may also include luciferase.

[0182] The methods, kits, and compositions disclosed herein may include fluorescent labels. Fluorescent labels may be organic dyes (e.g., FITC), biofluoresceins (e.g., green fluorescent protein), or quantum dots. A non-limiting list of fluorescent labels includes fluorescein isothiocyanate (FITC), DyLight Fluors, fluorescein, rhodamine (tetramethylrhodamine isothiocyanate, TRITC), coumarin, fluorescein yellow, and BODIPY. The label may be a fluorophore. Exemplary fluorophores include, but are not limited to, indolecarbonylcyanine (C3), indoledicarbonylcyanine (C5), Cy3, Cy3.5, Cy5, Cy5.5, Cy7, Texas Red, Pacific Blue, Oregon Green 488, and Alexa. -355, Alexa Fluor 488, Alexa Fluor 532, Alexa Fluor 546, Alexa Fluor-555, Alexa Fluor 568, Alexa Fluor 594, Alexa Fluor 647, Alexa Fluor 660, Alexa Fluor 680, JOE, Lissamine, Rhodamine Green, BODIPY, FITC (Fluorescence Isothiocyanate), FAM (Fluorescence Aminocyanate), Phycoerythrin, Rhodamine, dRhodamine, TAMRA (Carboxy-TetramethylRhodamine), ROX™ (Carboxy-X-Rhodamine), LIZ™, VICTM NED™ PET™, SYBR, PicoGreen, RiboGreen, etc. Fluorescent labels can be green fluorescent protein (GFP), red fluorescent protein (RFP), yellow fluorescent protein, and phycobiliproteins (e.g., allophycocyanin, phycocyanin, phycoerythrin, and phycoerythrocyanin).

[0183] Some embodiments may include one or more connectors. Certain embodiments may include one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, twenty or more, thirty or more, forty or more, fifty or more, sixty or more, seventy or more, eighty or more, ninety or more, one hundred or more, 125 or more, 150 or more, 175 or more, 200 or more, 250 or more, 300 or more, 350 or more, 400 or more, 500 or more, 600 or more, 700 or more, 800 or more, 900 or more, or one thousand or more connectors. The one or more connectors may be different, similar, identical, or a combination of more than one.

[0184] Suitable adapters include any chemical or biological compound capable of attaching to the labels, primers, and / or capture probes disclosed herein. If an adapter attaches to a label and primer or capture probe, a suitable adapter will be able to adequately separate the label and primer or capture probe. A suitable adapter will not significantly interfere with the ability of the primer and / or capture probe to hybridize with nucleic acid molecules, portions thereof, or variants or derivatives thereof. A suitable adapter will not significantly interfere with the ability to detect the label. Adapters can be rigid. Adapters can be flexible. Adapters can be semi-rigid. Adapters can be proteolytically stable (e.g., resistant to proteolytic cleavage). Adapters can be proteolytically unstable (e.g., sensitive to proteolytic cleavage). Adapters can be helical. Adapters can be non-helical. Adapters can be coiled. The linker can be a (3-chain). The linker can contain a turn conformation. The linker can be a single chain. The linker can be a long chain. The linker can be a short chain. The linker can contain at least about 5 residues, at least about 10 residues, at least about 15 residues, at least about 20 residues, at least about 25 residues, at least about 30 residues, or at least about 40 residues or more.

[0185] Examples of linkers include, but are not limited to, hydrazones, disulfides, thioethers, and peptide linkers. Linkers can be peptide linkers. Peptide linkers can contain proline residues. Peptide linkers can contain arginine, phenylalanine, threonine, glutamine, glutamic acid, or any combination thereof. Linkers can be heterobifunctional crosslinking agents.

[0186] Some implementations may include performing one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, ten or more, eleven or more, twelve or more, thirteen or more, fourteen or more, fifteen or more, twenty or more, twenty-five or more, thirty or more, thirty-five or more, forty or more, forty-five or more, or fifty or more assays on a sample comprising one or more nucleic acid molecules. The two or more assays may be different, similar, identical, or a combination thereof. For example, some implementations include performing two or more sequencing reactions. In another example, some implementations include performing two or more assays, wherein at least one of the two or more assays includes a sequencing reaction. In yet another example, some implementations include performing two or more assays, wherein at least two of the two or more assays include a sequencing reaction and a hybridization reaction. The two or more assays may be performed sequentially, simultaneously, or in a combination thereof. For example, two or more sequencing reactions can be performed simultaneously. In another example, some embodiments include performing a hybridization reaction followed by a sequencing reaction. In yet another example, some embodiments include performing two or more hybridization reactions simultaneously, followed by two or more sequencing reactions simultaneously. Two or more assays can be performed by one or more devices. For example, two or more amplification reactions can be performed by a PCR machine. In another example, two or more sequencing reactions can be performed by two or more sequencers.

[0187] C. Device

[0188] Some embodiments may include one or more devices. Some embodiments may include one or more assays that include one or more devices. Some embodiments may include using one or more devices to perform one or more steps or assays. Some embodiments may include using one or more devices in one or more steps or assays. For example, performing a sequencing reaction may include one or more sequencers. In another example, generating a subset of nucleic acid molecules may include using one or more magnetic separators. In yet another example, one or more processors may be used to analyze one or more nucleic acid samples. Exemplary devices include, but are not limited to, sequencers, thermal cyclers, real-time PCR instruments, magnetic separators, transport devices, hybridization chambers, electrophoresis apparatus, centrifuges, microscopes, imagers, fluorometers, photometers, plate readers, computers, processors, and bioanalysts.

[0189] Some implementations may include one or more sequencers. One or more sequencers may include one or more HiSeq, MiSeq, HiScan, Genome Analyzer IIx, SOLiD sequencers, Ion Torrent PGM, 454GS Junior, Pac Bio RS, or combinations thereof. One or more sequencers may include one or more sequencing platforms. One or more sequencing platforms may include 454Life Technologies / Roche's GS FLX, Solexa / Illumina's Genome Analyzer, Applied Biosystems' SOLiD, Complete Genomics' CGA Platform, Pacific Biosciences' PacBio RS, or combinations thereof.

[0190] Some implementations may include one or more thermal cyclers. One or more thermal cyclers can be used to amplify one or more nucleic acid molecules. Some implementations may include one or more real-time PCR instruments. One or more real-time PCR instruments may include a thermal cycler and a fluorometer. One or more thermal cyclers can be used to amplify and detect one or more nucleic acid molecules.

[0191] Some implementations may include one or more magnetic separators. One or more magnetic separators can be used to separate paramagnetic and ferromagnetic particles from a suspension. One or more magnetic separators may include one or more LifeStep™ biomagnetic separators, SPHEROTM FlexiMag separators, SPHEROTM MicroMag separators, SPHEROTM HandiMag separators, SPHEROTM MiniTube Mag separators, SPHEROTM UltraMag separators, DynaMag™ magnets, DynaMag™-2 magnets, or combinations thereof.

