Enhanced detection of target DNA by fragment size analysis

Abstract

The present invention provides a computer-implemented method for detecting variant nucleic acids from a sample containing cell-free nucleic acids. The method comprises (a) providing data representing the fragment size of nucleic acid fragments obtained from the sample and / or data representing a measure of the deviation from the copy number neutral point in the nucleic acid fragments obtained from the sample; b) processing the data from step a) according to a classification algorithm, wherein based on multiple cell-free nucleic acid fragment size characteristics and / or deviation characteristics from the copy number neutral point, the operation of the classification algorithm classifies the sample data into at least one of the following: a first category containing the variant nucleic acid and a second category not containing the variant nucleic acid; and c) outputting the classification of the sample from step b, thereby determining whether the sample contains the variant nucleic acid, or the probability that the sample contains the variant nucleic acid. Related methods are also provided.

Description

Technical Field

[0001] The present invention relates in part to methods for detecting the presence of target DNA, such as circulating tumor DNA (ctDNA), from, for example, a cell-free DNA (cfDNA) source (e.g., plasma or other biological fluids). In particular, the methods of the present invention can be used for diagnosis, treatment, and in particular monitoring of cancer. Background Art

[0002] The plasma of cancer patients contains circulating tumor DNA (ctDNA), but this valuable source of information is diluted by the much larger amount of non-cancer DNA: thus, ctDNA accounts for only a small fraction of total cell-free DNA (cfDNA) (1, 2). High-depth targeted sequencing of selected genomic regions can be used to detect low levels of ctDNA, but more extensive analyses using methods such as whole exome sequencing (WES) and shallow whole genome sequencing (sWGS) are typically only informative when ctDNA levels are approximately 10% or greater (3-5). ctDNA concentrations in patients with advanced cancer can exceed 10% of total cfDNA (6-8), but are much lower in patients with low tumor burden (9-12) and in patients with some cancer types such as glioma and renal cancer (6). Current strategies to improve ctDNA detection rely on increasing sequencing depth in combination with multiple error correction methods (2, 13, 14). However, methods that focus solely on mutation analysis do not exploit potential differences in chromatin organization or fragment size in ctDNA (15-17). The results of increasingly in-depth sequencing are also confounded by the possibility of false-positive results in the detection of non-cancerous cell mutations or clonal expansion of normal epithelial cells or clonal hematopoiesis of indeterminate potential (CHIP) (13, 18, 19).

[0003] Source cells and the mechanism of cfDNA release into the blood can mark cfDNA with specific fragmentation features, potentially providing accurate information about cell type, gene expression, carcinogenic potential or therapeutic effects (15, 16, 20). cfDNA fragments usually show a prominent pattern at 167bp, indicating release from apoptotic caspase-dependent cleavage (21-24). Circulating fetal DNA has been shown to be shorter than maternal DNA in plasma, and these size differences have been used to improve the sensitivity of non-invasive prenatal diagnosis (22, 25-27). The size distribution of tumor-derived cfDNA has only been explored in a few studies covering a small number of cancer types and patients, and conflicting results have been shown (28-33). The limitation of previous studies is that determining the specific size of tumor-derived DNA fragments requires detailed characterization of matching tumor-derived changes (30, 33), and a broader understanding and impact of potential biological differences have not been explored previously. Mouliere, Pikorz, Chandrananda, Moore et al., 2017, BioRxiv Preprint, doi: http: / / dx.doi.org / 10.1101 / 134437 reported that selection of short fragments in plasma improved the detection of circulating tumor DNA (ctDNA) in patients with recurrent high-grade serous ovarian cancer.

[0004] Although ctDNA detection has shown promise in the field of cancer care, there remains an unmet need for methods and systems that maximize the signal-to-noise ratio in the context of ctDNA detection. A related problem is the need to distinguish somatic cancer mutations from mutations present in non-cancerous cells, clonal expansion of normal epithelial cells, or CHIP. The present invention seeks to provide solutions to these needs and provides further related advantages. Summary of the invention

[0005] The inventors hypothesize that differences in the length of circulating DNA fragments can be exploited to enhance the sensitivity of non-invasive genomic analysis for detecting the presence of ctDNA and cancer. As described in detail herein, analysis of size-selected cfDNA identified clinically actionable mutations and copy number changes that would otherwise be undetectable. The identification of patients with advanced cancer was improved by such a predictive model: it integrated fragment length and copy number analysis of cfDNA with an AUC>0.99 compared to AUC<0.80 without fragment features. Compared to AUC<0.5 without fragment features, improved detection of ctDNA from patients with glioma, renal cancer, and pancreatic cancer was achieved at AUC>0.91. Detection of ctDNA from gliomas that do not metastasize outside the central nervous system (CNS) has previously been reported to be very challenging (6). Fragment size analysis and selective sequencing of specific fragment sizes can facilitate ctDNA detection and can be an alternative to more in-depth mutation sequencing for clinical applications, early diagnosis, and research on tumor biology.

[0006] Thus, in a first aspect, the present invention provides a computer-implemented method for detecting variant nucleic acids (eg, DNA or RNA) from a sample containing cell-free nucleic acid (eg, DNA or RNA), comprising:

[0007] a) providing data representing the fragment size of nucleic acid fragments obtained from said sample and / or data representing a measure of the deviation from copy number neutrality in nucleic acid fragments obtained from said sample;

[0008] b) causing a processor of a computer to process the data from step a) according to a classification algorithm, the classification algorithm having been trained on a training set comprising a plurality of cell-free nucleic acid samples comprising variant nucleic acids and a plurality of samples not comprising variant nucleic acids, wherein execution of the classification algorithm classifies the sample data into one of at least two classes based on a plurality of cell-free nucleic acid fragment size features and / or deviations from copy number neutrality features, the at least two classes comprising a first class comprising variant nucleic acids and a second class not comprising variant nucleic acids; and

[0009] c) outputting the classification of the sample from step b) and thereby determining whether the sample contains a variant nucleic acid, or determining the probability that the sample contains a variant nucleic acid.

[0010] In some embodiments, the sample containing cell-free nucleic acid is a cell-free DNA (cfDNA) sample, and wherein the variant nucleic acid is a variant DNA. In particular, the variant DNA can be selected from: circulating tumor DNA (ctDNA), circulating bacterial DNA, circulating pathogen DNA, circulating mitochondrial DNA, circulating fetal DNA, circulating DNA derived from donor organs or donor tissues, circulating DNA released by physiologically altered tissues or cells, circulating extrachromosomal DNA, and double minute circular DNA. In a particularly preferred embodiment, the variant DNA is ctDNA.

[0011] In some embodiments, data representing fragment sizes of nucleic acid fragments (eg, DNA or RNA fragments) comprises fragment sizes inferred from sequence reads, fragment sizes determined by fluorescence, or fragment sizes determined by densitometry.

[0012] In some embodiments, the present invention provides a computer-implemented method for detecting variant DNA from a sample containing cell-free DNA (cfDNA), comprising:

[0013] a) providing sequence data representing the fragment size of cfDNA fragments obtained from the sample and / or sequence data representing a measure of deviation from copy number neutrality in cfDNA fragments obtained from the sample;

[0014] b) causing a processor of a computer to process the sequence data from step a) according to a classification algorithm, the classification algorithm having been trained on a training set comprising a plurality of cfDNA samples containing variant DNA and a plurality of samples not containing variant DNA, wherein running of the classification algorithm classifies the sample data into one of at least two categories based on a plurality of cfDNA fragment size features and / or deviation features from a copy number neutral point, the at least two categories comprising a first category containing variant DNA and a second category not containing variant DNA; and

[0015] C) output the classification of the sample from step b), and thereby determine whether the sample contains variant DNA, or determine the probability that the sample contains variant DNA. As described in the embodiments herein, the classification algorithm can learn from the analysis of cfDNA fragmentation characteristics and somatic copy number alteration (SCNA), and improve the detection of ctDNA with relatively low cost and shallow sequencing methods. In addition, cfDNA fragmentation characteristics and / or SCNA analysis can be used to classify cancer and healthy samples with high accuracy.

[0016] In some embodiments, the classification algorithm is run to classify the sample data into one of the at least two, three, four, or at least five categories based at least on a plurality of cfDNA fragment size features selected from:

[0017] (i) the proportion of fragments with a size range of 20 to 150 bp (P20 to 150);

[0018] (ii) the proportion of fragments in the size range of 100 to 150 bp (P100 to 150);

[0019] (iii) the proportion of fragments with a size range of 160 to 180 bp (P160 to 180);

[0020] (iv) proportion of fragments in the size range of 180 to 220 bp (P180 to 220);

[0021] (v) proportion of fragments with a size range of 250 to 320 bp (P250 to 320);

[0022] (vi) the ratio of P(20 to 150) / P(160 to 180);

[0023] (vii) the ratio of the ratio P (100 to 150) divided by the ratio of fragments in the size range of 163 to 169 bp;

[0024] (viii) the ratio of the ratio P(20 to 150) / P(180 to 220); and

[0025] (ix) Amplitude oscillations with a 10 bp periodicity in fragment size density It should be understood that the sequence data representing the fragment sizes of cfDNA fragments in step a) comprises the cfDNA fragment size features used by the classification algorithm.

[0026] In some embodiments, multiple cfDNA fragment size features include: P (160 to 180), P (180 to 220), P (250 to 320), and amplitude oscillations with a 10 bp periodicity in fragment size density. As described in the Examples herein, both linear and nonlinear machine learning algorithms independently identified the same four fragment size features P (160 to 180), P (180 to 220), P (250 to 320), and amplitude oscillations with a 10 bp periodicity in fragment size density, as well as SCNA features (i.e., trimmed median absolute deviation (t-MAD) scores from copy number neutrality), although there were some differences in the rank order of the features. Highly accurate classification was obtained using only four fragmentation features (see Fig.26 ).

[0027] In some embodiments, sample data is classified into one of the at least two categories based at least on the execution of a deviation from copy number neutrality feature classification algorithm, wherein the deviation from copy number neutrality is a trimmed median absolute deviation (t-MAD) score from copy number neutrality or an ichorCNA feature.

[0028] ichorCNA is a tool for estimating tumor fractions in cell-free DNA from ultra-low-pass whole genome sequencing (ULP-WGS, 0.1× coverage). The code for ichorCNA is available at the following URL: https: / / github.com / broadinstitute / ichorCNA. ichorCNA uses a probabilistic model implemented as a hidden Markov model (HMM) to simultaneously segment the genome, predict large-scale copy number changes, and estimate tumor fractions for ultra-low-pass whole genome sequencing (ULP-WGS) samples. The methodology and probabilistic model are described in: Adalsteinsson, Ha, Freeman, et al. Scalable whole-exome sequencing of cell-free DNA reveals high concordance with metastatic tumors. (2017) Nature Communications Nov 6; 8(1): 1324. doi: 10.1038 / s41467-017-00965-y (the contents of which are incorporated herein by reference). The analysis workflow consists of 2 tasks:

[0029] GC content bias correction (using HMMcopy)

[0030] a. Calculate read coverage from ULP-WGS

[0031] b. Data Correction and Normalization

[0032] CNA prediction and estimation of cfDNA tumor fraction.

[0033] In particular, when the deviation feature from copy number neutrality comprises a t-MAD score, the score can be determined by trimming genomic regions that show high copy number variability in a whole genome dataset derived from healthy subjects, and then calculating the median absolute deviation from log2R=0 in non-trimmed regions of the genome.

[0034] In some embodiments according to the present invention, the classification algorithm performs random forests (RF) analysis, logistic regression (LR) analysis or support vector machine (SVM) analysis. The classification algorithm can provide an output as a correct classification probability (e.g., the probability that the sample in question has been correctly classified as a healthy category or a cancer category according to the training set on which the classification algorithm has been trained).

[0035] In some embodiments, the performance of the classification algorithm when trained on the training set is evaluated by the area under the curve (AUC) value of the receiver operating characteristic (ROC) analysis. The classification algorithm model showing the highest AUC value is usually selected as the classification algorithm model with the best performance.

[0036] In some embodiments, the classification algorithm has been trained on a training set comprising at least 10, 20, 30, 40 or at least 50 samples from healthy subjects and at least 10, 20, 30, 40 or at least 50 samples from subjects known to have cancer. In particular, the samples used in the training set can be those shown in Table 2.

[0037] In some embodiments, the sequence data provided in step a) represents whole genome sequence (WGS) reads, Tailored Panel Sequencing (TAPAS) sequence reads, Integration of Variant Read (INVAR) TAPAS (see co-pending patent application GB1803596.4 filed on March 6, 2018, which is incorporated herein by reference), hybrid capture sequence reads, tagged amplicon deep sequencing (TAm-Seq) reads, focused exome sequence reads, or whole exome sequence reads. In particular, the sequence data provided in step a) may represent shallow whole-genome sequence (sWGS) reads, optionally 0.4× depth WGS reads.

[0038] In some embodiments, the data provided in step a) represents fragment sizes of a plurality of nucleic acid fragments (eg, DNA fragments) from a substantially cell-free fluid sample from a subject having or suspected of having cancer.

[0039] In some embodiments, the sequence data provided in step a) represents sequence reads of a plurality of DNA fragments from a substantially cell-free liquid sample from a subject having or suspected of having cancer.

[0040] In some embodiments, the cancer can be selected from melanoma, lung cancer, bile duct cancer, bladder cancer, esophageal cancer, colorectal cancer, ovarian cancer, glioma, pancreatic cancer, renal cancer, and breast cancer.

[0041] In some embodiments, the sample is a plasma sample, a urine sample, a saliva sample, a cerebrospinal fluid sample, a serum sample, or other nucleic acid-containing (eg, DNA-containing) biological fluid sample.

[0042] In some embodiments, wherein the variant DNA is ctDNA, the method is used to detect the presence, growth, prognosis, regression, treatment response, or recurrence of cancer in the subject from which the sample was obtained.

[0043] In some embodiments, the presence of ctDNA in the sample is distinguished from cfDNA containing somatic mutations of non-cancer origin. It is particularly contemplated herein that the inclusion of fragment size information on each read can enhance mutation calling algorithms from high-depth sequencing in order to distinguish tumor-derived mutations from somatic variants of other origins (including clonal expansion of non-cancerous cells) or background sequencing noise. In certain embodiments, the method can distinguish variant sequence reads representing clonal expansion of normal epithelial cells or clonal hematopoiesis of uncertain potential (CHIP) from variant sequence reads representing ctDNA.

[0044] In certain embodiments, the fragment size data provided in step a) represents sequence reads of a plurality of DNA fragments from a substantially cell-free fluid sample of a subject, and wherein the method is used to determine whether the sample comprises ctDNA or cfDNA from CHIP. In particular, the classification algorithm may have been trained on a training set that also comprises a plurality of cfDNA samples obtained from subjects with CHIP, and wherein based on a plurality of cfDNA fragment size features and / or features of deviations from copy number neutrality, the at least two categories further comprise a third category containing cfDNA derived from CHIP.

[0045] In a second aspect, the invention provides a method for detecting variant nucleic acids from a sample containing cell-free nucleic acid, comprising:

[0046] analyzing a sample containing cell-free nucleic acid or a library derived from a sample containing cell-free nucleic acid to determine fragment sizes of nucleic acid fragments in the sample or the library, wherein the sample is obtained from a subject; and

[0047] The method of the first aspect of the invention is performed using fragment size.

[0048] In some embodiments, the analyzing comprises:

[0049] sequencing nucleic acids from a nucleic acid-containing sample or library and inferring fragment sizes from sequence reads;

[0050] Measuring the fragment sizes of nucleic acids from a nucleic acid-containing sample or library by fluorescence; and / or

[0051] The fragment sizes of nucleic acids from nucleic acid-containing samples or libraries are measured by optical density.

