Assay plate for amplifying cfDNA molecules associated with cancer

Abstract

The present invention provides an assay plate for amplifying cfDNA molecules associated with cancer. Various methods of designing, making, and using the cancer assay plate for diagnosing cancer are also provided.

Description

[0001] This application is a divisional application of application number 201980037495.6 (PCT application number is PCT / US2019 / 025358), application date is April 2, 2019, and invention name is “Methylation marker and target methylation probe board”.

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS

[0003] This application claims the benefit of U.S. Provisional Patent Application No. 62 / 651,643, filed April 2, 2018, and U.S. Provisional Patent Application No. 62 / 738,271, filed September 28, 2018, which are hereby incorporated by reference in their entireties. Background Art

[0004] Deoxyribonucleic acid (DNA) methylation plays an important role in regulating gene expression. Abnormal DNA methylation is relevant to many disease processes, including cancer. DNA methylation analysis using methylation sequencing (e.g., whole genome bisulfite sequencing (WGBS)) is increasingly considered to be a valuable diagnostic tool for detecting, diagnosing and / or monitoring cancer. For example, the specific pattern of different methylation regions can be used as a molecular marker for various diseases.

[0005] However, WGBS is not ideally suited for product assays because the vast majority of the genome is not differentially methylated in cancer or has local CpG density too low to provide a reliable signal. Only a few percent of the genome may be useful for classification.

[0006] Furthermore, there are various challenges in identifying multiple differentially methylated regions across various diseases. First, determining differentially methylated regions in a disease cohort is only weighted when compared to a cohort of multiple control subjects; therefore, if the control group is small, the determination will lose confidence in the smaller control group. Additionally, methylation status can vary across a cohort of control subjects, making it difficult to interpret when determining whether multiple regions are differentially methylated across a disease cohort. Furthermore, cytosine methylation at one CpG site is strongly correlated with methylation at a subsequent CpG site. Generalizing this dependency is a challenge in itself.

[0007] Therefore, there is no cost-effective method that can accurately diagnose diseases by detecting multiple differentially methylated regions. Summary of the Invention

[0008] Early detection of cancer in many subjects is important because it allows for early treatment, resulting in a greater chance of survival. Targeted detection of methylation patterns specific to a cancer or tissue of origin (i.e., the organ, organ group, body region, or cell type where the cancer develops or originates) using cell-free DNA (cfDNA) fragments can enable early detection of cancer by providing a cost-effective, non-invasive method for analyzing information related to cancer classification. Compared to whole genome sequencing (WGS) or whole genome bisulfite sequencing (WGBS), this method can increase the sequencing depth of the target region and reduce costs by using a panel of targeted genomic regions rather than sequencing all nucleic acids in a test sample (also known as "whole genome sequencing").

[0009] To this end, the present disclosure provides a variety of cancer assay panels (e.g., a variety of bait sets) for detecting cancer and various tissues or sources by detecting methylation patterns of multiple target genomic regions. The cancer assay panels can detect and differentiate methylation patterns of general cancers or different cancer types, for example, (1) blood cancer, (2) breast cancer, (3) colorectal cancer, (4) esophageal cancer, (5) head and neck cancer, (6) hepatobiliary cancer, (7) lung cancer, (8) ovarian cancer, and (9) pancreatic cancer.

[0010] The multiple cancer assay panels can further provide information about a cancer stage for each cancer type. The present description also provides a method for diagnosing cancer using a multiple cancer assay panel, wherein the diagnosing cancer further comprises a cancer type and / or a cancer stage. Further provided herein are multiple methods for identifying multiple genomic sites having methylation patterns that are specific to cancer or various cancer types, and a list of multiple genomic sites that can be used for the diagnosis of cancer and / or cancer-origin tissue. The multiple methods described herein also include methods for designing multiple probes to effectively amplify nucleic acids derived from selected multiple genomic regions without removing excess undesired or non-target nucleic acids, and methods for making cancer assay panels using the multiple probes. Methods for amplifying nucleic acids derived from selected multiple genomic regions by methods other than hybridization capture are also described.