[0192] Some implementations may include one or more bioanalysts. Generally, a bioanalyst is a chip-based capillary electrophoresis apparatus that can analyze RNA, DNA, and proteins. One or more bioanalysts may include the Agilent 2100 bioanalyst.

[0193] Some implementations may include one or more processors. One or more processors may analyze, compile, store, sort, combine, evaluate, or otherwise process one or more data and / or results from one or more assays, based on or derived from one or more data and / or results from one or more assays, one or more outputs from one or more assays, based on or derived from one or more assays, one or more outputs from one or more data and / or results, one or more outputs based on or derived from one or more data and / or results, or a combination thereof. One or more processors may transmit one or more data, results, or outputs from one or more assays, based on or derived from one or more assays, one or more outputs from one or more data or results, one or more outputs based on or derived from one or more data or results, or a combination thereof. One or more processors may receive and / or store requests from users. One or more processors may generate or produce one or more data, results, or outputs. One or more processors may generate or produce one or more biomedical reports. One or more processors may transmit one or more biomedical reports. One or more processors may analyze, compile, store, sort, combine, evaluate, or otherwise process information from one or more databases, one or more data or results, one or more outputs, or a combination thereof. One or more processors can analyze, compile, store, sort, combine, evaluate, or otherwise process information from 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 30, or more databases. One or more processors can transmit one or more requests, data, results, outputs, and / or information to one or more users, processors, computers, computer systems, memory locations, devices, databases, or a combination thereof. One or more processors can receive one or more requests, data, results, outputs, and / or information from one or more users, processors, computers, computer systems, memory locations, devices, databases, or a combination thereof. One or more processors can retrieve one or more requests, data, results, outputs, and / or information from one or more users, processors, computers, computer systems, memory locations, devices, databases, or a combination thereof.

[0194] Some implementations may include one or more memory locations. These memory locations may store information, data, results, output, requests, or a combination thereof. These memory locations may receive information, data, results, output, requests, or a combination thereof from one or more users, processors, computers, computer systems, devices, or a combination thereof.

[0195] The methods described herein can be implemented using one or more computers and / or computer systems. A computer or computer system may include an electronic storage location (e.g., a database, a memory) having machine-executable code for implementing the methods provided herein, and one or more processors for executing the machine-executable code.

[0196] The code can be pre-compiled and configured for use with a machine having a processor suitable for executing the code, or it can be compiled at runtime. The code can be provided in a programming language, which can be selected to enable the code to be executed in a pre-compiled or compiled manner.

[0197] One or more computers and / or computer systems can analyze, compile, store, sort, combine, evaluate, or otherwise process one or more data and / or results from one or more assays, based on or derived from one or more data and / or results from one or more assays, one or more outputs from one or more assays, based on or derived from one or more assays, one or more outputs from one or more data and / or results, or a combination thereof. One or more computers and / or computer systems can transmit one or more data, results, or outputs from one or more assays, based on or derived from one or more assays, one or more outputs from one or more data or results, based on or derived from one or more data or results, or a combination thereof. One or more computers and / or computer systems can receive and / or store requests from users. One or more computers and / or computer systems can generate or produce one or more data, results, or outputs. One or more computers and / or computer systems can generate or produce one or more biomedical reports. One or more computers and / or computer systems can transmit one or more biomedical reports. One or more computers and / or computer systems can analyze, compile, store, sort, combine, evaluate, or otherwise process information from one or more databases, one or more data or results, one or more outputs, or combinations thereof. One or more computers and / or computer systems can analyze, compile, store, sort, combine, evaluate, or otherwise process information from 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 30, or more databases. One or more computers and / or computer systems can transmit one or more requests, data, results, outputs, and / or information to one or more users, processors, computers, computer systems, memory locations, devices, or combinations thereof. One or more computers and / or computer systems can receive one or more requests, data, results, outputs, and / or information from one or more users, processors, computers, computer systems, memory locations, devices, or combinations thereof. One or more computers and / or computer systems can retrieve one or more requests, data, results, outputs and / or information from one or more users, processors, computers, computer systems, memory locations, devices, databases or combinations thereof.

[0198] D. Database

[0199] Some implementations may include one or more databases. Some implementations may include at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 30 or more databases. Databases may include genomics, proteomics, pharmacogenomics, biomedical and scientific databases. Databases may be publicly available databases. Optionally or additionally, databases may include proprietary databases. Databases may be commercially available databases. Databases include, but are not limited to, Cosmic, GnomAD, DBSNP, Mills Indels, MendelDB, PharmGKB, Varimed, Regulome, Curated BreakSeq junctions, Human Mendelian Inheritance Online (OMIM), Human Genome Mutation Database (HGMD), NCBI db SNP, NCBI RefSeq, GENCODE, GO (Gene Ontology), and Kyoto Encyclopedia of Genes and Genomes (KEGG).

[0200] Some implementations may include analyzing one or more databases. Some implementations may include analyzing at least about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 30 or more databases. Analyzing one or more databases may include one or more algorithms, computers, processors, memory locations, devices or combinations thereof.

[0201] Some implementations may include identifying one or more nucleic acid regions based on data and / or information from one or more databases. Some implementations may include identifying one or more sets of nucleic acid regions based on data and / or information from one or more databases. Some implementations may include identifying one or more nucleic acid regions and / or one or more sets of nucleic acid regions based on data and / or information from at least about two databases. Some implementations may include identifying one or more nucleic acid regions and / or one or more sets of nucleic acid regions based on data and / or information from at least about three databases. Some implementations may include identifying one or more nucleic acid regions and / or one or more sets of nucleic acid regions based on data and / or information from at least about 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 30 or more databases.

[0202] Some implementations may include analyzing one or more results based on data and / or information from one or more databases. Some implementations may include analyzing one or more sets of results based on data and / or information from one or more databases. Some implementations may include analyzing one or more merged results based on data and / or information from one or more databases. Some embodiments may include analyzing one or more results, result sets, and / or merged results based on data and / or information from at least about two or more databases. Some embodiments may include analyzing one or more results, result sets, and / or merged results based on data and / or information from at least about three or more databases. Some implementations may include analyzing one or more results, result sets, and / or merged results based on data and / or information from at least about four, five, six, seven, eight, nine, ten, eleven, thirteen, eleven, twelfth, thirteenth, eleventh, eleventh, thirteenth, and thirteenth or more databases.