[0052] In some embodiments, the present invention provides a method for detecting variant DNA from a sample containing cell-free DNA (cfDNA), comprising:

[0053] sequencing a cfDNA-containing sample obtained from a subject or a library derived from a cfDNA-containing sample to obtain a plurality of sequence reads;

[0054] Processing the sequence reads to determine sequence data representing fragment sizes of cfDNA fragments obtained from the sample and / or sequence data representing a measure of deviation from copy number neutrality in cfDNA fragments obtained from the sample; and

[0055] The method of the first aspect of the invention is performed using sequence data.

[0056] In some embodiments, sequencing includes generating a sequencing library from a sample and performing whole genome sequencing, panel-on-demand sequencing (TAPAS) sequencing, hybridization capture sequencing, TAm-Seq sequencing, focused exome sequencing, or whole exome sequencing, optionally generating an indexed sequencing library and performing shallow whole genome sequencing (e.g., a depth of 0.4×).

[0057] In some embodiments, processing the sequence reads includes one or more of the following steps:

[0058] Aligning the sequence reads to a reference genome of the same species as the subject (e.g., human reference genome GRCh37 for human subjects);

[0059] Removal of contaminating adapter sequences;

[0060] PCR and optical duplicates were removed;

[0061] removing sequence reads of low mapping quality; and

[0062] In case of multiplexed sequencing, the signals are de-multiplexed by excluding mismatches in the sequencing barcodes.

[0063] In some embodiments, the variant DNA is selected from: circulating tumor DNA (ctDNA), circulating bacterial DNA, circulating pathogen DNA, circulating mitochondrial DNA, circulating fetal DNA and circulating DNA derived from donor organs or donor tissues, circulating DNA released by physiologically altered tissues or cells, circulating extrachromosomal DNA, and double minute circular DNA.

[0064] In some embodiments, processing the sequence reads to determine sequence data representing fragment sizes of cfDNA fragments obtained from the sample and / or sequence data representing a measure of deviation from copy number neutrality in cfDNA fragments obtained from the sample comprises determining one or more (e.g., 2, 3, 4, 5 or more) features selected from:

[0065] (i) the proportion of fragments with a size range of 20 to 150 bp (P20 to 150);

[0066] (ii) the proportion of fragments in the size range of 100 to 150 bp (P100 to 150);

[0067] (iii) the proportion of fragments with a size range of 160 to 180 bp (P160 to 180);

[0068] (iv) proportion of fragments in the size range of 180 to 220 bp (P180 to 220);

[0069] (v) proportion of fragments with a size range of 250 to 320 bp (P250 to 320);

[0070] (vi) the ratio of P(20 to 150) / P(160 to 180);

[0071] (vii) the ratio of the ratio P (100 to 150) divided by the ratio of fragments in the size range of 163 to 169 bp;

[0072] (viii) the ratio of the ratio P(20 to 150) / P(180 to 220); and

[0073] (ix) Amplitude oscillations in fragment size density with a periodicity of 10 bp.

[0074] In some embodiments, the plurality of cfDNA fragment size characteristics include: P(160-180), P(180-220), P(250-320), and amplitude oscillations with a 10 bp periodicity in fragment size density.

[0075] In some embodiments, the fragment sizes of the cfDNA fragments are inferred from the sequence reads using the mapped positions of the ends of the reads in the genome after the sequence reads are aligned to a reference genome of the species from which the sample was obtained.

[0076] In some embodiments, processing the sequence reads to determine sequence data representing a measure of deviation from copy number neutrality in cfDNA fragments obtained from the sample comprises determining a trimmed median absolute deviation (t-MAD) score or an ichorCNA score from copy number neutrality. In particular, the t-MAD score can be determined by trimming genomic regions that show high copy number variability in a whole genome dataset derived from a healthy subject, and then calculating the median absolute deviation from log2R=0 in non-trimmed regions of the genome.

[0077] In some embodiments, the sample comprises a plurality of DNA fragments from a substantially cell-free fluid of a subject having or suspected of having cancer. In some specific instances, the cancer may be selected from melanoma, lung cancer, bile duct cancer, bladder cancer, esophageal cancer, colorectal cancer, ovarian cancer, glioma, pancreatic cancer, kidney cancer, and breast cancer.

[0078] In some embodiments, the sample is a plasma sample, a urine sample, a saliva sample, a cerebrospinal fluid sample, a serum sample, or other DNA-containing biological fluid sample.

[0079] According to any aspect of the invention, the sample may be or may have been subjected to one or more processing steps to remove intact cells, for example by centrifugation.

[0080] In certain embodiments, where the variant DNA is ctDNA, the method can be used to detect the presence, growth, prognosis, regression, treatment response, or recurrence of cancer in the subject from which the sample was obtained.

[0081] In some embodiments, the presence of ctDNA is distinguished from the presence of cfDNA containing somatic mutations of non-cancerous origin, optionally CHIP origin.

[0082] In some embodiments, somatic mutations comprising cfDNA fragments are classified as tumor-derived or CHIP-derived based on multiple fragment size features determined from sequence reads.

[0083] In some embodiments, the variant DNA is ctDNA and classifying a sample as containing ctDNA or not containing ctDNA, or determining the probability that a sample contains ctDNA, is used to predict whether the sample or another sample from the same subject will be susceptible to further ctDNA analysis.

[0084] In some cases, further ctDNA analysis includes sequencing at a greater sequencing depth and / or performing targeted sequencing of the ctDNA in the sample.

[0085] In some embodiments, the sample is subjected to the further ctDNA analysis when the probability that the sample comprises ctDNA as determined by the classification algorithm is at least 0.5 (eg, at least 0.6 or at least 0.75).

[0086] In some embodiments:

[0087] The sample is a plasma sample, and the probability that the sample contains ctDNA as determined by the classification algorithm is used to determine whether ctDNA will be detectable in the urine sample; or

[0088] The sample is a urine sample, and wherein the probability that the sample contains ctDNA as determined by the classification algorithm is used to determine whether ctDNA will be detectable in the plasma sample. As shown in Example 8, a relatively high probability that the plasma sample contains ctDNA as shown by the classification algorithm is associated with an increased probability that a usable ctDNA test can be performed with a urine sample (see also Fig. 27 ).

[0089] In a third aspect, the invention provides a method for improving the detection of circulating tumor DNA (ctDNA) in a cell-free DNA (cfDNA) sample, comprising performing in vitro and / or on-board size selection to enrich DNA fragments less than 167bp in length and / or DNA fragments with a size range of 250 to 320bp. In some embodiments, size selection is for enriching DNA fragments with a length in the range of 90 to 150bp. In some cases, size selection may include excluding high molecular weight DNA, such as those derived from leukocytes when the sample comprises a serum sample.

[0090] In some embodiments, the sample may be obtained from a subject having or suspected of having a cancer selected from the group consisting of melanoma, bile duct cancer, colorectal cancer, glioma, pancreatic cancer, renal cancer, and breast cancer.

[0091] In some embodiments, size selection comprises in vitro size selection performed on DNA extracted from a cfDNA-containing sample and / or in vitro size selection performed on a library created from DNA extracted from a cfDNA-containing sample. In particular, in vitro size selection may comprise agarose gel electrophoresis.

[0092] In some embodiments, size selection comprises on-board size selection of the sequence reads.

[0093] In some specific cases, the sequence reads may include paired-end reads generated by sequencing DNA from both ends of fragments present in a library generated from a cfDNA-containing sample. After aligning the sequence reads to a reference genome of the species from which the sample was obtained (e.g., the human reference genome GRCh37 of a human subject), the mapped positions of the ends of the reads in the genome can be used to infer the original lengths of the DNA fragments in the cfDNA-containing sample.

[0094] In some embodiments, DNA fragments outside the range of 90 to 150 bp in length are substantially excluded (see, e.g. Figure 6 B).

[0095] In some embodiments, size selection is performed on a genome-wide basis or an exome-wide basis. As described herein, in contrast to previous studies limited to specific genomic loci, cancer types, or cases (30, 32, 33), the inventors identified size differences between mutant cfDNA and non-mutant cfDNA on a genome-wide and pan-cancer scale.

[0096] In certain embodiments, in vitro size selection is performed prior to shallow whole genome sequencing (sWGS), or sWGS sequencing reads are size selected on-board.

[0097] In certain embodiments, the method further comprises performing somatic copy number distortion analysis and / or mutation calling on the sequence reads after size selection. In some specific cases, the somatic copy number distortion analysis may include processing the sequence reads to determine a trimmed median absolute deviation (t-MAD) score or ichorCNA score from a copy number neutral point. For example, the t-MAD score may be determined by trimming a genomic region showing high copy number variability in a whole genome data set derived from a healthy subject, and then calculating the median absolute deviation from log2R=0 in a non-trimmed region of the genome.

[0098] In certain embodiments, the analysis of somatic copy number aberrations may include detecting amplification of one or more genes selected from NF1, TERT, and MYC.As described in the Examples herein, analysis of plasma cfDNA after size selection revealed a large number of SCNAs not observed in the same sample without size selection.

[0099] In certain embodiments, mutation calling comprises detecting mutations in one or more genes selected from BRAF, ARID1A, and NF1. As described in the Examples herein, size selection enriches for mutant allele fractions (MAF) of nearly all mutations.

[0100] In some embodiments, the cancer is a high ctDNA cancer selected from colorectal cancer, bile duct cancer, breast cancer, and melanoma.

[0101] In some embodiments, the cancer is a low ctDNA cancer selected from pancreatic cancer, renal cancer, and glioma.

[0102] In certain embodiments, the sample can be a plasma sample, a urine sample, a saliva sample, a cerebrospinal fluid sample, a serum sample, or other DNA-containing biological fluid sample.

[0103] In some embodiments, the method also includes detecting the presence, growth, prognosis, regression, treatment response or recurrence of cancer in the subject from which the sample is obtained. Improving the detection of ctDNA, mutation calling and / or SCNA detection according to the method of this aspect of the invention can contribute to the early detection of cancer and continuous cancer monitoring, and can inform treatment strategies.

[0104] In some embodiments, the method can be performed on a sample obtained before cancer treatment of a subject and a sample obtained after cancer treatment of a subject. As described herein, the size selected samples indicated tumor progression 69 and 87 days prior to detection by imaging or t-MAD analysis without size selection (see Fig.10 E and F).

[0105] According to any aspect of the invention, the subject can be a human, a companion animal (e.g., a dog or cat), a laboratory animal (e.g., a mouse, a rat, a rabbit, a pig, or a non-human primate), a livestock or a farm animal (e.g., a pig, a cow, a horse, or a sheep). Preferably, the subject is a human patient. In some cases, the subject is a human patient who has been diagnosed with cancer, is suspected of having cancer, or has been classified as being at risk of developing cancer.

[0106] Embodiments of the invention will now be described by way of example and not limitation with reference to the accompanying drawings. However, various further aspects and embodiments of the invention will be apparent to those skilled in the art in view of this disclosure.

[0107] The invention includes combinations of the described aspects and preferred features, unless such combinations are clearly not permitted or indicated to be explicitly avoided. These and further aspects and embodiments of the invention are described in more detail below and with reference to the accompanying examples and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0108] Figure 1 Shown is a flow chart summarizing the different experiments completed in this study and the corresponding number of samples used in each step.

[0109] Figure 2A survey of plasma DNA fragmentation using whole genome sequencing in a pan-cancer context is shown. A, The size spectrum of cfDNA can be determined by paired-end sequencing of plasma samples and reflects its organization around nucleosomes. cfDNA is released in the blood circulation in a variety of ways, each of which leaves a mark on the fragment size. The size spectrum of cfDNA was inferred by analyzing sWGS (n=344 plasma samples from 65 healthy controls and 200 cancer patients), and the size spectrum of mutant ctDNA was inferred by personalized capture sequencing (n=18 plasma samples). B, Fragment size distribution of 344 plasma samples from 200 cancer patients. Based on previous literature (3), patients were divided into two groups: orange represents cancer samples previously observed to have low ctDNA levels (renal cancer, bladder cancer, pancreatic cancer, and glioma) and blue represents cancer samples observed to have higher ctDNA levels (breast cancer, melanoma, ovarian cancer, lung cancer, colorectal cancer, bile duct cancer, etc., see Table 1). C, The proportion of cfDNA fragments below 150bp in the cancer groups defined in B. Kruskal-Wallis test for size distribution differences showed significant differences between the group of cancer types that released a large amount of ctDNA and the group that released a small amount of ctDNA and the group of healthy individuals (p < 0.001). D, Proportion of cfDNA fragments below 150 bp by cancer type (all samples). Cancer types represented by less than 4 individuals were grouped in the "other" category. The red line represents the median proportion of each cancer type.

[0110] Figure 3 The cfDNA size distribution of all plasma samples of healthy individuals and cancer patients included in this study is shown, depending on their cancer type determined by sWGS. The plasma samples shown here are collected from: renal cancer (n = 33), glioblastoma (n = 11), bladder cancer (n = 19), breast cancer (n = 34), melanoma (n = 21), pancreatic cancer (n = 7), ovarian cancer (n = 59), lung cancer (n = 8), colorectal cancer (n = 21), bile duct cancer (n = 14), cervical cancer (n = 1), penile cancer (n = 1), endometrial cancer (n = 1)), thymoma (n = 1), hepatocellular carcinoma (n = 1). The size spectrum of cfDNA from healthy individuals (n = 46) is also shown.

[0111] Figure 4Depicted is the size spectrum of mutant ctDNA determined using animal models and personalized capture sequencing.A, in the case of using sWGS, a mouse model with xenografted human tumor cells can distinguish DNA fragments released by cancer cells (reads aligned with the human genome) from DNA released by healthy cells (reads aligned with the mouse genome).B, fragment size distribution from plasma extracted from mice with xenografted human ovarian tumors, which shows ctDNA (red) derived from tumor cells and cfDNA (blue) derived from non-cancerous cells.Two vertical lines indicate 145bp and 167bp.Indicates the read fraction shorter than 150bp.C, developed for the design of personalized hybrid capture sequencing groups for the size spectrum of mutant DNA and non-mutant DNA in plasma from 19 patients with advanced cancer.The capture group contains somatic mutations identified in tumor tissue by WES.Then an average of 165 mutations per patient are analyzed from matched plasma samples.The reads are aligned and divided into fragments carrying reference sequences or mutant sequences.Calculate the fragment size of paired end reads. D, Size profiles of mutant and non-mutant DNA in plasma from 19 patients with advanced cancer were determined by tumor-guided capture sequencing. The fraction of reads shorter than 150 bp is indicated.

[0112] Figure 5 The insertion size distribution (A to S) determined by hybrid capture sequencing for 19 patients included in the mutant DNA size distribution analysis is shown. The size distribution of mutant DNA fragments is shown in red, and the distribution of non-tumor reference cfDNA from the same sample is shown in gray. The vertical dotted lines represent 145bp and 167bp. The insertion size is determined by summarizing the insertion sizes observed for mutant DNA and reference DNA from all samples of each patient.

[0113] Figure 6Shown is the improvement of tumor score from plasma sequencing with size selection. A, plasma samples collected from ovarian cancer patients were analyzed in parallel without size selection or using computer and in vitro size selection. B, the accuracy of in vitro and on-machine size selection determined for 20 healthy control groups. The size distribution before size selection is shown in green, after on-machine size selection (with sharp cutoff at 90 and 150bp) is shown in blue, and after in vitro size selection is shown in orange. C, when the ctDNA MAF of TP53 mutation as determined by TAm-Seq is 0.271, SCNA analysis of sWGS of plasma DNA from ovarian cancer patients collected before starting treatment. The inferred amplification is shown in blue and the deletion is shown in orange. The copy number neutral point region is shown in gray. D, SCNA analysis of plasma samples from the same patient as group C collected three weeks after the start of treatment. The MAF of TP53 mutation is 0.068 and ctDNA is not detected by sWGS at this time point (before size selection). E, Analysis of the same plasma sample as D after in vitro size selection of fragments with lengths ranging from 90 bp to 150 bp. After in vitro size selection, the MAF of TP53 mutations increased to 0.402, and SCNAs were clearly shown by sWGS. Compared with C and D, more SCNAs were detected (e.g., in chr2, chr9, chr10).