[0011] The present invention discloses a variety of assay panels for amplifying multiple cfDNA molecules for cancer diagnosis, the assay panels comprising:

[0012] at least 500 different pairs of polynucleotide probes, wherein each of the at least 500 pairs of probes: (i) comprises two different probes configured to overlap with each other through an overlapping sequence of 30 or more nucleotides, and (ii) is configured to hybridize to a modified fragment obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions,

[0013] Each of the one or more genomic regions comprises at least five methylation sites and has an abnormal methylation pattern in a plurality of cancerous training samples.

[0014] In some embodiments, the overlapping sequence comprises at least 40, 50, 75, or 100 nucleotides.

[0015] In some embodiments, the plurality of assay plates comprises at least 50, 60, 70, 80, 90, 100, 120, 150, 200, 300, or 400 pairs of probes.

[0016] In some embodiments, the plurality of assay plates comprises at least 1000, 2000, 2500, 5000, 6000, 7500, 10000, 15000, 20000, or 25000 pairs of probes.

[0017] The present invention also discloses a variety of assay panels for amplifying multiple cfDNA molecules for cancer diagnosis, the assay panels comprising:

[0018] at least 1000 polynucleotide probes, wherein each of the at least 1000 probes is configured to hybridize to a modified polynucleotide obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions,

[0019] Each of the one or more genomic regions comprises at least five methylation sites and has an abnormal methylation pattern in a plurality of cancerous training samples.

[0020] In some embodiments, the processing of the plurality of cfDNA molecules comprises converting unmethylated C (cytosine) in the plurality of cfDNA molecules to U (uracil).

[0021] In some embodiments, each of the plurality of polynucleotide probes in the assay plate is conjugated to an affinity moiety.

[0022] In some embodiments, the affinity moiety is a biotin moiety.

[0023] In some embodiments, the plurality of training samples are a plurality of samples derived from a plurality of subjects determined to have cancer.

[0024] In some embodiments, when a methylation state vector representing the genomic region in the cancerous training samples appears at a frequency lower than a threshold in the reference samples, a genomic region in the cancerous training samples has an abnormal methylation pattern.

[0025] In some embodiments, the threshold is 0.1, 0.01, 0.001, or 0.0001.

[0026] In some embodiments, each of the one or more genomic regions is hypermethylated or hypomethylated in the plurality of cancerous training samples.

[0027] In some embodiments, at least 80%, 85%, 90%, 92%, 95%, or 98% of the at least 5 methylation sites are methylated or unmethylated in the plurality of cancerous training samples.

[0028] In some embodiments, at least 3%, 5%, 10%, 15%, 20%, 30%, or 40% of the plurality of probes on the assay plate do not contain G (guanine).

[0029] In some embodiments, at least 80%, 85%, 90%, 92%, 95%, 98% of the plurality of probes on the assay plate have only CpG or only CpA at the plurality of CpG detection sites.

[0030] In some embodiments, each of the plurality of probes on the assay plate comprises fewer than 20, 15, 10, 8, or 6 CpG detection sites.

[0031] In some embodiments, each of the plurality of probes on the assay plate is designed to have sequence homology or sequence complementarity to fewer than 20, 15, 10, or 8 off-target genomic regions.

[0032] In some embodiments, the fewer than 20 off-target genomic regions are identified using a k-mer seeding strategy.

[0033] In some embodiments, the fewer than 20 off-target genomic regions are identified using a k-mer seeding strategy in combination with local alignment at multiple seeding positions.

[0034] In some embodiments, the plurality of assay plates comprises at least 1,000, 2,000, 2,500, 5,000, 10,000, 12,000, 15,000, 20,000, or 25,000 probes.

[0035] In some embodiments, the at least 1,000 or 500 pairs of probes or the at least 1,000 probes simultaneously comprise at least 200,000, 400,000, 600,000, 800,000, 1 million, 2 million, or 4 million nucleotides.

[0036] In some embodiments, each of the plurality of probes on the assay plate comprises at least 50, 75, 100, or 120 nucleotides.

[0037] In some embodiments, each of the plurality of probes on the assay plate comprises fewer than 300, 250, 200, or 150 nucleotides.