[0203] Some implementations may include comparing one or more results based on data and / or information from one or more databases. Some implementations may include comparing one or more sets of results based on data and / or information from one or more databases. Some implementations may include comparing one or more merged results based on data and / or information from one or more databases. Some embodiments may include comparing one or more results, result sets, and / or merged results based on data and / or information from at least about two or more databases. Some embodiments may include comparing one or more results, result sets, and / or merged results based on data and / or information from at least about three or more databases. Some implementations may include comparing one or more results, result sets, and / or merged results based on data and / or information from at least about four, five, six, seven, eight, nine, ten, eleven, thirteen, thirteen, eleven, twentieth, thirteenth, eleventh, eleventh, eleventh, thirteenth, and thirteenth, or ...

[0204] Some implementations may include biomedical databases, genomic databases, biomedical reports, disease reports, case-control analyses, and rare variant discovery analyses based on data and / or information from one or more databases, one or more assays, one or more data or results, one or more outputs based on or derived from one or more assays, one or more outputs based on or derived from one or more data or results, or a combination of the above.

[0205] E. Datasets and Analysis

[0206] Some implementations may include one or more data, one or more datasets, one or more merged data, one or more merged datasets, one or more results, a set or more sets of results, one or more merged results, or a combination thereof. Data and / or results may be based on or derived from one or more measurements, one or more databases, or a combination thereof. Some implementations may include analyzing one or more data, one or more datasets, one or more merged data, one or more merged datasets, one or more results, a set or more sets of results, one or more merged results, or a combination thereof. Some implementations may include processing one or more data, one or more datasets, one or more merged data, one or more merged datasets, one or more results, a set or more sets of results, one or more merged results, or a combination thereof.

[0207] Some implementations may include at least one analysis and at least one processing of one or more data, one or more datasets, one or more merged data, one or more merged datasets, one or more results, a set or more sets of results, one or more merged results, or a combination thereof. Some implementations may include one or more analyses and one or more processing of one or more data, one or more datasets, one or more merged data, one or more merged datasets, one or more results, a set or more sets of results, one or more merged results, or a combination thereof. Some implementations may include at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000 or more unique analyses of one or more data, one or more datasets, one or more merged data, one or more merged datasets, one or more results, a set or more sets of results, one or more merged results, or a combination thereof. Some implementation schemes may include at least 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000 or more unique processing of one or more data, one or more datasets, one or more merged data, one or more merged datasets, one or more results, a set or more sets of results, one or more merged results, or a combination of the above. One or more analyses and / or one or more processing may occur simultaneously, sequentially or in a combination of the above.

[0208] One or more analyses and / or one or more treatments may occur within 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 15, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000 or more time points. Time points may occur within time intervals of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 30, 35, 40, 45, 50, 55, 60 or more hours. The time point can occur within a time period of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 30, 35, 40, 45, 50, 55, 60 or more days. The time point can also occur within a time period of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 30, 35, 40, 45, 50, 55, 60 or more weeks. The time point can occur within a time period of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 30, 35, 40, 45, 50, 55, 60 or more months. The time point can also occur within a time period of 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 30, 35, 40, 45, 50, 55, 60 or more years.

[0209] Some implementations may include one or more data. One or more data may include one or more raw data based on or derived from one or more assays. One or more data may include one or more raw data based on or derived from one or more databases. One or more data may include data at least partially based on or derived from one or more raw data analyses. One or more data may include data at least partially based on or derived from one or more raw data processes. One or more data may include data entirely based on or derived from one or more raw data analyses. One or more data may include data entirely based on or derived from one or more raw data processes. Data may include sequencing read data or expression data. Data may include biomedical, scientific, pharmacological, and / or genetic information.

[0210] Some implementations may include one or more merged data sets. One or more merged data sets may include two or more datasets. One or more merged data sets may include two or more datasets. One or more merged data sets may include one or more raw data sets based on or derived from one or more assays. One or more merged data sets may include one or more raw data sets based on or derived from one or more databases. One or more merged data sets may include data at least partially based on or derived from one or more raw data analyses. One or more merged data sets may include data at least partially based on or derived from one or more raw data processing methods. One or more merged data sets may include data entirely based on or derived from one or more raw data analyses. One or more merged data sets may include data entirely based on or derived from one or more raw data processing methods. One or more merged data sets may include sequencing read data or expression data. One or more merged data sets may include biomedical, scientific, pharmacological, and / or genetic information.

[0211] Some implementations may include one or more datasets. One or more datasets may include one or more data sets. One or more datasets may include one or more merged data sets. One or more datasets may include one or more raw data sets based on or derived from one or more assays. One or more datasets may include one or more raw data sets based on or derived from one or more databases. One or more datasets may include data at least partially based on or derived from one or more raw data analyses. One or more datasets may include data at least partially based on or derived from one or more raw data processes. One or more datasets may include data entirely based on or derived from one or more raw data analyses. One or more datasets may include data entirely based on or derived from one or more raw data processes. Datasets may include sequencing read data or expression data. Datasets may include biomedical, scientific, pharmacological, and / or genetic information.

[0212] Some implementations may include one or more merged datasets. One or more merged datasets may include two or more datasets. One or more merged datasets may include two or more merged datasets. One or more merged datasets may include one or more raw data based on or derived from one or more assays. One or more merged datasets may include one or more raw data based on or derived from one or more databases. One or more merged datasets may include data at least partially based on or derived from one or more raw data analyses. One or more merged datasets may include data at least partially based on or derived from one or more raw data processes. One or more merged datasets may include data entirely based on or derived from one or more raw data analyses. One or more merged datasets may include data entirely based on or derived from one or more raw data processes. Some implementations may also include further processing and / or analysis of the merged datasets. One or more merged datasets may include sequencing read data or expression data. One or more merged datasets may include biomedical, scientific, pharmacological, and / or genetic information.

[0213] Some implementations may include one or more results. One or more results may include one or more data, datasets, merged data, and / or merged datasets. One or more results may be based on or derived from one or more data, datasets, merged data, and / or merged datasets. One or more results may be generated from one or more assays. One or more results may be based on or derived from one or more assays. One or more results may be based on or derived from one or more databases. One or more results may include results from analyses at least partially based on or derived from one or more data, datasets, merged data, and / or merged datasets. One or more results may include results from processing at least partially based on or derived from one or more data, datasets, merged data, and / or merged datasets. One or more results may include results from analyses entirely based on or derived from one or more data, datasets, merged data, and / or merged datasets. One or more results may include results from processing entirely based on or derived from one or more data, datasets, merged data, and / or merged datasets. Results may include sequencing read data or expression data. Results may include biomedical, scientific, pharmacological, and / or genetic information.

[0214] Some implementations may include one or more sets of results. One or more sets of results may include one or more data, datasets, pooled data, and / or pooled datasets. One or more sets of results may be based on or derived from one or more data, datasets, pooled data, and / or pooled datasets. One or more sets of results may be generated from one or more assays. One or more sets of results may be based on or derived from one or more assays. One or more sets of results may be based on or derived from one or more databases. One or more sets of results may include a result set analyzed at least partially based on or derived from one or more data, datasets, pooled data, and / or pooled datasets. One or more sets of results may include a result set processed at least partially based on or derived from one or more data, datasets, pooled data, and / or pooled datasets. One or more sets of results may include a result set analyzed entirely based on or derived from one or more data, datasets, pooled data, and / or pooled datasets. One or more sets of results may include a result set processed entirely based on or derived from one or more data, datasets, pooled data, and / or pooled datasets. The result set may include sequencing read data or expression data. The result set may include biomedical, scientific, pharmacological, and / or genetic information.