[0114] Figure 7 The insert size distribution for each plasma sample from the 13 ovarian cancer patients from the OV04 cohort collected before and after treatment, as determined by sWGS, is shown. The distribution of cell-free DNA (cfDNA) without size selection is shown in green, and the distribution of the same cfDNA sample after size selection is shown in orange. The vertical lines represent the range of fragments between 90 and 150 bp selected with the PippinHT box. It should be noted that patients OV04-292 and OV04-300 showed altered fragmentation spectra, indicating that there may be problems with sample preparation or storage before analysis.

[0115] Figure 8 Quality control assessment of in vitro size selection estimated with sWGS and targeted sequencing is shown. A, Size distribution of DNA fragments of plasma samples included in the size selection study assessed by sWGS before size selection (green) and after in vitro size selection (orange). The two vertical dashed lines indicate the size selection range of 90 bp to 150 bp. B, Ratio of the non-reference allele fraction corresponding to the sequencing background noise as determined during targeted sequencing (TAm-Seq) of plasma DNA samples from ovarian cancer patients with and without in vitro size selection.

[0116] Fig. 9 The second quality control assessment of in vitro and on-machine size selection is shown. 20 plasmas are selected from healthy controls, DNA is extracted and sWGS is performed on these samples without size selection, with in vitro and on-machine size selection. A, the size spectrum determined for each sample and condition. B, the score of repeated reads is improved, and therefore these are removed for any downstream size selection analysis. In order to determine whether size selection can introduce more sequencing noise during analysis, the QC metric called median absolute pairwise difference (MAPD) ​​algorithm is used to find sequencing noise. MAPD measures the absolute difference between the log2 CN ratios of each pair of adjacent regions (neighboring bin), and finds the median in all regions. Higher MAPD scores reflect larger noise, which is generally associated with poor quality samples. Regardless of the size selection conditions, all samples show that MAPD scores are 0.01 (+-0.01). C, in addition to noise estimation, the ctDNA scores between 20 control samples as estimated by t-MAD scores are also compared. The t-MAD scores for samples without size selection were not significantly different from the t-MAD determined after on-board size selection (t-test, p=0.43), but a significant difference was observed with samples after in vitro size selection (t-test, p=0.0068). Even though the t-MAD values ​​were improved after in vitro size selection, the mean (0.011) and maximum (0.016) values ​​detected were still confined within the threshold limits empirically determined from the entire control group (n=65). D, The DNA yield recovered after in vitro size selection was determined (because on-board size selection is not affected by this technical bias).

[0117] Fig.10Shown is the quantification of ctDNA enrichment by sWGS with on-machine size selection and t-MAD.A, the workflow of quantifying the tumor score from SCNA as a whole genome score named t-MAD.B, the correlation between the MAF of SNV determined by digital PCR or hybrid capture sequencing and the t-MAD score determined by sWGS.The data includes 97 samples from cancer patients of various cancer types with matching MAF measurements and t-MAD scores.The Pearson correlation (coefficient r) between MAF and t-MAD scores is calculated for all cases where MAF>0.025 and t-MAD>0.015.Linear regression indicates a fit (purple solid line) with a slope of 0.44.C, the comparison of t-MAD scores determined by sWGS between healthy samples, cancer type patients with low ctDNA amounts in circulation, and samples collected from cancer type patients with high ctDNA amounts in circulation.All samples that can calculate t-MAD are included. D, ROC analysis of the classification of these plasma samples from high ctDNA cancer samples (n=189) and plasma samples from healthy controls (n=65) using t-MAD, the area under the curve (AUC) without the selected size is 0.69 (black solid curve). After applying on-machine size selection to samples from cancer patients, we observed an AUC of 0.90 (black dotted curve). E, t-MAD was determined from longitudinal plasma samples of colorectal cancer patients. t-MAD was analyzed before and after on-machine size selection of DNA fragments of 90 to 150bp, and then compared with the RECIST status of the patient. F, on-machine size selection was applied to 6 long-term follow-up patients. t-MAD scores were determined before and after on-machine size selection of short DNA fragments. Dark blue circles indicate samples in which ctDNA was detected both with and without on-machine size selection. Light blue circles indicate samples in which ctDNA was detected only after on-machine size selection. Hollow circles indicate samples in which ctDNA was not detected by any analysis. The time to assessment of RECIST status is indicated by a red bar for progression, or by an orange bar for regression or stable disease.

[0118] Fig.11 A comparison of MAF and t-MAD scores is shown according to the cancer type for which matching data is available. Data from ovarian cancer, breast cancer, bile duct cancer, colorectal cancer, and lung cancer are detailed. Other cancer types are grouped in the "Other" category. Samples are labeled according to their t-MAD scores, with t-MAD < 0.015 filled in light purple and t-MAD > 0.015 filled in dark purple. When n > 5 and t-MAD > 0.015, Pearson correlation, p-value, and slope are indicated.

[0119] Fig.12 Plasma DNA from breast cancer patients is shown, which is spiked into pooled plasma DNA from healthy individuals. It was diluted in a step series of 10 times, 100 times and 1000 times. A total of 10ng DNA was used for initial DNA library preparation. The allele fraction of the net sample TP53 mutation was estimated to be about 45.6% by both WES and TAm-Seq, and used as a reference for dilution. In the dilution series data, the t-MAD score showed that SCNAs with very low coverage and mutant AF (as low as about 0.4% AF, or 100× diluted samples) were detected. In addition, the sequencing data has been size-selected on-board for short fragments (90 to 150bp), which improves the t-MAD score of lower AF.

[0120] Fig.13 The available RECIST volume (in mm) determined by CT scan is shown to be compared with tMAD score and fragmentation characteristics. RECIST volume is compared with: tMAD score (A), fragment is 20 to 150bp ratio (B), fragment is 100 to 150bp ratio and fragment is 163 to 169bp ratio ratio (C), fragment is 20 to 150bp ratio and fragment is 180 to 220bp ratio ratio (D), statistical amplitude is 10bp peak value and valley value (E), and fragment is 250 to 350bp ratio (F). Correlation and P value are calculated for each comparison.

[0121] Fig.14Quantification of ctDNA enrichment by sWGS in the presence of in vitro size selection is shown. A, the effect of in vitro size selection on t-MAD scores. For each of the 48 plasma samples collected from 35 patients, the t-MAD score was determined from sWGS after in vitro size selection (y-axis) and without size selection (x-axis). In vitro size selection improved the t-MAD scores of almost all samples, with the median increasing by 2.1 times (ranging from 1.1 to 6.4 times). The t-MAD scores for 46 samples from healthy individuals determined from sWGS were <0.015 both before and after in vitro size selection. B, ROC analysis of the classification of these plasma samples from cancer samples (n=48) and plasma samples from healthy controls (n=46) using t-MAD, with an area under the curve (AUC) of 0.64 (green curve) without size selection. After applying on-machine size selection to samples from cancer patients and healthy patients, an AUC of 0.78 was observed (blue curve), and after in vitro size selection, an AUC of 0.97 (orange curve). C, Comparison of t-MAD scores determined from sWGS between matched ovarian cancer samples with and without in vitro size selection. A t-test for the difference in means showed that the tumor score (measured by t-MAD) was significantly improved with in vitro size selection (p < 0.0001). D, Detection of SCNAs in 15 genes frequently mutated in recurrent ovarian cancer, SCNAs were measured in plasma samples collected during treatment of 35 patients. Patients are ranked from left to right by increasing tumor score as quantified by tMAD (before in vitro size selection). SCNA is marked as detected gene if the relative copy number in the region is greater than 0.05. Empty squares represent copy number neutral regions, lower left triangles in light blue indicate SCNA detection without size selection, and upper right triangles in dark blue indicate SCNA detection after in vitro size selection.

[0122] Fig.15 Analysis of each of 48 plasma samples collected from 35 ovarian cancer patients with and without size selection is shown. A, There is a negative correlation between the ctDNA score represented by the t-MAD score and the enrichment level (Pearson, -0.49, p < 0.001). B, For almost all samples, the t-MAD score determined from sWGS with size selection is higher than that without size selection, with the median increased by 2.1 times. The enrichment factor determined by t-MAD with size selection varies from sample to sample, but is higher for samples with low initial t-MAD scores. Values ​​from healthy individuals are added for comparison purposes.

[0123] Fig.16 SCNA analysis of segmental log2 ratios determined after sWGS is shown. This was performed using a list of 29 genes frequently mutated in recurrent ovarian cancer from plasma samples collected at baseline and after treatment for 13 patients. The log2 ratios represent samples without size selection and with in vitro size selection of shorter DNA.

[0124] Fig.17 Shown is the improvement of detecting somatic changes by WES in multiple cancer types with size selection.A, MAF analysis of mutations detected by WES in 6 HGSOC patients without size selection and with in vitro and on-machine size selection.B, comparison of size-selected WES data with unselected WES data for evaluating the number of mutations detected in plasma samples from 6 HGSOC patients.For each patient, the first bar in light blue shows the number of mutations called without size selection, the second bar quantifies the number of mutations called after adding those determined with on-machine size selection, and the third bar in dark blue shows the number of mutations called after adding the mutations called after in vitro size selection.C, patients (n=16) were retrospectively selected from groups with different cancer types (colorectal cancer, cholangiocarcinoma, pancreatic cancer, prostate cancer) participating in early clinical trials.Matched tumor tissue DNA can be used for each plasma sample, and 2 patients also have biopsies collected at relapse. Tumor tissue DNA and plasma DNA samples were subjected to WES, and on-machine size selection was applied to the data. 2061 / 2133 (97%) of the shared mutations detected by WES showed higher MAF after on-machine size selection. D, compared with the mutations called by the WES of the matched tumor tissue, mutations detected only after computer selection of the WES data from 16 patients (as in C). 3 of the 16 patients did not identify additional mutations after on-machine size selection. Of the 82 mutations detected in plasma after on-machine size selection, 23 (28%) had low signal levels in tumor WES data and were not identified initially in these samples.

[0125] Fig.18The mutant allele fraction (MAF) of each single nucleotide variant (single nucleotide variant, SNV) called by WES on the OV04 sample without and with selected size is shown. A, for most mutations detected from plasma samples of 6 HGSOC patients, the MAF (vertical) determined by WES with in vitro size selection is higher than the MAF (horizontal) without in vitro size selection. B, enrichment is also observed in the same samples after on-machine size selection from WES data.

[0126] Fig.19 Depicted are mutations in 9 clinically important genes detected by WES with and without size selection of short DNA fragments. All plasma samples submitted to WES were analyzed (6 ovarian cancer cases from the OV04 study, and 16 cancers from the CoPPO study). Mutations called without size selection, as well as new mutations called by WES after in vitro and on-machine size selection, were integrated.

[0127] Fig. 20 Shown are: A, for most samples, including samples collected at baseline (red circles) and after initial treatment (blue triangles), the MAF of TP53 mutations determined by TAm-Seq was higher than that without size selection in the presence of in vitro size selection. Only 26 samples collected from 13 patients and samples collected before and after treatment are shown. The dashed area highlights samples that initially had low MAF (<5%), where methods such as whole exome sequencing (at approximately 100× sequencing depth) would be ineffective, and where in vitro size selection enriched MAF to >5% and can therefore be used for large-scale analysis. B, comparison of MAF detected by TAm-Seq before treatment and after initial treatment with in vitro size selection (yellow triangles) and without size selection (green circles).

[0128] Fig.21The size distribution of mutant DNA and non-mutant DNA obtained from personalized sequencing is shown. A small portion of the 10 patients in this figure are sub-selected. The selected loci correspond to clinically validated variants (WES based on tumor tissue DNA). The left figure shows the size distribution of mutant DNA, and the right figure is the size distribution of corresponding non-mutant DNA. The colors represent the t-MAD values ​​of these patients (red indicates t-MAD score>0.02 and blue indicates t-MAD score<0.02). Mutant ctDNA confirmed enrichment in the size range of 90 to 150bp (as previously described in the manuscript). Non-mutants show lower enrichment in the size range of 90 to 150bp, but vary depending on the patient. Patients with the highest ctDNA concentration as determined by t-MAD have enrichment of shorter non-mutant DNA, while patients with lower t-MAD values ​​have fewer short fragments. This shows that even in non-mutant DNA, tumor signals (= non-mutant ctDNA) can be detected by analyzing the size of cfDNA fragments.

[0129] Fig. 22Depicted is the potential for enhancing ctDNA detection by combining SCNA and fragment size features. A, schematically illustrates the selection of different size ranges and features in the fragment size distribution. For each sample, the fragmentation feature includes the ratio (P) of fragments within a specific size range, the ratio between certain ranges, and the quantification of the 10bp oscillation amplitude within the range of 90 to 145 size bp calculated from periodic "peaks" and "valleys". B, the principal component analysis (Principal Component Analysis, PCA) of cancer and healthy samples is compared using data from t-MAD scores and fragmentation features. The fragmentation features shown in gray are not included in the following steps. C, the workflow of the predictive analysis of combined SCNA and fragment size features. Plasma DNA sWGS data from healthy controls are divided into a training set (60% of samples) and a validation set (for both validation data 1 and validation set 2). sWGS data of plasma samples from a pan-cancer cohort of 182 samples from patients with cancer types (colorectal cancer, cholangiocarcinoma, lung cancer, ovarian cancer, breast cancer) with high ctDNA levels were divided into training sets (60% of samples) and validation sets (validation data 1, and healthy individual validation sets). Another sWGS dataset of 57 samples from cancer types (glioma, renal cancer, pancreatic cancer) exhibiting low ctDNA levels was used as validation data 2 together with the healthy individual validation set. D, ROC curves of validation data 1 (samples from cancer patients with high ctDNA levels = 68, healthy = 26) for 3 prediction models established based on the pan-cancer training cohort (cancer = 114, healthy = 39). The beige curve represents the ROC curve for classification using only t-MAD, the long green dashed line represents the logistic regression model combining the first 5 features (t-MAD score, 10bp amplitude, P (160 to 180), P (180 to 220) and P (250 to 320)) based on recursive feature elimination, and the red dashed line shows the results of the random forest classifier trained on the combination of the same 5 features independently selected for the best RF prediction model. E, ROC curve of the validation data 2 (samples from cancer patients with low ctDNA levels = 57, healthy = 26) of the same 3 classifiers as D. The beige curve represents the model using only t-MAD, the long green dashed line represents the logistic regression model combining the first 5 features (t-MAD score, 10bp amplitude, P (160 to 180), P (180 to 220) and P (250 to 320)), and the red dashed line shows the results of the random forest classifier trained on the same combination of 5 prediction features. F, plot showing the probability of classifying all samples in the two validation datasets as cancer using the RF model. The samples are separated by cancer type and sorted by the RF probability of being classified as cancer within each sample.The horizontal dashed line indicates a probability of 50%, and the light long dashed line indicates a probability of 33%.

[0130] Fig.23 ROC analysis of cfDNA fragmentation signatures between healthy samples and samples from patients with high ctDNA cancer is shown.

[0131] Fig.24 A comparison of t-MAD scores with 9 fragmentation features determined by sWGS is shown in 147 plasma samples from cancer patients included in the training and validation datasets of the classifier model. A correlation score was estimated for each cross comparison and the value is shown on the lower left side of the figure.

[0132] Fig.25 Performance metrics of different algorithms are shown: logistic regression (on t-MAD score and fragmentation features); and random forest (RF), on training set data from sWGS (n=153; 114 cancer samples and 39 healthy controls). The median ROC score and accuracy value for each model are shown, as well as a 0.95 confidence level.