[0038] In some embodiments, each of the plurality of probes on the assay plate comprises 100 to 150 nucleotides.

[0039] In some embodiments, at least 30% of the plurality of genomic regions are exons or introns.

[0040] In some embodiments, at least 15% of the plurality of genomic regions are exons.

[0041] In some embodiments, at least 20% of the plurality of genomic regions are exons.

[0042] In some embodiments, less than 10% of the plurality of genomic regions are intergenic regions.

[0043] In some embodiments, the multi-cancer assay panel further comprises: a plurality of virus-specific probes, wherein each of the plurality of virus-specific probes is configured to hybridize to a viral genomic fragment from cfDNA. In some embodiments, the viral genomic fragment is from MCV, EBV, HBV, HCMV, HCV, HHV5, HPV16, or HPV18. In some embodiments, the multi-cancer assay panel comprises at least 50, 100, 200, 500, 1000, 2000, or 3000 virus-specific probes.

[0044] In some embodiments, each of the one or more genomic regions is selected from one of Table 1 or Tables 11 to 15 (or a combination thereof). In some embodiments, each of the one or more genomic regions is selected from Table 13. In some embodiments, each of the one or more genomic regions is selected from Table 14. In some embodiments, each of the one or more genomic regions is selected from Table 15.

[0045] In some embodiments, a plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90% or 95% of the plurality of genomic regions in one or more of Tables 1, 11 to 15.

[0046] In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 13. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 14.

[0047] In some embodiments, a plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90% or 95% of the plurality of genomic regions in one or more of the plurality of genomic regions in Table 15.

[0048] In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 500, 1,000, 5000, 10,000, or 15,000 genes in one or more of Tables 1, 11 to 15. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 500, 1,000, 5000, 10,000, or 15,000 genes in Table 13. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 500, 1,000, 5000, 10,000, or 15,000 genes in Table 14. In some embodiments, the plurality of probes on the assay plate are collectively configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 500, 1,000, 5000, 10,000, or 15,000 genes in Table 15.

[0049] The present invention also discloses multiple assay panels for amplifying multiple cfDNA molecules for cancer diagnosis, the assay panels comprising multiple polynucleotide probes, wherein each of the multiple polynucleotide probes is configured to hybridize with a modified fragment obtained from processing the multiple cfDNA molecules, wherein each of the multiple cfDNA molecules corresponds to or is derived from one or more genomic regions selected from any one of Tables 1 to 24.

[0050] In some embodiments, each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from any one of Tables 2 to 10 or Tables 16 to 24 (or a combination thereof). In some embodiments, each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from any one of Tables 13, 14, or 15. In some embodiments, each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 13. In some embodiments, each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 14. In some embodiments, each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 15.

[0051] In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions of any one of Tables 1 to 24. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions of any one of Tables 2 to 10 or 16 to 24 (or a combination thereof). In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 13. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 14. In some embodiments, a plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90% or 95% of the plurality of genomic regions in one or more of the plurality of genomic regions in Table 15.

[0052] In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 50, 60, 70, 80, 90, 100, 120, 150, 200, 500, 1,000, 5000, 10,000, or 15,000 genes of any one of Tables 1 to 24. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 50, 60, 70, 80, 90, 100, 120, 150, or 200 genes of any one of Tables 2 to 10 or 16 to 24.

[0053] In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 500, 1,000, 5000, 10,000, or 15,000 genes in Table 13. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 500, 1,000, 5000, 10,000, or 15,000 genes in Table 14. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 500, 1,000, 5000, 10,000, or 15,000 genes in Table 15.

[0054] In some embodiments, the processing of the plurality of cfDNA molecules comprises converting unmethylated C (cytosine) in the plurality of cfDNA molecules to U (uracil). In some embodiments, each of the plurality of probes on the assay plate is conjugated to an affinity moiety, wherein the affinity moiety is not a nucleic acid affinity moiety. In some embodiments, the affinity moiety is a biotin moiety. In some embodiments, at least 3%, 5%, 10%, 15%, 20%, 30%, or 40% of the plurality of probes on the assay plate do not contain G (guanine). In some embodiments, at least 80%, 85%, 90%, 92%, 95%, or 98% of the plurality of probes on the assay plate have only CpG or only CpA at a plurality of CpG detection sites.