[0215] Some implementations may include one or more merged results. Merged results may include one or more results, result sets, and / or merged result sets. Merged results may be based on or derived from one or more results, result sets, and / or merged result sets. One or more merged results may include one or more data, datasets, merged data, and / or merged datasets. One or more merged results may be based on or derived from one or more data, datasets, merged data, and / or merged datasets. One or more merged results may be generated from one or more assays. One or more merged results may be based on or derived from one or more assays. One or more merged results may be based on or derived from one or more databases. One or more merged results may include merged results that are at least partially based on or derived from the analysis of one or more data, datasets, merged data, and / or merged datasets. One or more merged results may include merged results that are at least partially based on or derived from the processing of one or more data, datasets, merged data, and / or merged datasets. One or more merged results may include merged results that are entirely based on or derived from the analysis of one or more data, datasets, merged data, and / or merged datasets. One or more merged results may include merged results that are entirely based on or derived from the processing of one or more data, datasets, merged data, and / or merged datasets. Merged results may include sequencing read data or expression data. The results of the merging can include biomedical, scientific, pharmacological, and / or genetic information.

[0216] Some implementations may include one or more combined results. A combined result set may include one or more results, result sets, and / or combined results. A combined result set may be based on or derived from one or more results, result sets, and / or combined results. A combined result set may include one or more data, datasets, merged data, and / or merged datasets. A combined result set may be based on or derived from one or more data, datasets, merged data, and / or merged datasets. A combined result set may be generated from one or more measurements. A combined result set may be based on or derived from one or more measurements. A combined result set may be based on or derived from one or more databases. A combined result set may include a combined result set analyzed at least partially based on or derived from one or more data, datasets, merged data, and / or merged datasets. A combined result set may include a combined result set processed at least partially based on or derived from one or more data, datasets, merged data, and / or merged datasets. A combined result set may include a combined result set analyzed entirely based on or derived from one or more data, datasets, merged data, and / or merged datasets. One or more merged results may include merged result sets that are entirely based on or derived from one or more data, datasets, merged data, and / or merged datasets. Merged result sets may include sequencing read data or expression data. Merged result sets may include biomedical, scientific, pharmacological, and / or genetic information.

[0217] Some implementations may include one or more outputs, output sets, merged outputs, and / or merged output sets. The methods, libraries, kits, and systems described herein may generate one or more outputs, output sets, merged outputs, and / or merged output sets. An output set may include one or more outputs, one or more merged outputs, or a combination thereof. A merged output may include one or more outputs, one or more sets of outputs, one or more sets of merged outputs, or a combination thereof. A merged output set may include one or more outputs, one or more sets of outputs, one or more merged outputs, or a combination thereof. One or more outputs, output sets, merged outputs, and / or merged output sets may be based on or derived from one or more data, one or more datasets, one or more merged data sets, one or more merged datasets, one or more results, one or more sets of results, one or more merged results, or a combination thereof. One or more outputs, output sets, merged outputs, and / or merged output sets may be based on or derived from one or more databases. One or more outputs, output sets, merged outputs, and / or merged output sets may include one or more biomedical reports, biomedical outputs, rare variant outputs, pharmacogenetics outputs, population study outputs, case-control outputs, biomedical databases, genomic databases, disease databases, or web content.

[0218] Some implementations may include one or more biomedical outputs, a set or more sets of biomedical outputs, one or more combined biomedical outputs, or one or more combined biomedical outputs. The methods, libraries, kits, and systems described herein may generate one or more biomedical outputs, a set or more sets of biomedical outputs, one or more combined biomedical outputs, or one or more combined biomedical outputs. A collection of biomedical outputs may include one or more biomedical outputs, one or more combined biomedical outputs, or a combination thereof. A combined biomedical output may include one or more biomedical outputs, a set or more sets of biomedical outputs, or a combination thereof. One or more biomedical outputs, a set or more sets of biomedical outputs, one or more merged biomedical outputs, or a set or more merged biomedical outputs may be based on or derived from one or more data, one or more datasets, one or more merged data sets, one or more merged datasets, one or more results, a set or more sets of results, one or more merged results, one or more outputs, a set or more sets of outputs, one or more merged outputs, or a combination of the above. One or more biomedical outputs may include biomedical information about the subject. The biomedical information about the subject may predict, diagnose, and / or predict one or more biomedical characteristics. One or more biomedical characteristics may include the state of a disease or condition, the genetic risk of a disease or condition, reproductive risk, genetic risk to the fetus, risk of adverse drug reactions, efficacy of drug treatment, prediction of optimal drug dosage, transplant tolerance, or a combination of the above.

[0219] Some implementations may include one or more biomedical reports. The methods, libraries, kits, and systems described herein may generate one or more biomedical reports. One or more biomedical reports may be based on or derived from one or more data, one or more datasets, one or more pooled data, one or more pooled datasets, one or more results, a set of results, one or more pooled results, one or more outputs, a set of outputs, one or more pooled outputs, a set of combined outputs, one or more biomedical outputs, a set of combined biomedical outputs, a set of combined biomedical outputs, or a combination thereof. Biomedical reports may predict, diagnose, and / or predict one or more biomedical characteristics. One or more biomedical characteristics may include the state of a disease or condition, the genetic risk of a disease or condition, reproductive risk, genetic risk to the fetus, risk of adverse drug reactions, efficacy of drug treatment, prediction of optimal drug dosage, transplant tolerance, or a combination thereof.

[0220] Some implementations may also include the transmission of one or more data, information, results, outputs, reports, or combinations thereof. For example, data / information based on or derived from one or more measurements is transmitted to another device and / or instrument. In another instance, a combination of data, results, outputs, biomedical outputs, biomedical reports, or the like is transmitted to another device and / or instrument. Information obtained from an algorithm may also be transmitted to another device and / or instrument. Information based on analysis of one or more databases may be transmitted to another device and / or instrument. The transmission of data / information may include the transmission of data / information from a first source to a second source. The first and second sources may be in the same general location (e.g., in the same room, building, block, or campus). Alternatively, the first and second sources may be in multiple locations (e.g., multiple cities, states, countries, continents, etc.). Data, results, outputs, biomedical outputs, and biomedical reports may be transmitted to patients and / or healthcare providers.