[0133] Fig.26 LR and RF models are shown, which detect cancer from healthy samples with separate fragmentation features. A, ROC curves of the first validation sample set (cancer = 68, healthy = 26) of 2 classifiers established based on the pan-cancer training group (cancer = 114, healthy = 39). The orange curve represents the ROC of the logistic regression model trained only with fragmentation features without t-MAD, and the red dotted line shows the results of the random forest classifier trained on the combination of the best 3 predicted fragmentation features (amplitude_10bp, P (160 to 180) and P (250 to 320)). B, ROC curves of the second validation sample set (cancer = 57, healthy = 26) of 2 classifiers established based on the same training set as A. The orange curve represents the logistic regression model trained only with fragmentation features, and the red dotted curve shows the results of the random forest classifier trained on the combination of 3 predicted features (amplitude_10bp, P (160 to 180) and P (250 to 320)). C, plot showing the probability of classifying the second validation dataset (described in B) as cancer using the RF model. Samples are sorted by cancer type and probability of being classified as cancer. The horizontal dashed line represents a probability of 50%.

[0134] Fig. 27The probability of cancer classification for a given renal cell carcinoma (renal cell carcinoma, RCC) patient plasma sample by a random forest (RF) model is shown, as shown on the y-axis. Patient plasma samples are shown on the x-axis. For each patient, the ctDNA content of the same plasma sample (and in some cases the urine supernatant matched) was evaluated by INVAR-TAPAS and t-MAD analysis. The circle indicates a patient in which ctDNA is not detected in any liquid by any method. The triangle indicates a patient in which ctDNA is detected in any liquid by any method. DETAILED DESCRIPTION

[0135] Some aspects and embodiments of the present invention will now be discussed with reference to the accompanying drawings. Other aspects and embodiments will be apparent to those skilled in the art. All documents mentioned in this text are incorporated herein by reference.

[0136] In describing the present invention, the following terminology will be employed and is intended to be defined as indicated below.

[0137] As used herein, "computer-implemented method" should be taken to mean a method whose execution involves the use of a computer, a computer network or other programmable devices, wherein one or more features of the method are fully or partly implemented by a computer program.

[0138] As used herein, a "sample" may be a biological sample, such as a cell-free DNA sample, a cell (including circulating tumor cells) or a tissue sample (such as a biopsy), a biological fluid, an extract (such as a protein or DNA extract obtained from a subject). In particular, the sample may be a tumor sample, a biological fluid sample containing DNA, a blood sample (including a plasma or serum sample), a urine sample, a cervical smear, a cerebrospinal fluid sample, or a non-tumor tissue sample. Urine and cervical smears have been found to contain cells, and therefore suitable samples for use according to the present invention may be provided. Other sample types suitable for use according to the present invention include fine needle aspirates, lymph nodes, surgical margins, bone marrow, or other tissues from the tumor microenvironment, in which trace amounts of tumor DNA may be found or expected to be found. The sample may be a sample freshly obtained from a subject (such as a blood extract), or may be a sample that has been processed and / or stored (such as frozen, fixed, or subjected to one or more purification, enrichment, or extraction steps, including centrifugation) before determination. The sample may be obtained from one or more of the above-mentioned biological samples by an enrichment or amplification process. For example, the sample may comprise a DNA library generated from a biological sample, and may optionally be a barcoded or otherwise labeled DNA library. Multiple samples may be obtained from a single patient, such as continuously during a treatment process. In addition, multiple samples may be obtained from multiple patients. Sample preparation may be as described in the Materials and Methods section of this article.

[0139] As used herein, "and / or" should be deemed to specifically disclose each of the two specific features or components with or without the other. For example, "A and / or B" should be deemed to specifically disclose each of the following: (i) A, (ii) B, and (iii) A and B, just as if each were individually listed herein.

[0140] Provide sequence reads

[0141] Sequence read data can be provided or directly obtained, for example, by sequencing a cfDNA sample or library, or by obtaining or providing generated sequencing data, for example, by retrieving sequence read data from a non-volatile or volatile computer memory, a data storage, or a network location. In the case of obtaining sequence reads by sequencing a sample, the median mass of the input DNA can be 1 to 100 ng, for example, 2 to 50 ng or 3 to 10 ng in some cases. DNA can be amplified to obtain a library having, for example, 100 to 1000 ng DNA. Sequence reads can be in a suitable data format, such as FASTQ.

[0142] Sequence data processing and error suppression

[0143] Prior to or as part of the step of compressing the reads into read families, the sequence read data (e.g., FASTQ files) may be subjected to one or more processing or cleaning steps. For example, the sequence data files may be processed using one or more tools selected from, for example, FastQC v0.11.5, tools for removing adapter sequences (e.g., cutadapt v1.9.1). The sequence reads (e.g., trimmed sequence reads) may be aligned to a suitable reference genome (e.g., human reference genome GRCh37 of a human subject).

[0144] "Read" or "sequencing read" as used herein may be understood to mean a sequence that has been read from one molecule and read once. Depending on the sequencing performed, each molecule may be read many times.

[0145] A "classifier" or "classification algorithm" can be a model or algorithm that maps input data, such as cfDNA fragment size features, to categories, such as cancerous or non-cancerous sources. In some embodiments, the present invention provides a method for detecting, classifying, predicting or monitoring cancer in an object. In particular, one or more pattern recognition algorithms can be used to evaluate data obtained from sequence analysis, such as fragment length and / or copy number (e.g., trimmed median absolute deviation "t-MAD" from the copy number neutral point). Such an analysis method can be used to form a prediction model that can be used to classify test data. For example, a convenient and particularly effective classification method uses multivariate statistical analysis modeling, first using data from samples of known categories (e.g., from subjects known to suffer from specific cancers) ("modeling data") to form a model ("predictive mathematical model"), and secondly classifying unknown samples (e.g., "test samples") according to category.

[0146] Pattern recognition is to analyze data using multivariate statistics (both parametric and non-parametric), and thereby classify samples based on a series of observed measured values ​​and predict the values ​​of some dependent variables. There are two main methods. One group of methods is called "unsupervised", and these simply reduce data complexity in a reasonable manner, and also produce a display diagram that can be interpreted by the human eye. However, this type of method may not be suitable for developing a clinical assay that can be used to classify samples derived from an object without relying on an initial sample group for training a prediction algorithm.

[0147] Another approach is called "supervised," in which a training set of samples with known categories or outcomes is used to generate a mathematical model, which is then evaluated with an independent validation data set. Here, a "training set" of sequence information, such as fragmentation features and / or copy number features, is used to build a statistical model that correctly predicts the category of each sample. The training set is then tested with independent data (called a test or validation set) to determine the robustness of the computer-based model. These models are sometimes called "expert systems," but can be based on a range of different mathematical procedures, such as support vector machines (SVMs), decision trees, k-nearest neighbors, and naive Bayes ( Bayes), each of which is intended to be used in accordance with the present invention herein. As described in detail in the Examples herein, logistic regression (LR) and random forests (RF) are used for variable selection and classification of samples as "healthy" or "cancer". Supervised methods can use data sets with reduced dimensionality (e.g., the first few main components), but typically use unreduced data with all dimensions. The robustness of the predictive model can also be checked by omitting selected samples from the analysis using cross-validation. Custom panel sequencing (TAPAS)

[0148] The custom group sequencing used herein refers to the sequencing of the target region and / or gene. This can be selected or customized capture groups using targeted target genes, such as genes that are usually mutated in cancer and / or genes found to carry mutations in the tumor of the target object (e.g., by sequencing matching tumor tissue DNA and plasma DNA samples). In some cases, the size range of the capture group can be 0.5Mb to 5Mb, such as 1Mb to 3Mb.

[0149] The following are given by way of example and should not be construed as limiting the scope of the claims.

[0150] Example

[0151] Materials and methods

[0152] Study Design

[0153] Collected 344 plasma samples from 200 patients with various cancer types and 65 plasma samples from 65 healthy controls. Among these patients, 172 individuals were recruited by prospective clinical research conducted at Addenbrooke's Hospital (Addenbrooke's Hospital, Cambridge, UK) in Cambridge, UK, and the research was approved by local research ethics committee (REC reference number: 07 / Q0106 / 63; and NRES Committee East of England (NRES Committee East of England)-Cambridge Center 03 / 018)). Written informed consent was obtained from all patients, and blood samples were collected before and after starting to be treated with surgery or chemotherapeutic agents. According to the manufacturer's instructions, DNA was extracted from 2mL plasma using QIAamp circulating nucleic acid kit (Qiagen) or QIAsymphony (Qiagen). In addition, 28 patients were recruited as part of the Copenhagen Prospective Personalized Oncology (CoPPO) program (reference: PMID: 25046202) conducted at Rigshospitalet, Copenhagen, Denmark, which was approved by the local research ethics committee. Baseline tumor tissue biopsies were available from all 28 patients, as well as re-biopsies collected from two patients at relapse, including matched plasma samples. Brain tumor patients were recruited as part of the BLING study (REC-15 / EE / 0094) at Addenbrooke's Hospital in Cambridge, UK. Bladder cancer patients were recruited at the Netherlands Cancer Institute, Amsterdam, The Netherlands, and were approved in accordance with national guidelines (N13KCM / CFMPB250) (47). 65 plasma samples were obtained from healthy control individuals using a similar protocol (Seralab). Plasma samples were frozen and thawed no more than twice to reduce artificial fragmentation of cfDNA. Figure 1 Describe the study as a flow chart.

[0154] In vitro size selection

[0155] 8 to 20 ng of DNA were loaded into a 3% agarose box (HTC3010, Sage Bioscience) and size selected with PippinHT (Sage Bioscience) according to the manufacturer's protocol. Quality control of in vitro size selection was performed on 20 healthy control samples. For any downstream size selection analysis, duplicate reads observed in the case of in vitro selection were removed. A QC metric called the median absolute pairwise difference (MAPD) ​​algorithm was used to determine sequencing noise. MAPD measures the absolute difference between the log2 CN ratios of each pair of adjacent regions and determines the median of all regions. Higher MAPD scores reflect greater noise and are generally associated with samples of poor quality. All samples exhibited a MAPD score of 0.01 (+-0.01) regardless of the size selection conditions.

[0156] TAm-Seq

[0157] Using primers designed to assess single nucleotide variants (SNVs) and small deletion mutations across selected hotspots and the entire coding region of TP53, tagged amplicon deep sequencing libraries were prepared as previously described (34). Libraries were sequenced using MiSeq or HiSeq 4000 (Illumina).

[0158] Shallow whole genome sequencing (sWGS)

[0159] Indexed sequencing libraries were prepared using commercially available kits (ThruPLEX-Plasma Seq and / or Tag-Seq, Rubicon Genomics). Libraries were pooled in equimolar amounts and sequenced on a HiSeq 4000 (Illumina) to <0.4× coverage depth, generating paired-end reads of 150 bp. Sequence data were analyzed using an in-house pipeline. After removal of contaminating adapter sequences, paired-end sequence reads were aligned to the human reference genome (GRCh37) using BWA-mem (48). PCR and optical duplicates were marked using the MarkDuplicates (Picard Tools) feature, and these were excluded from downstream analysis along with reads of low mapping quality and supplementary alignments. Where necessary, reads in all samples were downsampled to 10 million for comparison purposes.

[0160] Somatic copy number aberration analysis

[0161] The analysis was performed in R using a software suite for shallow whole-genome sequencing copy number analysis called CNAclinic (https: / / github.com / sdchandra / CNAclinic) and the QDNAseq pipeline (49). Sequencing reads were randomly sampled to 10 million reads / dataset and assigned to non-overlapping regions of the same size (30 Kbp) over the entire genome length. The read counts in each region were corrected to account for sequence GC content and mappability. Regions that overlapped with "blacklist" regions (derived from the ENCODE project and 1000 Genome Project databases) that were prone to artifacts were excluded from downstream analysis. The read counts in the test samples were normalized by counts from the same treated healthy individuals and log2 converted to the obtained copy number ratio / genomic region. The read counts of the healthy controls were normalized by their whole-genome count median. The regions were then segmented using both the cyclic binary segmentation and hidden Markov model algorithms. The average log2R value for each region was calculated.

[0162] An internal empirical blacklist of abnormal read count regions was constructed. First, the median read counts per 30Kbp genomic region were calculated using 65 sWGS data sets from healthy plasma as a function of GC content and mappability. Then a 2D LOESS surface was applied and the difference between the actual count and the LOESS fitted value was calculated. The median of these residual values ​​in the 65 controls for each genomic region was calculated, and regions with median residuals greater than 4 standard deviations were blacklisted. The average segmental log2R value of each test sample overlapping the cfDNA blacklist was trimmed, and the median absolute value was calculated. The score is defined as the trimmed median absolute deviation (t-MAD) from log2R=0. R code for reproducing this analysis is provided in https: / / github.com / sdchandra / tMAD (which is incorporated herein by reference in its entirety).

[0163] Whole Exome Sequencing (WES)

[0164] Index sequencing libraries were prepared as described above (see method, sWGS). Plasma DNA libraries from each sample were prepared and merged together for exome capture (TruSeq exome enrichment kit, Illumina). The pool was concentrated using a SpeedVac vacuum concentrator (Eppendorf). Exome enrichment was performed according to the manufacturer's protocol. The enriched library was quantified using quantitative PCR (KAPA library quantification, KAPA Biosystems), and the DNA fragment size was observed by a bioanalyzer (2100Bioanalyzer, Agilent Genomics), and was merged in an equimolar ratio for paired-end next-generation sequencing on HiSeq4000 (Illumina). Sequencing reads were separated by signals so that there was zero mismatch in the barcode. Paired-end alignments with the GRCh37 reference genome were performed using BWA-mem for all exome sequencing data (germline / plasma / tumor tissue DNA). Picard marker PCR was used to repeat. The Genome Analysis Tool Kit (GATK) was used for base quality score recalibration and local realignment.

[0165] Mutation Calling

[0166] For all bases with PHRED quality ≥ 30, the mutant allele fraction (MAF) of each single base locus is calculated with MuTect2. Then filter parameters are applied so that if no mutant reads of the allele are observed in the germline DNA at the locus covering at least 10×, and if at least 4 reads (wherein at least 1 read on each chain (forward and reverse)) supporting the mutant are found in the plasma data, mutations are called. At the locus where coverage is <10× and there is no mutant read in normal DNA, if the previous plasma sample does not show evidence of mutation and is fully covered (10× or more), mutations are called in plasma. The method of integrated signal amplification called tumor non-invasive inquiry is used to aggregate the mutations called before and after size selection. This method combines different subsets of mutations called from the same plasma DNA sample using different processing methods. The mutation aggregation used in this study is formalized as follows: aggregated mutations=mutations detected without size selection U (mutations detected with in vitro size selection U mutations detected with on-machine size selection).

[0167] On-board size selection

[0168] Paired end reads are generated by sequencing the DNA at both ends of the fragment present in the library. The original length of the DNA can be inferred using the mapping position of the end of the read in the genome. Once the comparison is completed, Samtools software is used to select the paired reads corresponding to the fragment length within a specific range. Mutect2 is used to call mutations from the data of the size selection on the machine, as described in the previous part.

[0169] Tumor-guided capture sequencing

[0170] Matched tumor tissue DNA and plasma DNA samples of 19 patients with advanced cancer collected from RigsHospitalet (Copenhagen, Denmark) were sequenced by WES. Variants were called from these samples by mutation calling (see above). Hybridization-based capture for longitudinal plasma sample analysis was designed to use SureDesign (Agilent) to cover these variants for each patient. The median of 160 variants per patient was included, and in addition, 41 common target genes for pan-cancer analysis were included in the tumor-guided sequencing group. Index sequencing libraries were prepared according to sWGS (see above). Plasma DNA libraries from each sample were prepared and merged together for tumor-guided capture sequencing (SureSelect, Agilent). The merged product was concentrated using SpeedVac vacuum concentrator (Eppendorf). Capture enrichment was performed according to the manufacturer's protocol. The enriched libraries were quantified using quantitative PCR (KAPA library quantification, KAPA Biosystems) and DNA fragment size controlled by a bioanalyzer (2100 Bioanalyzer, Agilent Genomics) and combined in equal molar ratios for paired-end next-generation sequencing on HiSeq4000 (Illumina). Sequencing reads were signal separated to make zero mismatches in the barcode. Paired-end alignments with the GRCh37 reference genome were performed using BWA-mem for all exome sequencing data (including germline, plasma, and tumor tissue DNA, where produced). Picard was used to mark PCR repeats. Base quality score recalibration and local re-alignment were performed using the Genome Analysis Tool Kit (GATK).