[0055] The present invention further discloses various methods for providing sequence information that indicates the presence or absence of cancer, the methods comprising the steps of obtaining a test sample comprising a plurality of cfDNA test molecules; processing the plurality of cfDNA test molecules to obtain a plurality of bisulfite-converted test fragments; contacting the plurality of bisulfite-converted test fragments with an assay plate to amplify a subset of the plurality of bisulfite-converted test fragments by hybridization capture; and sequencing the subset of the plurality of bisulfite-converted test fragments to obtain a set of a plurality of sequence reads. In some embodiments, the assay plate is selected from the plurality of assay plates described above.

[0056] In some embodiments, the methods further comprise the step of determining a cancer classification by evaluating the set of the plurality of sequence reads, wherein the cancer classification is: the presence or absence of cancer; a stage of cancer; the presence or absence of one type of cancer; or the presence or absence of at least 1, 2, 3, 4, or 5 different types of cancer.

[0057] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from one of Tables 1, 12, 13, 14, and 15, wherein the cancer classification is the presence or absence of cancer, or a stage of cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions selected from one of Tables 1, 12, 13, 14, and 15.

[0058] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to a bisulfite-converted fragment obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 2, wherein the cancer classification is the presence or absence of a blood cancer or a stage of a blood cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to a bisulfite-converted fragment obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 2.

[0059] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to a bisulfite-converted fragment obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 3, wherein the cancer classification is the presence or absence of breast cancer or a stage of breast cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to a bisulfite-converted fragment obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 3.

[0060] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 4, wherein the cancer classification is the presence or absence of colorectal cancer or a stage of colorectal cancer. In some embodiments, the plurality of nucleotide probes together are configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 4.

[0061] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 5, wherein the cancer classification is the presence or absence of esophageal cancer or a stage of esophageal cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 5.

[0062] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 6, wherein the cancer classification is the presence or absence of head and neck cancer or a stage of head and neck cancer. In some embodiments, the plurality of nucleotide probes together are configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 6.

[0063] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 7, wherein the cancer classification is the presence or absence of hepatobiliary cancer or a stage of hepatobiliary cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 7.

[0064] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 8, wherein the cancer classification is the presence or absence of lung cancer or a stage of lung cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 8.

[0065] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 9, wherein the cancer classification is the presence or absence of ovarian cancer or a stage of ovarian cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 9.

[0066] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 10, wherein the cancer classification is the presence or absence of pancreatic cancer or a stage of pancreatic cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 10.

[0067] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 16, wherein the cancer classification is the presence or absence of a blood cancer or a stage of a blood cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 16.

[0068] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 17, wherein the cancer classification is the presence or absence of breast cancer or a stage of breast cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 17.

[0069] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 18, wherein the cancer classification is the presence or absence of colorectal cancer or a stage of colorectal cancer. In some embodiments, the plurality of nucleotide probes together are configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 18.

[0070] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 19, wherein the cancer classification is the presence or absence of esophageal cancer or a stage of esophageal cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 19.

[0071] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 20, wherein the cancer classification is the presence or absence of head and neck cancer or a stage of head and neck cancer. In some embodiments, the plurality of nucleotide probes together are configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 20.

[0072] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 21, wherein the cancer classification is the presence or absence of hepatobiliary cancer or a stage of hepatobiliary cancer. In some embodiments, the plurality of nucleotide probes are collectively configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 21.

[0073] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 22, wherein the cancer classification is the presence or absence of lung cancer or a stage of lung cancer. In some embodiments, the plurality of nucleotide probes together are configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 22.

[0074] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 23, wherein the cancer classification is the presence or absence of ovarian cancer or a stage of ovarian cancer. In some embodiments, the plurality of nucleotide probes together are configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 23.