[0221] Transmission can be based on the analysis of one or more data, results, information, databases, outputs, reports, or a combination thereof. For example, the transmission of a second report is based on the analysis of a first report. Optionally, the transmission of a report is based on the analysis of one or more data or results. Transmission can be based on receiving one or more requests. For example, the transmission of a report can be based on receiving a request from a user (e.g., a patient, healthcare provider, individual).

[0222] Data / information transmission can include digital or analog transmission. Digital transmission can include the physical transmission of data (digital bit streams) over point-to-point or point-to-multipoint communication channels. Examples of such channels are copper wires, optical fibers, wireless communication cables, and storage media. Data can be represented as electromagnetic signals, such as voltage, radio waves, microwaves, or infrared signals.

[0223] Analog transmission can include the transmission of continuously varying analog signals. Messages can be represented by a sequence of pulses in line code (baseband transmission) or by a finite set of continuously varying waveforms using digital modulation methods (passband transmission). Passband modulation and the corresponding demodulation (also known as detection) can be performed by a modem. According to the most common definition of digital signals, both baseband and passband signals representing bitstreams are considered digital transmissions, while an alternative definition considers only baseband signals as digital signals and passband transmission of digital data as a form of digital-to-analog conversion.

[0224] Some implementations may include one or more sample identifiers. Sample identifiers may include tags, barcodes, and other indicators that can be linked to one or more samples and / or subsets of nucleic acid molecules. Some implementations may include one or more processors, one or more memory locations, one or more computers, one or more monitors, one or more computer software, or one or more algorithms for linking data, results, outputs, biomedical outputs, and / or biomedical reports to samples.

[0225] Some implementations may include a processor for correlating the expression levels of one or more nucleic acid molecules with prognostic disease outcomes. Some implementations may include one or more of a variety of related techniques, including lookup tables, algorithms, multivariate models, and linear or nonlinear combinations of expression models or algorithms. Expression levels may be converted into one or more probability scores reflecting the likelihood that a patient providing the sample may exhibit a specific disease outcome. The model and / or algorithm may be provided in a machine-readable format and may optionally further specify treatment options for patients or patient categories.

[0226] In some cases, the methods and systems described herein are used to generate outputs including the detection and / or quantification of genomic DNA regions, such as regions containing DNA polymorphisms (e.g., germline variants or somatic cell variants). In some cases, the detection of one or more genomic regions is based on one or more algorithms, depending on the source of the data input or database described elsewhere in this specification. Each of one or more algorithms can be used to receive, combine, and generate data including genomic region detection (i.e., polymorphism). In some embodiments, the methods and systems may include the detection of genomic regions based on one or more, two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, nine or more, or ten or more algorithms. Algorithms may be machine learning algorithms, computer-implemented algorithms, machine-executed algorithms, automated algorithms, etc.

[0227] Feature selection techniques can be used to analyze the data obtained from each nucleic acid sample, including filtering techniques that assess feature relevance by examining the intrinsic properties of the data, wrapping methods that embed model hypotheses into the search of a subset of features, and embedding techniques in which the search for an optimal set of features is built into the algorithm or model.

[0228] In some cases, the detection of one or more genomic regions is based on one or more statistical models. Statistical models or filtering techniques that can be used in the methods of this invention include (1) parametric methods, such as using two-sample t-tests, ANOVA analysis, Bayesian frames, and Gamma distribution models; (2) model-free methods, such as using Wilcoxon rank-sum tests, inter-class sum-of-squares tests, rank-product methods, random permutation methods, or TNoM, which involve setting a threshold point for the difference in fold change in expression between two datasets and then detecting the threshold point in each gene that minimizes the number of misclassifications; and (3) multivariate methods, such as bivariate methods, correlation-based feature selection (CFS), minimum redundancy maximum correlation (MRMR), Markov blanket filter methods, Markov models, hidden Markov models (HMMs), and uncorrelated shrinking centroid methods. In some cases, the hidden Markov model (HMM) is given an internal state, which is set according to the total copy number of chromosomes in the first or second nucleic acid sample. In an example, for diploid chromosomes, the internal states of the HMM can be homozygous deletion (locally zero copy), heterozygous deletion (locally one copy), normal (locally two copies), duplication (more than two copies), and a reference gap (existing as a state to distinguish gaps from homozygous deletions). In another case, for haploid chromosomes (e.g., X or Yin males), the internal states of the HMM can be homozygous deletion (locally zero copy), normal (locally two copies), duplication (more than two copies), and a reference gap (existing as a state to distinguish gaps from homozygous deletions). For example, for haploid chromosomes, there may be no available heterozygous deletion state. In another case, for trisomy and / or tetrasomy, additional intermediate states of the HMM state may have additional intermediate states, where these intermediate states can explain various CNV possibilities. In another implementation, the Hidden Markov Model is used to filter the output by examining the measurement insert size of readings near breakpoints of detected features.

[0229] Other models or algorithms useful in the methods of this invention include sequential search methods, genetic algorithms, distribution estimation algorithms, random forest algorithms, weight vectors of support vector machine algorithms, weights of logistic regression algorithms, etc. Bioinformatics. 1 Oct 2007; 23(19):2507-17 provides an overview of the relative advantages of the algorithms or models described above for data analysis. Illustrative algorithms include, but are not limited to, methods for reducing the number of variables, such as principal component analysis, partial least squares, independent component analysis, methods for directly handling large numbers of variables such as statistical methods and machine learning-based methods. Statistical methods include penalized logistic regression, microarray predictive analysis (PAM), centroid-based methods, support vector machine analysis, and regularized linear discriminant analysis.

[0230] In some implementations, HMM-based detection algorithms can detect large or fairly large CNVs in "segmentation." In some cases, small detection gaps may exist along the length of the true CNV due to fluctuations in the coverage signal. In one instance, a 1-megabase-pair (Mbp) deletion might be detected as a small number of individual nominal detections with small gaps between them. To mitigate this, a merging operation can be employed to identify adjacent detection pairs with gaps smaller than either of the two enclosing detections. The merging operation then measures the median coverage level within the gap. If the median coverage exceeds a predefined threshold, the two detections are merged into a single large detection spanning both original detections (including the enclosing detection gap). In this instance, the true feature spans both detections, and the gap is a statistical artifact. Using real sequencing data from samples known to have large CNVs, this merging operation can allow significantly better fidelity relative to the true properties of the CNV.

[0231] The methods and systems provided herein may also include the use of the feature selection algorithms provided herein. In some embodiments of the invention, feature selection is provided using the LIMMA software package (Smyth, GK (2005). Limma: linear models for microarray data. Bioinformatics and Computational Biology Solutions using Rand Bioconductor, R. Gentleman, V. Carey, S. Dudoit, R. Irizarry, W. Huber (eds.), Springer, New York, pp. 397-420).