[0171] Classification Analysis

[0172] 304 samples (182 high ctDNA cancer samples, 57 low ctDNA cancer samples and 65 healthy controls) were subjected to preliminary analysis. For each sample, the following features were calculated from sWGS data: t-MAD, amplitude_10bp, P(20 to 150), P(160 to 180), P(20 to 150) / P(160 to 180), P(100 to 150), P(100 to 150) / P(163 to 169), P(180 to 220), P(250 to 320), P(20 to 150) / P(180 to 220) (see Table 2). The data were arranged in a matrix in which the rows represent each sample and the columns contain the above features, wherein the additional "category" column has a binary label of "cancer" / "health". The following analysis was performed in R using random forests, caret and PROC packages. The caret software package is available and described at the following URL: http: / / topepo.github.io / caret / index.html. Exemplary source code for the classification algorithms described in the Examples herein is provided below in the section entitled " Code " is shown in the section. Pairwise correlations between features were calculated to assess multicollinearity in the dataset. Single variables were selected for removal from pairs with Pearson correlations > 0.75. Highly correlated fragmentation features that were composites of separate variables already in the dataset, such as P(20 to 150) / P(180 to 220), were preferentially removed. Zero variance and linear dependence of features were also assessed, but without labeling. After this preprocessing, the following 5 variables were selected for further analysis: t-MAD, amplitude_10bp, P(160 to 180), P(180 to 220), and P(250 to 320) (see Table 2). All 57 low ctDNA samples were set aside (set The remaining data matrix of high ctDNA cancer samples and healthy controls (n = 247) was randomly divided into 1 training dataset and 1 validation dataset in a ratio of 60:40, where different cancer types and healthy samples were represented in similar proportions. Therefore, the training data contained 153 samples (cancer = 114, healthy = 39), while the first validation set of high ctDNA cancer contained 94 samples (cancer = 68, healthy = 26). This validation dataset was only used for the final evaluation of the classifier.

[0173] Samples were classified as healthy or cancer using a linear machine learning algorithm and a nonlinear machine learning algorithm, namely logistic regression (LR) and random forest (RF). Each algorithm was paired with recursive feature selection to determine the best predictor. The analysis was performed with caret within a box of 5 repetitions of 10-fold cross validation on the training set. The algorithm was configured to explore all possible subsets of features. The best model for each classifier was selected using the ROC metric. In addition, the logistic regression model was trained using only t-MAD as a predictor to evaluate performance differences without adding fragmentation features. Finally, the classifier was tested using 68 high ctDNA cancer samples, 57 low ctDNA cancer samples, and 26 healthy controls set aside for validation, and its performance was quantified using the area under the curve in the ROC analysis.

[0174] A secondary analysis was performed on the same training cohort and validation cohort, where the only difference was the features used in the model. Here, we tested the predictive power of the fragmentation features without adding information from SCNA (i.e., t-MAD). Therefore, the features used were: Amplitude_10bp, P(160 to 180), P(180 to 220), and P(250 to 320).

[0175] Quantification of 10 bp periodic oscillations

[0176] The amplitude of the 10bp periodic oscillation observed in the size distribution of cfDNA samples is determined from sWGS data as follows. Local maxima and minima within the range of 75bp to 150bp were identified. The average value of their positions in the sample was calculated (for minimum values: 84, 96, 106, 116, 126, 137, 148 and maximum values: 81, 92, 102, 112, 122, 134, 144). In order to calculate the oscillation amplitude with 10bp periodicity observed below 150bp, the sum of the minimum values ​​was subtracted from the sum of the heights of the maximum values. The greater the difference, the more obvious the peak. The height of the x bp peak is defined as the number of fragments with a length of x divided by the total number of fragments. In order to define the local maximum, y positions are selected so that y is the maximum value in the interval [y-2, y+2]. The same principle is used to select the minimum value.

[0177] Example 1: Studying the fragmentation characteristics of tumor cfDNA

[0178] A catalog of cfDNA fragmentation signatures was generated using 344 plasma samples from 200 patients with 18 different cancer types and an additional 65 plasma samples from healthy controls ( Figure 1 and Figure 2A). The size distribution of cfDNA fragments in cancer patients was different compared to healthy individuals in the size ranges of 90 to 150 bp, 180 to 220 bp, and 250 to 320 bp ( Figure 2 B and Figure 3 cfDNA fragment sizes were significantly longer in the plasma of healthy individuals and in the plasma of patients with advanced glioma, renal, pancreatic, and bladder cancers than in other advanced cancer types, including breast, ovarian, lung, melanoma, colorectal, and cholangiocarcinoma (p < 0.001, Kruskal-Wallis; Figure 2 C). The classification of 18 cancer types based on the proportion of cfDNA fragments in the 20 to 150 bp size range is very similar to the ranking by Bettegowda et al. based on ctDNA concentrations measured by individual mutation assays ( Figure 2 D) (6). In contrast to previous reports (6, 34), this classification analysis was performed without any prior knowledge of mutations or somatic copy number alterations (SCNAs), but allowed exploration of ctDNA content in different cancers.

[0179] Example 2: Estimation of mutant ctDNA

[0180] Two highly specific approaches were used to determine the size profile of mutant ctDNA in plasma. First, sWGS from plasma of mice bearing human ovarian cancer xenografts was used to infer the specific size profiles of ctDNA and non-tumor cfDNA ( Figure 4 A). ctDNA fragment size shifted to less than 167 bp ( Figure 4 B). Second, using deep sequencing with patient-specific hybrid capture panels developed from whole-exome analysis of matched tumor samples, we determined the size profiles of mutant ctDNA in the plasma of 19 cancer patients ( Figure 4 C). By sequencing hundreds of mutations in cfDNA at a depth of >300×, allele-specific reads from mutant and normal DNA were obtained. Enrichment of DNA fragments carrying tumor mutant alleles was observed in fragments approximately 20 to 40 bp shorter than the size of nucleosomal DNA (multiples of 167 bp) ( Figure 4 D). Mutant ctDNA is generally more fragmented than non-mutant cfDNA, with ctDNA having the greatest enrichment in fragments between 90 and 150 bp ( Figure 5 ), and enriched in the size range of 250 to 320 bp. These data also indicate that mutant DNA in plasma from patients with advanced cancer (before treatment) is consistently shorter than the predicted mono- and di-nucleosomal DNA fragment lengths ( Figure 4 D).

[0181] Example 3: Selection of tumor-derived DNA fragments

[0182] These data indicate that ctDNA is shorter than non-tumor cfDNA and suggest that biological differences in fragment length can be exploited to improve ctDNA detection. In 48 plasma samples from 35 patients with high-grade serous ovarian cancer (HGSOC), the feasibility of selective sequencing of shorter fragments was established using in vitro size selection with a benchtop microfluidic device followed by sWGS ( Figure 6 A. Figure 7 and Figure 8 The accuracy and quality of size selection were evaluated using plasma from 20 healthy individuals ( Figure 6 B and Fig. 9 ). The utility of on-board size selection of fragmented DNA was also explored using read pair mapping from unprocessed sWGS data ( Figure 6 A) On-board size selection is performed once the reads are aligned to the genomic reference by selecting paired-end reads corresponding to fragment lengths in the 90 to 150 bp size range. Figure 6 C. Figure 6 D and Figure 6 E illustrates the role of in vitro size selection in a case of HGSOC. SCNAs were identified in plasma cfDNA before treatment when ctDNA concentration was high ( Figure 6 C). Only a few focal SCNAs were observed in a follow-up plasma sample collected 3 weeks after the start of chemotherapy (no size selection, Figure 6 D). In vitro size selection of plasma samples after the same treatment showed a median 6.4-fold increase in the amplitude of detectable SCNAs without size selection. Selective sequencing of shorter fragments in this sample led to the detection of multiple additional SCNAs that were not observed without size selection ( Figure 6 E), and a genome-wide copy number profile similar to that obtained before treatment when ctDNA levels were 4-fold higher ( Figure 6 C) In conclusion, selection of short DNA fragments in plasma can enrich tumor content on a genome-wide scale.

[0183] Example 4: Quantifying the impact of size selection

[0184] To quantitatively assess enrichment after size selection on a genome-wide scale, a metric from sWGS data (<0.4× coverage) was developed, called t-MAD (trimmed median absolute deviation from copy number neutral, see Fig.10A). All sWGS data were downsampled to 10 million sequencing reads for comparison. To define the detection threshold, the t-MAD scores of sWGS data from 65 plasma samples from 46 healthy individuals were measured and the maximum value was found (median = 0.01, range 0.004 to 0.015). In the comparison of t-MAD and mutant allele fraction (MAF) in high ctDNA cancer types assessed by digital PCR (dPCR) or WES in 97 samples, there was a high correlation (Pearson correlation, r = 0.80) between t-MAD and MAF for samples with t-MAD greater than the detection threshold (0.015) or MAF > 0.025 ( Fig.10 B). Fig.11 The slope of the t-MAD vs. MAF fitted line is shown to differ between cancer types (range 0.17 to 1.12), reflecting possible differences in the range of SCNAs. The sensitivity of t-MAD for detecting low ctDNA levels was estimated using a spike of DNA from a patient with a TP53 mutation diluted into DNA from a pool of 7 healthy individuals ( Fig.12 ), which confirmed that the t-MAD score was linear with ctDNA levels down to a MAF of approximately 0.01. In addition, even in samples with a MAF as low as 0.004, there were t-MAD scores greater than the sample detection threshold (0.015). t-MAD also strongly correlated with tumor volume determined by RECIST1.1 (Pearson correlation, r = 0.6, p < 0.0001, n = 35) ( Fig.13 ).

[0185] Using t-MAD, ctDNA was detected in 69% (130 / 189) of samples from cancer types showing high ctDNA levels ( Fig.10 C). In cancer types suspected of having low ctDNA levels (glioma, renal cancer, bladder cancer, pancreatic cancer), ctDNA was detected in 17% (10 / 57) of cases ( Fig.10 C). To increase the sensitivity of detecting t-MAD, on-board size selection of 90 to 150 bp DNA fragments from high ctDNA cancers (n=189) and healthy controls (n=65) was used ( Fig.10 D). Receiver operating characteristic (ROC) analysis comparing the t-MAD scores of the samples revealed an area under the curve (AUC) of 0.90 after on-board size selection compared to an AUC of 0.69 without size selection ( Fig.10 D).

[0186] To investigate whether size-selective sequencing could improve detection of response or disease progression, we used sWGS of longitudinal plasma samples from six cancer patients ( Fig.10 E and F) and on-board size selection of cfDNA fragments between 90 and 150 bp. In two patients, size-selected samples showed tumor progression 60 and 87 days before detection by imaging or unselected t-MAD analysis ( Fig.10 E and 10F). Other longitudinal samples showed improved detection of ctDNA using t-MAD and size selection ( Fig.10 F) Validation in large clinical studies will be required to determine the potential of selective sequencing of ctDNA for clinical application.

[0187] Example 6: Identification of more clinically relevant mutations using size selection

[0188] The ability of size selection to improve the sensitivity of detecting new mutations in cfDNA was examined. To test the effect on copy number aberrations, 35 patients with HGSOC were studied, as HGSOC is a primary copy number-driven cancer (35). t-MAD was used to quantify the enrichment of ctDNA in 48 plasma samples (including samples collected before and after the start of chemotherapy) with in vitro size selection. In vitro size selection resulted in an improvement in t-MAD scores calculated from sWGS data of 47 / 48 plasma samples (98%, t-test, p = 0.06), with an average 2.5-fold and median 2.1-fold improvement ( Fig.14 A). The t-MAD scores were then compared to those obtained for plasma samples from healthy individuals by sWGS. 44 of 48 size-selected HGSOC plasma samples (92%) had t-MAD scores greater than the highest t-MAD value determined in the in vitro size-selected healthy plasma sample, compared to only 24 of 48 (50%) without size selection ( Fig.14 A and Fig.15 ). ROC analysis comparing t-MAD scores of samples from cancer patients (before and after starting treatment, n=48) and healthy controls (n=46) revealed that after in vitro size selection, the AUC was 0.97, with a maximum sensitivity and specificity of 90% and 98%, respectively. This is significantly better than detection by sWGS without size selection (AUC=0.64) ( Fig.14 B).

[0189] This was then explored to determine whether the increased sensitivity resulted in the detection of SCNAs of potential clinical value. In the genome, 33 / 35 (94%) of HGSOC patients had higher t-MAD scores for SCNAs evaluated after size selection, and the absolute levels of copy number (log2 ratio) values ​​were significantly improved after in vitro size selection (t-test for the mean, p=0.003) ( Fig.14C). The relative copy number values ​​of 15 genes frequently altered in HGSOC were then compared (Table 3). Analysis of plasma cfDNA after size selection revealed a large number of SCNAs not observed in the same samples without size selection ( Fig.14 D), including amplification of key genes such as NF1, TERT, and MYC ( Fig.16 ).

[0190] To exclude the possibility that size selection might simply improve the sensitivity of sWGS analysis, we examined whether enrichment of surrogates was found. Whole exome sequencing was performed on plasma cfDNA from 23 patients with 7 cancer types ( Figure 2 Comparison of the size distribution of fragments carrying mutant or non-mutant alleles can be done using WES data ( Fig.17 A) and to show whether size selection can identify additional mutations. Six patients with HGSOC were selected and WES of plasma DNA was performed with and without in vitro size selection in the 90 to 150 bp range, analyzing time points before and after the start of treatment (36). In addition, onboard size selection was performed for the same range of fragment sizes ( Fig.17 A). Analysis of MAF of SNVs revealed statistically significant enrichment of tumor fractions under both in vitro size selection (mean 4.19-fold, median 4.27-fold increase, t-test, p<0.001) and on-machine size selection (mean 2.20-fold, median 2.25-fold increase, t-test, p<0.001). Fig.17 A and Fig.18 ). Three weeks after starting treatment, ctDNA levels were generally low (36), and therefore post-treatment plasma samples were further analyzed using labeled amplicon deep sequencing (TAm-Seq) (37). Enrichment of MAF by in vitro size selection was observed to be 0.9 to 118-fold (mean 2.1-fold, median 1.5-fold) compared to the same samples without size selection ( Fig.19 ).

[0191] Size selection with both in vitro and in silico methods increased the number of mutations detected by WES by an average of 53% compared to no size selection ( Fig.17 B). A total of 1023 mutations were identified in the unsize-selected samples. An additional 260 mutations were detected by in vitro size selection, and an additional 310 mutations were called after onboard size selection ( Fig.17B and Table 4). Novel mutations were also detected in tumor specimens, which excludes the possibility that the increased sensitivity of mutation detection is the result of sequencing artifacts. On-board size selection was then used in an independent cohort of 16 patients for whom matched tumor tissue DNA was available. On-board size selection enriched for MAF of nearly all mutations (2061 / 2133, 97%), with an average increase in MAF of ×1.7 ( Fig.17 C). For 13 of 16 patients (81%), additional mutations were identified in plasma after onboard size selection. 23 of these 82 additional mutations (28%) were determined to be present in matched tumor tissue DNA ( Fig.17 D). Notably, this includes mutations in key cancer genes including BRAF, ARID1A, and NF1 ( Fig. 20 ).

[0192] Example 7: Detecting cancer by supervised machine learning combining cfDNA fragmentation and somatic alteration analysis.

[0193] It is important to note that although in vitro and on-board size selection improve the sensitivity of the assay, they also result in the loss of cfDNA for analysis. Regions of the cancer genome that are not altered by mutations are also excluded and cannot contribute to the analysis ( Fig.21 ). It is hypothesized that exploiting other biological characteristics of the cfDNA fragmentation profile could enhance ctDNA detection.