[0075] In some embodiments, the assay panel comprises a plurality of polynucleotide probes, wherein each of the plurality of polynucleotide probes is configured to hybridize to bisulfite-converted fragments obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 24, wherein the cancer classification is the presence or absence of pancreatic cancer or a stage of pancreatic cancer. In some embodiments, the plurality of nucleotide probes together are configured to hybridize to bisulfite-converted fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 24.

[0076] In some embodiments, the step of determining a cancer classification is performed by a method comprising the steps of: generating a test feature vector based on the set of the plurality of sequence reads; and applying the test feature vector to a model obtained by a training procedure, the model having a cancer set of multiple fragments from one or more training subjects having cancer and a non-cancer set of multiple fragments from one or more training subjects not having cancer, wherein the cancer set of multiple fragments and the non-cancer set of multiple fragments each comprise multiple training fragments.

[0077] In some embodiments, the training procedure comprises: obtaining sequence information of multiple training fragments from multiple training subjects; for each training fragment, determining whether the training fragment is hypomethylated or hypermethylated, wherein each of the multiple hypomethylated and hypermethylated training fragments comprises at least a threshold number of multiple CpG sites, and the multiple CpG sites respectively have at least a threshold percentage of being unmethylated or methylated; for each training subject, generating a training feature vector based on the multiple hypomethylated training fragments and generating a training feature vector based on the multiple hypermethylated training fragments; and training the model using the multiple training feature vectors from one or more training subjects without cancer and the multiple training feature vectors from one or more training subjects with cancer.

[0078] In some embodiments, the training procedure comprises: obtaining sequence information of multiple training fragments from multiple training subjects; for each training fragment, determining whether the training fragment is hypomethylated or hypermethylated, wherein each of the multiple hypomethylated and hypermethylated training fragments comprises at least a threshold number of multiple CpG sites, and the multiple CpG sites respectively have at least a threshold percentage of being unmethylated or methylated, and for each training subject, generating a training feature vector based on the multiple hypomethylated training fragments and generating a training feature vector based on the multiple hypermethylated training fragments; and training the model using the multiple training feature vectors from one or more training subjects without cancer and the multiple feature vectors from one or more training subjects with cancer.

[0079] In some embodiments, the training program includes: obtaining sequence information of multiple training fragments from multiple training objects; for each training fragment, determining whether the training fragment is hypomethylated or hypermethylated, wherein each of the multiple hypomethylated and hypermethylated training fragments includes at least one threshold number of multiple CpG sites, and the multiple CpG sites each have at least one threshold percentage of being unmethylated or methylated, and for each of the multiple CpG sites in a reference genome: quantifying a number of multiple hypomethylated training fragments overlapping with the CpG site and a number of multiple hypermethylated training fragments overlapping with the CpG site; and generating a hypomethylation score and a hypermethylation score based on the number of multiple hypomethylated training fragments and multiple hypermethylated training fragments; for each training fragment , generating a total low methylation score based on the low methylation scores of the multiple CpG sites in the training segments, and generating a total high methylation score based on the high methylation scores of the multiple CpG sites in the training segments; for each training subject: ranking the multiple training segments based on the total low methylation score, and ranking the multiple training segments based on the total high methylation score; and generating a feature vector based on the ranking of the multiple training segments; obtaining multiple training feature vectors of one or more training subjects without cancer, and multiple training feature vectors of one or more training subjects with cancer; and training the model using the multiple feature vectors of the one or more training subjects without cancer and the multiple feature vectors of the one or more training subjects with cancer. In some embodiments, the model comprises one of a kernel logistic regression classifier, a random forest classifier, a hybrid model, a convolutional neural network, and an autoencoder model. In some embodiments, the method further comprises: obtaining a cancer probability of the test sample based on the model; and comparing the cancer probability with a threshold probability to determine whether the test sample is from a patient with cancer or a patient without cancer. In some embodiments, the method further comprises administering an anticancer agent to the subject. In some embodiments, the methods comprise administering an anticancer agent to a subject identified as having cancer by the methods disclosed herein. In some embodiments, the anticancer agent is a chemotherapeutic agent selected from the group consisting of an alkylating agent, an antimetabolite, an anthracycline, an antitumor antibiotic, a cytoskeletal disruptor (paclitaxel), a topoisomerase inhibitor, a mitotic inhibitor, a corticosteroid, a kinase inhibitor, a nucleotide analog, and a platinum-based drug.