[0232] In some embodiments of the invention, diagonal linear discriminant analysis, k-nearest neighbor algorithm, support vector machine (SVM) algorithm, linear support vector machine, random forest algorithm, or probabilistic model-based methods, or combinations thereof, are provided for detecting one or more genomic regions. In some embodiments, identified markers that distinguish samples (e.g., diseased vs. normal) or genomic regions (e.g., copy number variation vs. normal) are selected based on the statistical significance of differences in expression levels between target categories. In some cases, statistical significance is adjusted by applying Benjamini Hochberg or other false discovery rate (FDR) corrections.

[0233] In some cases, the algorithm may be supplemented with meta-analysis methods such as those described in Fishel and Kaufman et al. 2007 Bioinformatics 23(13):1599-606. In some cases, the algorithm may be supplemented with meta-analysis methods such as reproducibility analysis. In some cases, reproducibility analysis selects markers that appear in at least one group of predictive expression product markers.

[0234] Statistical evaluation of genomic region detection can provide one or more quantitative values ​​indicating one or more of the following: the probability of diagnostic accuracy; the probability of a condition, disease, or symptom; the probability of a specific condition, disease, or symptom; and the probability of success of a specific treatment intervention. Therefore, it is unlikely that physicians trained in genetics or molecular biology will not need to understand the raw data. Instead, the data are presented directly to physicians in the form of quantitative values ​​to guide patient care. Many methods known in the art can be used to statistically evaluate the results, including but not limited to: Student's t-test, two-sided t-test, Pearson rank-sum analysis, hidden Markov model analysis, qq graph analysis, principal component analysis, one-way ANOVA, two-way ANOVA, LIMMA, etc.

[0235] F. Illness or condition

[0236] Some implementation schemes may include predicting, diagnosing, and / or forecasting the state or outcome of a disease or condition in a subject based on one or more biomedical outputs. Predicting, diagnosing, and / or forecasting the state or outcome of a disease in a subject may include diagnosing a disease or condition, identifying a disease or condition, determining the stage of a disease or condition, assessing the risk of a disease or condition, assessing the risk of disease recurrence, assessing the efficacy of a drug, assessing the risk of adverse drug reactions, predicting the optimal drug dosage, predicting drug resistance, or a combination thereof.

[0237] The samples disclosed herein may be derived from subjects with cancer. Samples may include malignant tissue, benign tissue, or a mixture of both. The cancer may be recurrent and / or refractory. Examples of cancer include, but are not limited to, sarcoma, carcinoma, lymphoma, or leukemia. In some cases, samples containing cancerous tissue were obtained, but no matching normal samples were obtained. In some cases, no matching normal samples were available. In some cases, matching normal samples were obtained (e.g., for training and testing the models disclosed herein).

[0238] Sarcomas are cancers of bone, cartilage, fat, muscle, blood vessels, or other connective or supporting tissues. Sarcomas include, but are not limited to, bone cancer, fibrosarcoma, chondrosarcoma, Ewing's sarcoma, malignant angioendothelioma, malignant schwannoma, bilateral vestibular schwannoma, osteosarcoma, and soft tissue sarcomas (e.g., alveolar soft tissue sarcoma, angiosarcoma, phyllodes sarcoma, dermatofibrosarcoma, desmoid sarcoma, epithelioid sarcoma, extraskeletal osteosarcoma, fibrosarcoma, hemangiopericytoma, angiosarcoma, Kaposi's sarcoma, leiomyosarcoma, liposarcoma, lymphangiosarcoma, lymphosarcoma, malignant fibrous histiocytoma, neurofibrosarcoma, rhabdomyosarcoma, and synovial sarcoma).

[0239] Cancer originates from epithelial cells, which are the cells that cover the body's surface, produce hormones, and make up glands. By way of non-limiting examples, cancers include breast cancer, pancreatic cancer, lung cancer, colon cancer, colorectal cancer, rectal cancer, kidney cancer, bladder cancer, stomach cancer, prostate cancer, liver cancer, ovarian cancer, brain cancer, vaginal cancer, vulvar cancer, uterine cancer, oral cancer, penile cancer, testicular cancer, esophageal cancer, skin cancer, fallopian tube cancer, head and neck cancer, gastrointestinal stromal carcinoma, adenocarcinoma, melanoma of the skin or eye, anal cancer, small bowel cancer, endocrine system cancers, thyroid cancer, parathyroid cancer, adrenal cancer, urethral cancer, renal pelvis cancer, ureteral cancer, endometrial cancer, cervical cancer, pituitary adenocarcinoma, central nervous system (CNS) tumors, primary CNS lymphoma, brainstem glioma, and spinal cord axis tumors. Cancer can be skin cancer, such as basal cell carcinoma, squamous cell carcinoma, melanoma, non-melanoma, or actinic keratosis.

[0240] Cancer can be lung cancer. Lung cancer can begin from the airways that branch off from the trachea to supply the lungs (bronchioles) or the small air sacs (alveoli) of the lungs. Lung cancers include non-small cell lung cancer (NSCLC), small cell lung cancer, and mesothelioma. Examples of NSCLC include squamous cell carcinoma, adenocarcinoma, and large cell carcinoma. Mesothelioma can be a cancerous tumor of the endothelium of the lungs and the pleura or peritoneum of the abdomen. Mesothelioma can be caused by asbestos. Cancer can also be brain cancer, such as glioblastoma.

[0241] Cancer can be a tumor of the central nervous system (CNS). CNS tumors can be classified as gliomas or non-gliomas. Gliomas can be malignant gliomas, high-grade gliomas, or diffuse intrinsic pontine gliomas. Examples of gliomas include astrocytomas, oligodendrogliomas (or mixtures of oligodendroglioma and astrocytoma components), and ependymomas. Astrocytomas include, but are not limited to, low-grade astrocytomas, anaplastic astrocytomas, glioblastoma multiforme, pilocytic astrocytomas, xanthomorphic astrocytomas, and subependymal giant cell astrocytomas. Oligodendrogliomas include low-grade oligodendrogliomas (or oligodendroastrocytomas) and anaplastic oligodendrogliomas. Non-gliomas include meningiomas, pituitary adenomas, primary CNS lymphomas, and medulloblastomas. Cancer can be a meningioma.

[0242] Leukemia can be acute lymphoblastic leukemia, acute myeloid leukemia, chronic lymphoblastic leukemia, or chronic myeloid leukemia. Other types of leukemia include hairy cell leukemia, chronic myelomonocytic leukemia, and juvenile myelomonocytic leukemia.