[0194] Additional cfDNA fragmentation features were defined from the sWGS data, including (1) fragment proportions in multiple size ranges, (2) ratios of fragment proportions of different sizes, and (3) oscillations in fragment size density with a 10 bp periodicity ( Fig. 22 A). These fragmentation signatures were compared between cancer patients and healthy individuals ( Fig.23 ), and the feature representing the fragment ratio (P) of 20 to 150 bp showed the highest AUC (0.819). Principal component analysis (PCA) of samples represented by t-MAD and fragmentation features showed separation between healthy and cancerous samples, and fragmentation features clustered with t-MAD scores ( Fig. 22 B).

[0195] In addition, the potential of fragmentation signatures to enhance tumor DNA detection in plasma samples was explored. Prediction analysis was performed using t-MAD scores and 9 fragmentation signatures from 304 samples (239 from cancer patients and 65 from healthy controls) ( Fig. 22 C and Fig.24and Table 2). The nine fragmentation features determined from sWGS include five features based on the proportions of fragments (P) within the following defined size ranges: P(20 to 150), P(100 to 150), P(160 to 180), P(180 to 220), P(250 to 320); three features based on the ratios of these proportions: P(20 to 150) / P(160 to 180), P(100 to 150) / P(163 to 169), P(20 to 150) / P(180 to 220); and additional features based on the amplitude of oscillations with a 10 bp periodicity observed below 150 bp.

[0196] Logistic regression (LR) and random forest (RF) trained on 153 samples were used for variable selection and classification of samples into “healthy” or “cancer” and validated on two datasets of 94 and 83 independent samples ( Fig. 22 C). The best feature set for the LR model included t-MAD, 10 bp amplitude, P(160 to 180), P(180 to 220), and P(250 to 320). The same five variables were independently identified using the RF model (with some differences in their rankings). Fig.25 The performance metrics of different algorithms using cross validation on the training set data are shown. The source code for the classification algorithm is shown below in the section titled "Code". t-MAD ( Fig. 22 D and Fig.24 ), can distinguish cancer samples from healthy individuals with AUC = 0.764. The LR model improves the classification of samples to AUC = 0.908. The RF model (trained on a 153-sample training set) can distinguish cancer from healthy individuals even more accurately with AUC = 0.994 in a validation dataset (n = 94). On a second validation dataset containing low ctDNA cancer samples (n = 83) ( Fig. 22 E), t-MAD or LR alone did not perform well, with AUC values ​​of 0.421 and 0.532, respectively. However, the RF model was still able to distinguish samples from low ctDNA cancers from samples from healthy controls with AUC = 0.914. At 95% specificity, the RF model correctly classified 64 / 68 (94%) samples from high ctDNA cancers (colorectal cancer, bile duct cancer, ovarian cancer, breast cancer, melanoma) and 37 / 57 (65%) samples from low ctDNA cancers (pancreatic cancer, renal cancer, glioma) as cancer ( Fig. 22 F). In the second iteration of model training, only 4 fragmentation features are used and t-MAD is omitted ( Fig.26). The RF model can still distinguish cancer from healthy controls, although the AUC is slightly reduced (0.989 for cancer types with high ctDNA levels and 0.891 for cancer types with low ctDNA levels), indicating that the cfDNA fragmentation pattern is the most important predictive component.

[0197] Example 8: Using the Random Forest (RF) model to predict the detection of ctDNA in the body fluid of cancer patients

[0198] The random forest (RF) model according to the present invention and as described in Example 7 is based on the density or proportion of plasma cell-free DNA fragments with lengths of 20 to 150, 100 to 150, 160 to 180, 163 to 169, 180 to 220 and 250 to 320 bp and an oscillation amplitude with a periodicity of 10 bp, and can predict the probability that a given plasma sample has been collected from an individual with cancer.

[0199] Additionally, our data suggest that the output of this same RF classification model may allow for classification of cancer patient fluid samples into samples with sufficiently high ctDNA levels for detection by other methods (including those with higher sensitivity and / or allowing targeted analysis of specific somatic mutations) and those that do not have sufficiently high ctDNA levels.

[0200] After applying the RF model to plasma samples of patients with renal cell carcinoma (RCC), for those with a cancer probability >50% that passed the RF model:

[0201] By our Integration of Variant Reads from Custom Panel Sequencing (INVAR TAPAS) approach (see co-pending patent application GB1803596.4 filed on March 6, 2018, the contents of which are incorporated herein by reference), approximately 62% had detectable ctDNA in plasma;

[0202] Approximately 63% had detectable ctDNA in plasma by INVAR and / or t-MAD (the latter as described above);

[0203] By INVAR and / or t-MAD, approximately 81% had detectable ctDNA in plasma and / or urine. In contrast, only 11% of plasma samples with a cancer probability <50% by the RF model had detectable ctDNA. Fig. 27 Summarized in .

[0204] In summary, this analysis has the potential to highlight those cancer patients in which ctDNA analysis (by more sensitive or more targeted methods, such as INVAR-TAPAS) is more likely to produce information output. In turn, these samples are more likely to prove clinically useful, for example, potentially allowing the prediction of response to treatment, disease prediction and evaluation of clonal evolution by applying targeted methods by identifying resistance mutations. This can prove to be particularly relevant to those cancer types (such as renal cancer and glioblastoma) in which ctDNA detection is unreliable, even in the late stage of the disease where ctDNA detection is expected to be reliable (based on equivalent data from other cancer types). In addition, preliminary results (not shown) show that the above-mentioned RCC findings are confirmed in the glioblastoma group.

[0205]

[0206]

[0207]

[0208]

[0209]

[0210]

[0211]

[0212]

[0213]

[0214]

[0215]

[0216]

[0217]

[0218]

[0219]

[0220]

[0221]

[0222]

[0223]

[0224]

[0225]

[0226]

[0227]

[0228]

[0229]

[0230]

[0231]

[0232]

[0233]

[0234]

[0235]

[0236]

[0237]

[0238]

[0239]

[0240]

[0241]

[0242]

[0243]

[0244]

[0245]

[0246]

[0247]

[0248]

[0249]

[0250]

[0251]

[0252]

[0253]

[0254]

[0255]

[0256]

[0257]

[0258]

[0259]

[0260]

[0261]

[0262]

[0263]

[0264]

[0265]

[0266]

[0267]

[0268]

[0269]

[0270]

[0271] discuss:

[0272] Our results suggest that exploring the fundamental properties of cfDNA with fragment-specific analysis may provide more sensitive ctDNA analysis. We based our selection criteria on the biological observation that the distribution of ctDNA fragment sizes is different from that of normal cfDNA. Our work is based on a comprehensive investigation of the fragmentation patterns of plasma cfDNA in 200 patients with a variety of cancer types and 65 healthy individuals. We identified features that can determine the presence and amount of ctDNA in plasma samples without a priori knowledge of somatic aberrations. Although this catalog is the first of its kind, we note that it uses double-stranded DNA from plasma samples and is subject to potential biases caused by the DNA extraction and sequencing methods we used. Additional biological effects may contribute to further selective analysis of cfDNA. Other body fluids (urine, cerebrospinal fluid, saliva), different nucleic acids and structures, changes in the mechanism of release into the circulation, or sample processing methods may display different fragment size characteristics and provide additional exploitable biological patterns for selective sequencing.

[0273] Previous work has reported the size distribution of mutant ctDNA, but only considered limited genomic loci, cancer types, or cases (30, 32, 33). We determined the size difference between mutant DNA and non-mutant DNA on a genome-wide and pan-cancer scale. We developed a method to arrange mutant ctDNA according to size without using high-depth WGS. By sequencing >150 mutations per patient at high depth, we obtained a large number of reads that can be clearly identified as tumor-derived, and thus determined the size distribution of mutant ctDNA and non-mutant cfDNA in cancer patients. A potential limitation of our method is that capture-based sequencing is affected by probe capture efficiency, and therefore our data may not accurately reflect ctDNA fragments <100bp or >300bp.

[0274] Our work provides strong evidence that the modal size of ctDNA in many cancer types is less than 167 bp, which is the length of DNA wrapped around chromosomes. In addition, our work also shows that there is a high level of enrichment of mutant DNA fragments at sizes greater than 167 bp, particularly in the range of 250 to 320 bp. These longer fragments may explain previous observations that longer ctDNA can be detected in the plasma of cancer patients (29, 32). The source of these long fragments remains unknown, and their observation may be related to technical factors. However, it is possible that the mechanism by which cfDNA is compressed and released into the circulation may differ depending on its source, which would be reflected by different fragment sizes (38). Improving the characterization of these fragments will be important, particularly for future work combining ctDNA analysis with other entities in the blood, such as microvesicles and tumor-trained platelets (39, 40). Fragment-specific analysis not only improves the sensitivity of detecting rare mutations, but can also be used to track changes in ctDNA size distribution. Future work should address whether this approach can be used to elucidate the mechanistic effects of treatment on tumor cells, such as by distinguishing necrosis and apoptosis based on fragment size (41).

[0275] Whole genome and exome sequencing of plasma DNA at multiple time points during cancer treatment has been proposed as a non-invasive means to study cancer evolution and to identify possible mechanisms of treatment resistance (3). However, WGS and WES methods are costly and are therefore only applicable to samples with tumor DNA fractions > 5% to 10% so far (3-5, 42). We demonstrate that we can use in vitro and on-machine size selection to exploit differences in fragment length to enrich tumor content in plasma samples, which improves mutation and SCNA detection in sWGS and WES data. We demonstrate that size selection improves detection of mutations present in plasma at low allele fractions while maintaining low sequencing depth by sWGS and WES. Size selection can be implemented in a simple manner and at low cost, and is compatible with a wide range of downstream whole genome and targeted genome analyses, greatly improving the potential value and utility of liquid biopsies.

[0276] Size selection can be applied in silico, which does not incur additional costs, or in vitro, which adds a simple and low-cost intermediate step that can be applied to extracted DNA or libraries created therefrom. This approach, applied prospectively to new studies, can enhance the clinical utility of ctDNA detection and analysis and create opportunities for reanalysis of large amounts of existing data (4, 34, 43). A limitation of this technology is the potential loss of material and information, as some informative fragments may exist within size ranges that are filtered out or not prioritized in the analysis. This can be particularly problematic if only a few copies of the fragment of interest are present in plasma. Despite the potential loss of material, we demonstrate that classification algorithms can learn from cfDNA fragmentation features and SCNA analysis and improve ctDNA detection using inexpensive sequencing methods ( Fig. 22 ). In addition, the cfDNA fragment signature alone can be used to classify cancer and healthy samples with high accuracy (AUC = 0.989 for high ctDNA cancers and AUC = 0.891 for low ctDNA cancers) ( Fig.26 ).

[0277] Analyzing fragment size can provide improvements in other applications. Introducing fragment size information in each read segment can enhance mutation calling algorithms from high-depth sequencing to identify tumor-derived mutations from other sources such as somatic variants or background sequencing noise. In addition, cfDNA analysis of CHIP patients can be structurally different from ctDNA released during tumor cell proliferation (18, 19). Therefore, fragmentation analysis or selective sequencing strategies can be used to distinguish clinically relevant tumor mutations from mutations present in clonal expansion of normal cells. This is crucial for developing cfDNA-based methods for identifying early cancer patients.

[0278] Size selection may also have an impact on the detection of other types of DNA in body fluids, or be used to enrich signals for circulating bacterial or pathogen DNA and mitochondrial DNA. These DNA fragments are not associated with nucleosomes and are typically highly fragmented below 100 bp. Based on the recent connection between the microbiome and treatment efficiency, filtering such fragments may prove important (17, 44). In addition, recent work highlights a stronger correlation between ctDNA detection and cell proliferation than cell death (45). We hypothesize that the distribution pattern of ctDNA fragment size under 145 bp may reflect cfDNA released during cell proliferation, and that fragments under 167 bp may reflect cfDNA released by apoptosis or maturation / turnover of blood cells. The impact of other cancer characteristics (46) on ctDNA biology, structure, concentration, and release is unclear.

[0279] In summary, ctDNA fragment size analysis facilitates non-invasive genomic analysis of tumor DNA through size selection and machine learning methods. Size selection of shorter plasma DNA fragments enriches ctDNA and leads to the identification of a larger number of genomic alterations with both targeted and non-targeted sequencing at minimal additional cost. Combining cfDNA fragment size analysis with SCNA detection with a nonlinear classification algorithm improves the discrimination between samples from cancer patients and healthy individuals. Since the analysis of fragment size is based on the structural properties of ctDNA, size selection can be used for any downstream sequencing application. Our work can help overcome current limitations on the sensitivity of liquid biopsies and support expanded clinical and research applications. Our results suggest that leveraging the endogenous biological properties of cfDNA provides an alternative paradigm for deeper ctDNA sequencing.

[0280] Code

[0281] The following exemplary analysis code for the classification algorithm described in the above embodiment is in the R programming environment (see https: / / www.r-project.org / about.html). Features can be obtained from Table 2, where samples are divided into group A cancer ("high ctDNA cancer") and group B ("low ctDNA cancer"), and healthy controls are used in each (i.e., copies in each file).

[0282]

[0283]

[0284]

[0285]

[0286]

[0287]

[0288]

[0289]

[0290]

[0291] All references cited herein are incorporated by reference in their entirety and for all purposes to the same extent as if each individual publication or patent or patent application was specifically and individually indicated to be incorporated by reference in its entirety.

[0292] The specific embodiments described herein are provided by way of example and not by way of limitation.Any headings included herein are for convenience only and should not be construed as limiting the disclosure in any way.

[0293] References

[0294] 1. G. Siravegna, S. Marsoni, S. Siena, A. Bardelli, Integrating liquid biopsies into the management of cancer, Nat. Rev. Clin. Oncol. (2017), doi: 10.1038 / nrclinonc.2017.14.

[0295] 2.JCMWan, C.Massie, J.Garcia-Corbacho, F.Mouliere, JDBrenton, C.Caldas, S.Pacey, R.Baird, N.Rosenfeld, Liquid biopsies come of age: towards implementation of circulating tumor DNA, Nat.Rev.Cancer 17, 223-238 (2017).

[0296] 3.M.Murtaza, S.-J.Dawson, DWYTsui, D.Gale, T.Forshew, AMPiskorz, C.Parkinson, S.-F.Chin, Z.Kingsbury, ASCWong, F.Marass, S.Humphray, J.Hadfield, D.Bentley, TMChin, JDBrenton, C.Caldas, N.Rosenfeld, Non-invasive analysis ofacquired resistance to cancer therapy by sequencing of plasma DNA, Nature 497, 108-112 (2013).

[0297] 4.V.A.Adalsteinsson,G.Ha,S.S.Freeman,A.D.Choudhury,D.G.Stover,H.A.Parsons,G.Gydush,S.C.Reed,D.Rotem,J.Rhoades,D.Loginov,D.Livitz,D.Rosebrock,I.Leshchiner,J.Kim,C.Stewart,M.Rosenberg,J.M.Francis,C.-Z.Zhang,O.Cohen,C.Oh,H.Ding,P.Polak,M.Lloyd,S.Mahmud,K.Helvie,M.S.Merrill,R.A.Santiago,E.P.O′Connor,S.H.Jeong,R.Leeson,R.M.Barry,J.F.Kramkowski,z.Zhang,L.Polacek,J.G.Lohr,M.Schleicher,E.Lipscomb,A.Saltzman,N.M.Oliver,L.Marini,A.G.Waks,L.C.Harshman,S.M.Tolaney,E.M.Van Allen,E.P.Winer,N.U.Lin,M.Nakabayashi,M.-E.Taplin,C.M.Johannessen,L.A.Garraway,T.R.Golub,J.S.Boehm,N.Wagle,G.Getz,J.C.Love,M.Meyerson,Scalable whole-exome sequencing of cell-freeDNA reveals highconcordance with metastatic tumors,Nat.Commun.8,1324(2017).