[0243] Lymphoma is cancer of lymphocytes and can develop from B or T lymphocytes. The two main types of lymphoma are Hodgkin lymphoma (formerly known as Hodgkin's disease) and non-Hodgkin lymphoma. Hodgkin lymphoma is characterized by the presence of Reed-Sternberg cells. Non-Hodgkin lymphoma encompasses all lymphomas that do not involve Hodgkin's disease. Non-Hodgkin lymphoma can be indolent or aggressive. Non-Hodgkin's lymphomas include, but are not limited to, diffuse large B-cell lymphoma, follicular lymphoma, mucosa-associated lymphoid tissue lymphoma (MALT), small cell lymphoma, mantle cell lymphoma, Burkitt's lymphoma, mediastinal large B-cell lymphoma, Waldenstrom's macroglobulinemia, lymph node marginal zone B-cell lymphoma (NMZL), spleen marginal zone lymphoma (SMZL), extranodal marginal zone B-cell lymphoma, intravascular large B-cell lymphoma, primary exudative lymphoma, and lymphomatoid granuloma.

[0244] Some implementation schemes may include treatment and / or prevention of a disease or condition in a subject based on one or more biomedical outputs. One or more biomedical outputs may recommend one or more therapies. One or more biomedical outputs may suggest, select, specify, recommend, or otherwise determine a treatment and / or prevention process for a disease or condition. One or more biomedical outputs may suggest modification or continuation of one or more therapies. Modification of one or more therapies may include administering, initiating, reducing, increasing, and / or terminating one or more therapies. One or more therapies include anticancer, antiviral, antibacterial, antifungal, immunosuppressive therapies, or combinations thereof. One or more therapies may treat, alleviate, or prevent one or more diseases or indications.

[0245] Examples of anticancer therapies include, but are not limited to, surgery, chemotherapy, radiation therapy, immunotherapy / biological therapy, and photodynamic therapy. Anticancer therapies may include chemotherapy, monoclonal antibodies (e.g., rituximab, trastuzumab), cancer vaccines (e.g., therapeutic vaccines, prophylactic vaccines), gene therapy, or combinations thereof.

[0246] G. Systems, reagent kits, and libraries

[0247] The methods disclosed herein can be implemented via systems, kits, libraries, or combinations thereof. The methods of this invention may include one or more systems. The systems disclosed herein can be implemented via kits, libraries, or both. A system may include one or more components to perform any method or step of certain embodiments. For example, a system may include one or more kits, devices, libraries, or combinations thereof. A system may include one or more sequencers, processors, memory locations, computers, computer systems, or combinations thereof. A system may include a transport device.

[0248] The kit may include a variety of reagents for performing the various operations disclosed herein, including sample processing and / or analytical operations. The kit may include instructions for performing at least some of the operations disclosed herein. The kit may contain one or more capture probes, one or more beads, one or more tags, one or more adapters, one or more devices, one or more reagents, one or more buffers, one or more samples, one or more databases, or a combination of the above.

[0249] A library may contain one or more capture probes. A library may contain one or more subsets of nucleic acid molecules. A library may include one or more databases. A library may be generated or produced by any of the methods, kits, or systems disclosed herein. A database library may be generated from one or more databases. Methods for generating one or more libraries may include (a) aggregating information from one or more databases to generate an aggregated dataset; (b) analyzing the aggregated dataset; and (c) generating one or more database libraries from the aggregated dataset.

[0250] VI. Computing Environment

[0251] Figure 10An example of a computer system 1000 for implementing some embodiments disclosed herein is shown. The computer system 1000 may have a distributed architecture, where some components (e.g., memory and processor) are part of an end-user device, and some other similar components (e.g., memory and processor) are part of a computer server. The computer system 1000 includes at least a processor 1002, memory 1004, storage device 1006, input / output (I / O) peripherals 1008, communication peripherals 1010, and an interface bus 1012. The interface bus 1012 is configured to communicate, transmit, and transfer data, control, and commands among the various components of the computer system 1000. The processor 1002 may include one or more processing units, such as a CPU, GPU, TPU, systolic array, or SIMD processor. Memory 1004 and storage device 1006 include computer-readable storage media such as RAM, ROM, electrically erasable programmable read-only memory (EEPROM), hard disk drive, CD-ROM, optical storage device, magnetic storage device, electronic non-volatile computer storage, etc. Memory and other tangible storage media. Any such computer-readable storage medium may be configured to store instructions or program code embodying aspects of this disclosure. Memory 1004 and storage device 1006 also include computer-readable signal media. Computer-readable signal media include propagated data signals containing computer-readable program code. Such propagated signals may take any of a variety of forms, including but not limited to electromagnetic, optical, or any combination thereof. Computer-readable signal media include any computer-readable medium that is not a computer-readable storage medium but may communicate, propagate, or transmit a program for use in conjunction with computer system 1000.

[0252] In addition, memory 1004 includes an operating system, programs, and application programs. Processor 1002 is configured to execute stored instructions and includes, for example, a logic processing unit, a microprocessor, a digital signal processor, and other processors. Memory 1004 and / or processor 1002 can be virtualized and can be hosted in another computing system, such as a cloud network or data center. I / O peripherals 1008 include user interfaces such as a keyboard, a screen (e.g., a touchscreen), a microphone, a speaker, other input / output devices, and computing components such as a graphics processing unit, serial ports, parallel ports, a universal serial bus, and other input / output peripherals. I / O peripherals 1008 are connected to processor 1002 via any port coupled to interface bus 1012. Communication peripherals 1010 are configured to facilitate communication between computer system 1000 and other computing devices via a communication network and include, for example, a network interface controller, a modem, wireless and wired interface cards, an antenna, and other communication peripherals.

[0253] Although this subject matter has been described in detail with respect to specific embodiments thereof, it should be understood that those skilled in the art, upon gaining an understanding of the foregoing, can readily produce changes, variations, and equivalents to these embodiments. Therefore, it should be understood that this disclosure is presented for illustrative purposes rather than limiting, and does not exclude the inclusion of such modifications, variations, and / or additions to the subject matter that will be apparent to those skilled in the art. In fact, the methods and systems described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions, and changes may be made to the form of the methods and systems described herein without departing from the spirit of this disclosure. The appended claims and their equivalents are intended to cover such forms or modifications that fall within the scope and spirit of this disclosure.

[0254] Unless otherwise expressly stated, it should be understood that the discussion throughout this specification uses terms such as “processing,” “computing,” “calculating,” “determining,” and “identifying” to refer to the actions or processes of computing devices, such as one or more computers or similar one or more electronic computing devices, which manipulate or convert data represented as physical electronic or magnetic quantities in the memory, registers, or other information storage, transmission, or display devices of a computing platform.

[0255] The one or more systems discussed herein are not limited to any particular hardware architecture or configuration. A computing device may include any suitable arrangement of components that provides a result conditioned on one or more inputs. Suitable computing devices include microprocessor-based multipurpose computing systems that access stored software, ranging from general-purpose computing devices to dedicated computing devices that implement one or more embodiments of this subject matter, to program or configure the computing system. Any suitable programming, scripting, or other type of language or combination of languages ​​may be used to implement the teachings contained herein in software used for programming or configuring the computing device.

[0256] Implementations of the methods disclosed herein can be executed in the operation of such computing devices. The order of the boxes presented in the above examples can be changed—for example, the boxes can be reordered, combined, and / or decomposed into sub-boxes. Some boxes or processes can be executed in parallel.

[0257] The conditional language used in this document, such as “can,” “could,” “might,” “may,” “eg,” etc., unless otherwise explicitly stated or otherwise understood in the context in which they are used, is generally intended to convey that certain instances include, while other instances do not include, certain features, elements, and / or steps. Therefore, such conditional language is not generally intended to imply that one or more instances require features, elements, and / or steps in any way, or that one or more instances must include logic for determining whether such features, elements, and / or steps are included or will be performed in any particular instance, with or without author input or prompting.

[0258] The terms “comprising,” “including,” “having,” etc., are synonyms and are used inclusively in an open-ended manner, not excluding additional elements, features, actions, operations, etc. Furthermore, the term “or” is used in its inclusive meaning (and not its exclusionary meaning), such that, for example, when used to connect lists of elements, the term “or” means one, some, or all of the elements in the list. The terms “applies to” or “configured as” as used herein are open-ended and inclusive language, not excluding means applicable to or configured to perform additional tasks or steps. Additionally, the use of “based on” implies openness and inclusion, because a process, step, calculation, or other action “based on” one or more listed conditions or values ​​may actually be based on other conditions or values ​​beyond the listed values. Similarly, the use of “at least partially based on” implies openness and inclusion, because a process, step, calculation, or other action “at least partially based on” one or more listed conditions or values ​​may actually be based on other conditions or values ​​beyond the listed values. The headings, lists, and numbering included herein are for ease of interpretation only and are not intended to be restrictive.

[0259] The various features and processes described above can be used independently of each other or combined in various ways. All possible combinations and sub-combinations are intended to fall within the scope of this disclosure. Additionally, certain method or process blocks may be omitted in some embodiments. The methods and processes described herein are not limited to any particular order, and the blocks or states associated with them may be performed in other suitable orders. For example, the described blocks or states may be performed in an order different from that specifically disclosed, or multiple blocks or states may be combined in a single block or state. Exemplary blocks or states may be performed serially, in parallel, or in some other manner. Blocks or states may be added to or removed from the disclosed instances. Similarly, the exemplary systems and components described herein may be configured differently from those described. For example, elements may be added, removed, or rearranged compared to the disclosed instances.

Claims

1. A system for determining tumor purity, comprising one or more data processors and a non-transitory computer-readable storage medium containing instructions that, when executed on the one or more data processors, cause the one or more data processors to perform the following steps: Obtain nucleic acid sequence data of multiple nucleic acid molecules from a biological sample representing the target; The nucleic acid sequence data was compared with a reference genome. A set of genomic regions is identified based on aligned nucleic acid sequence data, wherein each genomic region in the set of genomic regions includes one or more nucleotide sequence variations relative to a corresponding genomic region in the reference genome; Determine the frequency of the B allele in each of the aforementioned genomic regions; Based on the B allele frequencies of the aforementioned set of genomic regions, the distribution of B allele frequencies in the biological sample was determined. The frequency distribution of the B allele is processed using a trained machine learning model to estimate a measure of tumor purity in the biological sample, wherein the trained machine learning model includes a convolutional neural network trained on a training dataset generated from nucleic acid sequence data of one or more tumor cells diluted to normal cells. and Output the metric.

2. The system of claim 1, wherein the nucleic acid sequence data is whole exome sequencing data.

3. The system of claim 1, wherein the nucleic acid sequence data is whole genome sequencing data.

4. The system of claim 1, further comprising: The biological sample was obtained from the object; and Multiple nucleic acid molecules from the biological sample are sequenced to generate the nucleic acid sequence data.

5. The system of claim 4, further comprising isolating the plurality of nucleic acid molecules prior to sequencing.

6. The system of claim 1, wherein identifying the genome region further comprises: Identify one or more candidate nucleotide sequence variations in the nucleic acid sequence data; and Calculate the reference and substitution read depths for each of the one or more candidate nucleotide sequence variations.

7. The system of claim 1, wherein the B allele frequency distribution is normalized.

8. The system of claim 1, wherein the trained machine learning model has a mean absolute error of less than about 0.

2.

9. The system of claim 1, further comprising outputting a report including information identifying the frequency distribution of the B allele.

10. The system of claim 1, further comprising outputting a report including a predicted metric for identifying the purity of the tumor.

11. The system of claim 10, wherein the report further includes information identifying at least one biomarker.

12. The system of claim 10, wherein the report further includes information identifying at least one prognostic marker.

13. The system of claim 10, wherein the report includes information identifying predicted somatic variations.

14. The system of claim 1, wherein the biological sample is derived from a human subject.

15. A computer program product comprising instructions configured to cause one or more data processors to perform the following steps: Obtain nucleic acid sequence data of multiple nucleic acid molecules from a biological sample representing the target; The nucleic acid sequence data was compared with a reference genome. A set of genomic regions is identified based on aligned nucleic acid sequence data, wherein each genomic region in the set of genomic regions includes one or more nucleotide sequence variations relative to a corresponding genomic region in the reference genome; Determine the frequency of the B allele in each of the aforementioned genomic regions; Based on the B allele frequencies of the aforementioned set of genomic regions, the distribution of B allele frequencies in the biological sample was determined. The frequency distribution of the B allele is processed using a trained machine learning model to estimate a measure of tumor purity in the biological sample, wherein the trained machine learning model includes a convolutional neural network trained on a training dataset generated from nucleic acid sequence data of one or more tumor cells diluted to normal cells. and Output the metric.

16. The computer program product of claim 15, wherein the nucleic acid sequence data is whole exome sequencing data.

17. The computer program product of claim 15, wherein the nucleic acid sequence data is whole genome sequencing data.

18. The computer program product of claim 15, wherein identifying the genome region further comprises: Identify one or more candidate nucleotide sequence variations in the nucleic acid sequence data; and Calculate the reference and substitution read depths for each of the one or more candidate nucleotide sequence variations.

19. The computer program product of claim 15, wherein the B allele frequency distribution is normalized.

20. The computer program product of claim 15, wherein the trained machine learning model has a mean absolute error of less than about 0.

2.

21. The computer program product of claim 15, further comprising instructions configured to cause one or more data processors to perform the following steps: The output includes a report that identifies the frequency distribution of the B allele.

22. The computer program product of claim 15, further comprising instructions configured to cause one or more data processors to perform the following steps: The output includes a report of a predicted metric for identifying the purity of the tumor.

23. The computer program product of claim 22, wherein the report further includes information identifying at least one biomarker.

24. The computer program product of claim 22, wherein the report further includes information identifying at least one prognostic marker.

25. The computer program product of claim 22, wherein the report includes information identifying predicted somatic variations.

26. The computer program product of claim 15, wherein the biological sample is derived from a human subject.

Images