[0298] 5.E.Heitzer,P.Ulz,J.Belic,S.Gutschi,F.Quehenberger,K.Fischereder,T.Benezeder,M.Auer,C.Pischler,S.Mannweiler,M.Pichler,F.Eisner,M.Haeusler,S.Riethdorf,K.Pantel,H.Samonigg,G.Hoefle,H.Augustin,J.B.Geigl,M.R.Speicher,Tumor-associated copy number changes in the circulation of patients withprostate cancer identified through whole-genome sequencing,Genome Med.5,30(2013).

[0299] 6.C.Bettegowda,M.Sausen,RJLeary,I.Kinde,Y.Wang,N.Agrawal,BRBartlett,HWang, B.Luber,RMLani,ESAntonarakis,NSAzad,A.Bardelli,H.Brem,JLCameron,CCLee,LAFac her,GLGallia,P.Gibbs,D.Le,RLGiuntoli,M.Goggins,MDHogarty,M.Holdhoff,S.-M.Ho ng,Y.Jiao,HHJuhl,JJKim,G.Siravegna,DALaheru,C.Lauricella,M.Lim,EJLipson,SKNM arie,GJNetto,KSctiner,A.Olivi,L.Olsson,GJRiggins,A.Sartore-Bianchi,K.Schmid tl.-M.Shih,SMOba-Shinjo,S.Siena,D.Theodorescu,J.Tie,TTHarkins,S.Veronese,T. -L.Wang,JDWeingart,CLWolfgang,LDWood,DXing,RHHruban,JWu,PJAllen,CMSchmidt ,MAChoti,VEVelculescu,KWKinzler,B.Vogelstein,N.Papadopoulos,LADiaz,Detection of Circulating Tumor DNA in Early-and Late-Stage Human Malignancies,Sci.Transl.Med.6,224ra24-224ra24(2014).

[0300] 7.F.Diehl,M.Li,D.Dressman,Y.He,D.Shen,S.Szabo,L.A.Diaz,S.N.Goodman,K.A.David,H.Juhl,K.W.Kinzler,B.Vogelstein,Detection and quantification ofmutations in the plasma of patients with colorectal tumors,Proc.Natl.Acad.Sci.102,16368-16373(2005).

[0301] 8.S.-J.Dawson,D.W.Y.Tsui,M.Murtaza,H.Biggs,O.M.Rueda,S.-F.Chin,M.J.Dunning,D.Gale,T.Forshew,B.Mahler-Araujo,S.Rajah,S.Humphray,J.Becq,D.Halsall,M.Wallis,D.Bentley,C.Caldas,N.Rosenfeld,Analysis of CirculatingTumor DNA to Monitor Metastatic Breast Cancer,N.Engl.J.Med.368,1199-1209(2013).

[0302] 9.F.Diehl,K.Schmidt,M.A.Choti,K.Romans,S.Goodman,M.Li,K.Thornton,N.Agrawal,L.Sokoll,S.A.Szabo,K.W.Kinzler,B.Vogelstein,L.A.Diaz,Circulatingmutant DNA to assess tumor dynamics.,Nat.Med.14,985-90(2008).

[0303] 10.J.Tie,Y.Wang,C.Tomasetti,L.Li,S.Springer,I.Kinde,N.Silliman,M.Tacey,H.-L.Wong,M.Christie,S.Kosmider,I.Skinner,R.Wong,M.Steel,B.Tran,J.Desai,I.Jones,A.Haydon,T.Hayes,T.J.Price,R.L.Strausberg,L.A.Diaz,N.Papadopoulos,K.W.Kinzler,B.Vogelstein,P.Gibbs,Circulating tumor DNAanalysis detects minimal residual disease and predicts recurrence in patientswith stage II colon cancer.,Sci.Transl.Med.8,346ra92(2016).

[0304] 11.A.A.Chaudhuri,J.J.Chabon,A.F.Lovejoy,A.M.Newman,H.Stehr,T.D.Azad,M.S.Khodadoust,M.S.Esfahani,C.L.Liu,L.Zhou,F.Scherer,D.M.Kurtz,C.Say,J.N.Carter,D.J.Merriott,J.C.Dudley,M.S.Binkley,L.Modlin,S.K.Padda,M.F.Gensheimer,R.B.West,J.B.Shrager,J.W.Neal,H.A.wakelee,B.W.Loo,A.A.Alizadeh,M.Diehn,Early Detection of Molecular Residual Disease inLocalized Lung Cancer by Circulating Tumor DNA Profiling.,Cancer Discov.7,1394-1403(2017).

[0305] 12.JDCohen,L.Li,Y.Wang,C.Thoburn,B.Afsari,L.Danilova,C.Douville,AAJaved,F.Wong,A.Mattox,RHHruban,CLWolfgang,MGGoggins,M.Dal Molin,T.-L.Wang,R.Roden,APKlein,J.Ptak,L.Dobbyn,J.Schaefer,N.Silliman, M.Popoli,JTVogelstein,JDBrowne,RESchoen,REBrand,J.Tie,P.Gibbs,H.-L.Wong ,ASMansfield,J.Jen,SMHanash,M.Falconi,PJAllen,S.Zhou,C.Bettegowda,LADiaz,C.Tomasetti,KWKinzler,B.Vogelstein,AMLennon,N.Papadopoulos,Detection and localization of surgically resectable cancers with a mu1ti-analyte blood test.,Science 359,926-930(2018).

[0306] 13.ISHaque,O.Element,Challenges in Using ctDNA to Achieve Early Detection of Cancer,bioRxiv,237578(2017).

[0307] 14.A.M.Newman,A.F.Lovejoy,D.M.Klass,D.M.Kurtz,J.J.Chabon,F.Scherer,H.Stehr,C.L.Liu,S.V Bratman,C.Say,L.Zhou,J.N.Carter,R.B.West,G.W.Sledge Jr,J.B.Shrager,B.W.Loo,J.W.Neal,H.A.Wakelee,M.Diehn,A.A.Alizadeh,Integrateddigital error suppression for improved detection of circulating tumor DNA,Nat.Biotechnoi.34,547-555(2016).

[0308] 15.P.Uiz,G.G.Thallinger,M.Auer,R.Graf,K.Kashofer,S.W.Jahn,L.Abete,G.Pristauz,E.Petru,J.B.Geigl,E.Heitzer,M.R.Speicher,Inferring expressed genesby whole-genome sequencing of plasma DNA,Nat.Genet.48,1273-1278(2016).

[0309] 16.M.W.Snyder,M.Kircher,A.J.Hill,R.M.Daza,J.Shendure,Cell-free DNAComprises an In Vivo Nucleosome Footprint that Informs Its Tissues-Of-Origin.,Cell 164,57-68(2016).

[0310] 17.P.Burnham,M.S.Kim,S.Agbor-Enoh,H.Luikart,H.A.Valantine,K.K.Khush,I.De Vlaminck,Single-stranded DNA library preparation uncovers the origin anddiversity of ultrashort cell-free DNA in plasma,Sci.Rep.6,27859(2016).

[0311] 18.G.Genovese,AK ,REHandseker,J.Lindberg,SARose,SFBakhoum,K.Chambert,E.Mick,BMNeale,M.Fromer,SMPurcell,O.Svantesson,M.Landen,M. ,S.Lehmann,SBGabriel,JLMoran,ESLander,PFSullivan,P.Sklar,H. ,CMHultman,SAMcCarroll,Clonal Hematopoiesis and Blood-Cancer Risk Inferred from Blood DNA Sequence,N.Engl.J.Med.371,2477-2487(2014).

[0312] 19.Y.Hu,B.Ulrich,J.Supplee,Y.Kuang,PHLizotte,N.Feeney,N.Guibert,MMAwad,K.-K.Wong,PAJanne,CPPaweletz,GROxnard,False positive plasma genotyping due to clonal hematopoiesis.,Clin.Cancer Res.,clincanres.0143.2018(2018).

[0313] 20.AJBronkhorst,JFWentzel,J.Aucamp,E.van Dyk,L.du Plessis,PJPretorius,Characterization of the cell-free DNA released by culturedcancer cells,Biochim.Biophys.Acta-Mol.Cell Res.1863,157–165(2016).

[0314] 21.S.Jahr,H.Hentze,S.Englisch,D.Hardt,F.O.Fackelmayer,R.D.Hesch,R.Knippers,DNA fragments in the blood plasma of cancer patients:quantitationsand evidence for their origin from apoptotic and necrotic cells.,CancerRes.61,1659-65(2001).

[0315] 22.Y.M.D.Lo,K.C.A.Chan,H.Sun,E.Z.Chen,P.Jiang,F.M.F.Lun,Y.W.zheng,T.Y.Leung,T.K.Lau,C.R.Cantor,R.W.K.Chiu,Maternal plasma DNA sequencingreveals the genome-wide genetic and mutational profile of the fetus.,Sci.Transl.Med.2,61ra91(2010).

[0316] 23.D.Chandrananda,N.P.Thorne,M.Bahlo,L.-S.Tam,G.Liao,E.Li,High-resolution characterization of sequence signatures due to non-random cleavageof cell-free DNA,BMC Med.Genomics 8,29(2015).

[0317] 24.P.Jiang,Y.M.D.Lo,The Long and Short of Circulating Cell-Free DNAand the Ins and Outs of Molecular Diagnostics,Trends Genet.32,360-371(2016).

[0318] 25.SCYYu,KCAChan,YWLZheng,P.Jiang,GJWLiao,H.Sun,R.Akolekar,TYLeung,ATJIGo,JMGvan Vugt,R.Minekawa,CBMOudejans,KHNicolaides,RWKChiu,YMDLo,Size-based molecular diagnostics using plasma DNA for noninvasive prenatal testing.,Proc.Natl.Acad.Sci.USA111,8583-8(2014).

[0319] 26.FMFLun,NBYTsui,KCAChan,TYLeung,TKLau,P.Charoenkwan,KCKChow,WYWLo,C.Wanapirak,T.Sanguansermsri,CRCantor,RWKChiu,YMDLo,Noninvasive prenatal diagnosis of Proc.Natl.Acad.Sci.USA105,19920-5(2008) Relative mutation dosage on DNA in maternal plasma.

[0320] 27.G.Minarik,G.Repiska,M.Hyblova,E.Nagyova,K.Soltys,J.Budis,F.Duris,R.Sysak,M.Gerykova Bujalkova,B.Vlkova-Izrael,O.Biro,B.Nagy,T.Szemes,Utilization of Benchtop Next Generation Sequencing Platforms Ion Torrent PGMand MiSeq in Noninvasive Prenatal Testing for Chromosome 21 Trisomy andTesting of Impact of In Silico and Physical Size Selection on Its AnalyticalPerformance.,PLoS One 10,e0144811(2015).

[0321] 28.M.B.Giacona,G.C.Ruben,K.A.Iczkowski,T.B.Roos,D.M.Porter,G.D.Sorenson,Cell-Free DNA in Human Blood Plasma,Pancreas 17,89-97(1998).

[0322] 29.N.Umetani,A.E.Giuliano,S.H.Hiramatsu,F.Amersi,T.Nakagawa,S.Martino,D.S.B.Hoon,Prediction of breast tumor progression by integrity offree circulating DNA in serum.,J.Clin.Oncol.24,4270-6(2006).

[0323] 30.F.Mouliere,B.Robert,E.Arnau Peyrotte,M.Del Rio,M.Ychou,F.Molina,C.Gongora,A.R.Thierry,T.Lee,Ed.High Fragmentation Characterizes Tumour-Derived Circulating DNA,PLoS One 6,e23418(2011).

[0324] 31.F.Mouliere,S.El Messaoudi,D.Pang,A.Dritschilo,A.R.Thierry,Multi-marker analysis of circulating cell-free DNA toward personalized medicine forcolorectal cancer,Mol.Oncol.8,927-941(2014).

[0325] 32.P.Jiang,C.W.M.Chan,K.C.A.Chan,S.H.Cheng,J.Wong,V.W.-S.Wong,G.L.H.Wong,S.L.Chan,T.S.K.Mok,H.L.Y.Chan,P.B.S.Lai,R.W.K.Chiu,Y.M.D.Lo,Lengthening and shortening of plasma DNA in hepatocellular carcinomapatients.,Proc.Natl.Acad.Sci.U.S.A.112,E1317-25(2015).

[0326] 33.H.R.Underhill,J.O.Kitzman,S.Hellwig,N.C.Welker,R.Daza,D.N.Baker,K.M.Gligorich,R.C.Rostomily,M.P.Bronner,J.Shendure,D.JKwiatkowski,Ed.FragmentLength of Circulating Tumor DNA,PLOS Genet.12,e1006162(2016).

[0327] 34.O.A.Zill,K.C.Banks,S.R.Fairclough,S.A.Mortimer,J.VVowles,R.Mokhtari,D.R.Gandara,P.C.Mack,J.I.Odegaard,R.J.Nagy,A.M.Baca,H.Eltoukhy,D.I,Chudova,R.B.Lanman,A,Talasaz,The Landscape of Actionable GenomicAlterations in Cell-Free Circulating Tumor DNA from 21,807 Advanced CancerPatients.,Clin.Cancer Res.,clincanres.3837.2017(2018).

[0328] 35.G.Macintyre,T.E.Gorancva,D.De Silva,D.Ennis,A.M.Piskorz,M.Eldridge,D.Sie,L.-A.Lewsley,A.Hanif,C.Wilson,S.Dowson,R.M.Glasspool,M.Lockley,E.Brockbank,A.Montes,A.Walther,S.Sundar,R.Edmondson,G.D.Hall,A.Clamp,C.Gourley,M.Hall,C.Fotopoulou,H.Gabra,J.Paul,A.Supernat,D.Millan,A.Hoyle,G.Bryson,C.Nourse,L.Mincarelli,L.N.Sanchez,B.Ylstra,M.Jimenez-Linan,L.Moore,O.Hofmann,F.Markowetz,I.A.McNeish,J.D.Brenton,Copy number signaturesand mutational processes in ovarian carcinoma,Nat.Genet.,1(2018).

[0329] 36.C.A.Parkinson,D.Gale,A.M.Piskorz,H.Biggs,C.Hodgkin,H.Addley,S.Freeman,P.Moyle,E.Sala,K.Sayal,K.Hosking,I.Gounaris,M.Jimenez-Linan,H.M.Earl,W.Qian,N.Rosenfeld,J.D.Brenton,E.R.Mardis,Ed.Exploratory Analysis ofTP53 Mutations in Circulating Tumour DNA as Biomarkers of Treatment Responsefor Patients with Relapsed High-Grade Serous Ovarian Carcinoma:ARetrospective Study,PLQS Med.13,e1002198(2016).

[0330] 37.T.Forshew,M.Murtaza,C.Parkinson,D.Gale,D.W.Y.Tsui,F.Kaper,S.-J.Dawson,A.M.Piskorz,M.Jimenez-Linan,D.Bentiey,J.Hadfield,A.P.May,C.Caldas,J.D.Brenton,N.Rosenfeld,Noninvasive identification and monitoring of cancermutations by targeted deep sequencing of plasma DNA.,Sci.Transl.Med.4,136ra68(2012).

[0331] 38.A.R.Thierry,S.El Messaoudi,P.B.Gahan,P.Anker,M.Stroun,Origins,structures,and functions of circulatingDNA in oncology,Cancer MetastasisRev.35,347-376(2016).

[0332] 39.M.G.Best,N.Sol,B.A.Tannous,P.Weaseling,T.Wurdinger,RNA-Seq ofTumor-Educated Platelets Enables Blood-Based Pan-Cancer,Multiclass,andMolecular Pathway Cancer Diagnostics,Cancer Cell 28,666-676(2015).

[0333] 40.MGBest,N.Sol,SGJGIn′t Veld,A.Vancura,M.Muller,A.-LNNiemeijer,AV Fejes,L.-A.Tjon Kon Fat,AEHuis In′t Veld,C.Leurs,TYLeLarge,LLMeijer,IEKooi,F.Rustenburg,P.Schellen,H.verschueren,E.Post,LEWedekind,J.Bracht,M.Esenkbrink,L.Wils. ,F.Favaro,JDSchoonhoven,J.Tannous,H.Meijers-Heijboer,G.Kazemier,E.Giovannetti,JCReijneveld,S.Idema,J.Killestein,M.Heger,SCde Jager,RTUrbanus,IEHoefer,G.Pasterkamp,C.Mannhalter,J.Gomez-Arroyo,H.-J.Bogaard,DPNoske,WPVandertop,D.van denBroek,B.Ylstra,RJANilsson,P.Wesseling,N.Karachaliou,R.Rosell,E.Lee-Lewandrowski,KBLewandrowski,BATannous,AJde Langen,EFSmit,MMvan,Werding,Werding. Intelligence-Enhanced Detection of Non-Small-CellLung Cancer Using Tumor-Educated Platelets.,Cancer Cell 32,238-252.e9(2017).

[0334] 41.A.L.Riediger,S.Dietz,U.Schirmer,M.Meister,I.Heinzmann-Groth,M.Schneider,T.Muley,M.Thomas,H.Sültmann,Mutation analysis of circulatingplasma DNA to determine response to EGFR tyrosine kinase inhibitor therapy oflung adenocarcinoma patients,Sci.Rep.6,33505(2016).

[0335] 42.J.Belic,M.Koch,P.Ulz,M.Auer,T.Gerhalter,S.Moban,K.Fischereder,E.Petru,T.Bauernhofer,J.B.Geigl,M.R.Speicher,E.Heitzer,Rapid Identificationof Plasma DNA Samples with Increased ctDNA Levels by a Modified FAST-SeqSApproach,Clin.Chem.61,838-849(2015).

[0336] 43.D.G.Stover,H.A.Parsons,G.Ha,S.S.Freeman,W.T.Barry,H.Guo,A.D.Choudhury,G.Gydush,S.C.Reed,J.Rhoades,D.Rotem,M.E.Hughes,D.A.Dillon,A.H.Partridge,N.Wagle,I.E.Krop,G.Getz,T.R.Golub,J.C.Love,E.P.Winer,S.M.Tolaney,N.U.Lin,V.A.Adalsteinsson,Association of Cell-Free DNA TumorFraction and Somatic Copy Number Alterations With Survival in MetastaticTriple-Negative Breast Cancer.,J.Clin.Oncol.36,543-553(2018).

[0337] 44.B.Routy,E.Le Chatelier,L.Derosa,CPMDuong,MTAlou,R.Daillère,A.Fluckiger,M.Messaoudene,C.Rauber,MPRoberti,M.Fidelle,C.Flament,V.Poirier-Co blades,P.Opolon,C.Klein,K.Iribarren,L.Mondragón,N.Jacquelot,B.Qu,G.Ferrere,C.Clémenson,L.Mezquita,JRMasip,C.Naltet,S.Brosseau ,C.Kaderbhai, C.Richard, N.Rizvi, F.Levenez, N.Galleron, B.Quinquis, N.Pons, B.Ryffel, V.Minard-Colin, P.Gonin, J.-C.Soria, E.Deutsch, Y.Loriot, F.Ghiringhelli, G.zalcman, F.Goldwasser, B.Escudier, M.D.Hellmann, A.Eggermont, D.Raoult, L.Albiges, G.Kroemer, L.zitvogel,Gut microbiome influences efficacy of PD-1-based immunotherapy against epithelial tumors.,Science 359, 91-97(2018).

[0338] 45.C.Abbosh, NJBirkbak, G.A.Wilson, M.Jamal-Hanjani, T.Constantin, R.Salari, J.Le Quesne, D.A.Moore, S.Veeriah, R.Rosenthal, T.Marafioti, E.Kirkizlar, T.B.K.Watkins, N.McGranahan, S.Ward, L.Martinson, J.Riley, F.Fraioli, M.Al Bakir, E. ,F.Zambrana,R.Endozo,WLBi,FMFennessy,N.Sponer,D.Johnson,J.Laycock,S.Shafi,J.Czyzews ka-Khan,A.Rowan,T.Chambers,N.Matthews,S.Turajlic,C.Hiley,SMLee,MDForster,T.Ahmad,M.F alzon,E.Borg,D.Lawrence,M.Hayward,S.Kolvekar,N.Panagiotopoulos,SMJanes,R.Thakrar,A.A hmed,F.Blackhall,Y.Summers,D.Hafez,A.Naik,A.Ganguly,S.Kareht,R.Shah,L.Joseph,A.Marie Quinn,PACrosbie,B.Naidu,G.Middleton,G.Langman,S.Trotter,M.Nicolson,H.Remmen,K.Kerr,M.Cherty,L.Gom ersall,DAFennell,A.Nakas,S.Rathinam,G.Anand,S.Khan,P.Russell,V.Ezhil,B.Ismail,M.Irwin-Sellers,V.P rakash,JFLester,M.Kornaszewska,R.Attanoos,H.Adams,H.Davies,D.Oukrif,AUAkarca,JAHartley,HLowe,S.L ock,N.Iles,H.Bell,Y.Ngai,G.Elgar,Z.Szailasi,RFSchwarz,J.Herrero,A.Stewart,SAQuezada,KSPeggs,P.Van Loo,C.Dive,CJLin,M.Rabinowitz,HJWLAerts,A.Hackshaw,JAShaw,BGzimmermann,TRACERxconsortium,PEACE consortium,C.Swanton,Phylogenetic ctDNA analysis depicting lung cancer evolution. 545,446-451(2017).

[0339] 46.D.Hanahan,R.A.Weinberg,Hallmarks of cancer:the next generation.,Cell 144,646-74(2011).

[0340] 47.K.M.Patel,K.E.van der Vos,C.G.Smith,F.Mouliere,D.Tsui,J.Morris,D.Chandrananda,F.Marass,D.van den Broek,D.E.Neal,V.J.Gnanapragasam,T.Forshew,B.W.van Rhijn,C.E.Massie,N.Rosenfeld,M.S.van der Heijden,Association OfPlasma And Urinary Mutant DNA With Clinical Outcomes In Muscle InvasiveBladder Cancer,Sci.Rep.7,5554(2017).

[0341] 48.H.Li,R.Durbin,Fast and accurate short read alignment with Burrows-Wheeler transform,Bioinformatics 25,1754-1760(2009).

[0342] 49.I.Scheinin,D.Sie,H.Bengtsson,M.A.van de Wiel,A.B.Olshen,H.F.vanThuijl,H.F.van Essen,P.P.Eijk,F.Rustenburg,G.A.Meijer,J.C.Reijneveld,P.Wesseling,D.Pinkel,D.G.Albertson,B.Ylstra,DNA copy number analysis of freshand formalin-fixed specimens by shallow whole-genome sequencing withidentification and exclusion of problematic regions in the genome assembly,Genome Res.24,2022-2032(2014).

Claims

1. A computer-implemented method for detecting variant nucleic acids from a sample comprising cell-free nucleic acid, comprising performing, by at least one processor, operations comprising: a) providing data representing the fragment sizes of nucleic acid fragments obtained from said sample; b) processing the data from step a) according to a classification algorithm, the classification algorithm having been trained on a training set comprising a plurality of cell-free nucleic acid samples containing the variant nucleic acid and a plurality of samples not containing the variant nucleic acid, wherein the variant DNA is circulating tumor DNA, wherein running of the classification algorithm classifies the sample data into one of at least two categories, the at least two categories comprising a first category containing the variant nucleic acid and a second category not containing the variant nucleic acid, wherein running of the classification algorithm classifies the sample data into one of the at least two categories based at least on a plurality of cfDNA fragment size features selected from the group consisting of: (i) the proportion of fragments with a size range of 20 to 150 bp; (ii) the proportion of fragments with a size range of 100 to 150 bp; (iii) the proportion of fragments in the size range of 160 to 180 bp; (iv) the proportion of fragments in the size range of 180 to 220 bp; (v) proportion of fragments with a size range of 250 to 320 bp; (vi) the ratio of the proportion of fragments in the size range of 20 to 150 bp divided by the proportion of fragments in the size range of 160 to 180 bp; (vii) the ratio of the proportion of fragments in the size range of 100 to 150 bp divided by the proportion of fragments in the size range of 163 to 169 bp; (viii) the ratio of the proportion of fragments in the size range of 20 to 150 bp divided by the proportion of fragments in the size range of 180 to 220 bp; and (ix) oscillation amplitude in fragment size density with a periodicity of 10 bp, And wherein the data representing the fragment sizes of the cfDNA fragments in step a) comprises the cfDNA fragment size features used by the classification algorithm; and c) outputting the classification of the sample from step b), and thereby determining whether the sample contains the variant nucleic acid, or determining the probability that the sample contains the variant nucleic acid.

2. The method of claim 1, wherein the at least one processor further performs operations including providing data representing a measure of deviation from copy number neutrality in nucleic acid fragments obtained from the sample, and wherein operation of the classification algorithm classifies the sample data based on a plurality of cell-free nucleic acid fragment size characteristics and characteristics of deviation from copy number neutrality.

3. The computer-implemented method of claim 1 or claim 2, wherein the data representing the fragment size of the nucleic acid fragment comprises the fragment size inferred from the sequence reads, the fragment size determined by fluorescence or the fragment size determined by optical density.

4. The computer-implemented method of claim 1 or claim 2, wherein the multiple cfDNA fragment size characteristics include: a fragment ratio in the size range of 160 to 180 bp, a fragment ratio in the size range of 180 to 220 bp, a fragment ratio in the size range of 250 to 320 bp, and an amplitude oscillation with a periodicity of 10 bp in the fragment size density.

5. The computer-implemented method of claim 1 or claim 2, wherein after aligning the sequence reads to a reference genome of the species from which the sample was obtained, the fragment sizes of the cfDNA fragments are inferred from the sequence reads using the mapping positions of the ends of the reads in the genome.

6. A computer-implemented method as described in claim 1 or claim 2, wherein the operation of the classification algorithm classifies the sample data into one of the at least two categories based at least on the following deviation feature from the copy number neutral point, wherein the deviation feature from the copy number neutral point is a trimmed median absolute deviation score from the copy number neutral point or an ichorCNA score.

7. The computer-implemented method of claim 6, wherein the t-MAD score is determined by trimming genomic regions showing high copy number variability in a whole genome dataset derived from healthy subjects, and then calculating the median absolute deviation from log2R=0 in non-trimmed regions of the genome.

8. The computer-implemented method of claim 1 or claim 2, wherein the classification algorithm performs a random forest (RF) analysis, a logistic regression analysis, or a support vector machine (SVM) analysis.

9. The computer-implemented method of claim 1 or claim 2, wherein the classification algorithm has been trained on a training set comprising at least 10 samples from healthy subjects and 10 samples from subjects known to have cancer.

10. The computer-implemented method of claim 9, wherein the training set comprises at least 10 healthy samples and 10 cancer samples shown in Table 2.

11. The computer-implemented method of claim 1 or claim 2, wherein the sequence data provided in step a) represents whole genome sequence reads, custom set sequencing sequence reads, labeled amplicon deep sequencing reads, hybrid capture sequence reads, focused exome sequence reads, or whole exome sequence reads.

12. The computer-implemented method of claim 11, wherein the sequence data provided in step a) represents shallow whole-genome sequence reads.

13. The computer-implemented method of claim 1 or claim 2, wherein the data provided in step a) represents fragment sizes of a plurality of DNA fragments from a substantially cell-free fluid sample from a subject having or suspected of having cancer.

14. The computer-implemented method of claim 13, wherein the cancer is selected from the group consisting of melanoma, lung cancer, bile duct cancer, bladder cancer, esophageal cancer, colorectal cancer, ovarian cancer, glioma, pancreatic cancer, kidney cancer, and breast cancer.

15. The computer-implemented method of claim 1 or claim 2, wherein the sample is a plasma sample, a urine sample, a saliva sample, a cerebrospinal fluid sample, a serum sample, or other biological fluid sample containing DNA.

16. The computer-implemented method of claim 1 or claim 2, wherein the method is used to detect the presence, growth, prognosis, regression, treatment response, or recurrence of cancer in a subject from which the sample was obtained.

17. The computer-implemented method of claim 16, wherein the presence of ctDNA in the sample is distinguished from cfDNA containing somatic mutations of non-cancerous origin.

18. The computer-implemented method of claim 17, wherein the non-cancerous source comprises a clonal expansion of a normal epithelial cell or a clonal hematopoietic cell of indeterminate potential.

19. The computer-implemented method of claim 18, wherein the fragment size data provided in step a) represents sequence reads of a plurality of DNA fragments from a substantially cell-free fluid sample of a subject, and wherein the method is used to determine whether the sample contains ctDNA or cfDNA from CHIP.

20. The computer-implemented method of claim 19, wherein the classification algorithm has been trained on a training set further comprising a plurality of cfDNA samples obtained from subjects with CHIP, and wherein the at least two classes further comprise a third class containing cfDNA derived from CHIP based on a plurality of cfDNA fragment size features and / or deviation from copy number neutrality features.

21. The computer-implemented method of claim 1 or claim 2, further comprising the at least one processor to perform operations comprising: Receiving data comprising a plurality of sequence reads obtained by sequencing a cfDNA-containing sample obtained from a subject or a library derived from a cfDNA-containing sample; Sequence reads obtained by sequencing the sample or a library derived from the sample are processed to determine sequence data representing fragment sizes of cfDNA fragments obtained from the sample and / or sequence data representing a measure of deviation from a copy number neutral point in the cfDNA fragments obtained from the sample.

22. The computer-implemented method of claim 21, wherein processing the sequence reads comprises one or more of the following steps: aligning the sequence reads to a reference genome of the same species as the subject; Removal of contaminating adapter sequences; PCR and optical duplicates were removed; Sequence reads with low mapping quality were removed; as well as In case of multiplex sequencing, signals are separated by excluding mismatches in the sequencing barcodes.

23. A computer-implemented method according to claim 21, wherein processing the sequence reads to determine sequence data representing a measure of deviation from a copy number neutral point in the cfDNA fragments obtained from the sample includes determining a trimmed median absolute deviation score or ichorCNA score from a copy number neutral point.

24. A computer-implemented method according to claim 23, wherein the t-MAD score is determined by trimming genomic regions showing high copy number variability in a whole genome dataset derived from healthy subjects, and then calculating the median absolute deviation from log2R=0 in non-trimmed regions of the genome.

25. The computer-implemented method of claim 18, wherein somatic mutations comprising cfDNA fragments are classified as tumor-derived or CHIP-derived based on a plurality of fragment size features determined from the sequence reads.

26. The computer-implemented method of claim 1 or claim 2, wherein the classification of the sample as containing ctDNA or not containing tDNA, or the determined probability that the sample contains ctDNA, is used to predict whether the sample or another sample from the same subject will be susceptible to further ctDNA analysis.

27. The computer-implemented method of claim 26, wherein the further ctDNA analysis comprises sequencing at a greater sequencing depth and / or performing targeted sequencing of the ctDNA in the sample.

28. The computer-implemented method of claim 26, wherein the probability that the sample comprises ctDNA as determined by the classification algorithm is at least 0.5, and wherein the further ctDNA analysis is performed on the sample.

29. The computer-implemented method of claim 26, wherein: The sample is a plasma sample, and wherein the probability that the sample comprises ctDNA as determined by the classification algorithm is used to determine whether ctDNA will be detectable in a urine sample; or The sample is a urine sample, and wherein the probability that the sample comprises ctDNA as determined by the classification algorithm is used to determine whether ctDNA will be detectable in the plasma sample.

30. A computer program product comprising instructions which, when executed by at least one processor, cause the at least one processor to perform the method according to any one of claims 1 to 29.

Images