[0080] The present invention further discloses various methods, comprising the steps of: obtaining a set of sequence reads of a plurality of modified test fragments, wherein the plurality of modified test fragments are or have been obtained by processing a set of a plurality of nucleic acid fragments from a test subject, wherein each of the plurality of nucleic acid fragments corresponds to or is derived from a plurality of genomic regions selected from one or more of Tables 1 to 24; and applying the set of sequence reads or a test feature vector obtained based on the set of sequence reads to a model obtained by a training procedure, the model having a cancer set of fragments from one or more training subjects having cancer and a non-cancer set of fragments from one or more training subjects not having cancer, wherein the cancer set of fragments and the non-cancer set of fragments comprise a plurality of training fragments.

[0081] In some embodiments, the methods further comprise: a step of obtaining the test feature vector, the step comprising: for each of the plurality of nucleic acid fragments, determining whether the nucleic acid fragment is hypomethylated or hypermethylated, wherein each of the plurality of hypomethylated and hypermethylated nucleic acid fragments comprises at least a threshold number of CpG sites, and the plurality of CpG sites have at least a threshold percentage of being unmethylated or methylated, respectively; for each of the plurality of CpG sites in a reference genome: quantifying a number of the plurality of hypomethylated nucleic acid fragments overlapping the CpG site and a number of the plurality of hypermethylated nucleic acid fragments overlapping the CpG site. a number of segments; and based on the number of multiple low-methylation nucleic acid fragments and the number of multiple high-methylation nucleic acid fragments, generating a low methylation score and a high methylation score; for each nucleic acid fragment, generating a total low methylation score based on the low methylation scores of the multiple CpG sites in the nucleic acid fragment and generating a total high methylation score based on the high methylation scores of the multiple CpG sites in the nucleic acid fragment; ranking the multiple nucleic acid fragments based on the total low methylation scores, and ranking the multiple nucleic acid fragments based on the total high methylation scores; and generating the test feature vector based on the ranking of the multiple nucleic acid fragments.

[0082] In some embodiments, the training procedure comprises the steps of: for each training segment, determining whether the training segment is hypomethylated or hypermethylated, wherein each of the multiple hypomethylated and hypermethylated training segments comprises at least a threshold number of multiple CpG sites, and the multiple CpG sites each have at least a threshold percentage of being unmethylated or methylated; for each training subject, generating a training feature vector based on the multiple hypomethylated training segments and generating a training feature vector based on the multiple hypermethylated training segments; and training the model using the multiple training feature vectors from one or more training subjects without cancer and the multiple training feature vectors from one or more training subjects with cancer.

[0083] In some embodiments, the training program includes the steps of: for each training segment, determining whether the training segment is hypomethylated or hypermethylated, wherein each of the multiple hypomethylated and hypermethylated training segments includes at least one threshold number of multiple CpG sites, and the multiple CpG sites each have at least one threshold percentage of being unmethylated or methylated, and for each of the multiple CpG sites in a reference genome: quantifying a number of multiple hypomethylated training segments overlapping with the CpG site and a number of multiple hypermethylated training segments overlapping with the CpG site; and generating a hypomethylation score and a hypermethylation score based on the number of multiple hypomethylated training segments and the number of multiple hypermethylated training segments; for each training segment, based on the number of the multiple hypomethylated training segments in the training segment, The low methylation scores of the multiple CpG sites generate a total low methylation score, and a total high methylation score is generated based on the high methylation scores of the multiple CpG sites in the training fragments; for each training object: the multiple training fragments are ranked based on the total low methylation score, and the multiple training fragments are ranked based on the total high methylation score; and a feature vector is generated based on the ranking of the multiple training fragments; multiple training feature vectors of one or more training objects without cancer and multiple training feature vectors of one or more training objects with cancer are obtained; and the model is trained using the multiple feature vectors of the one or more training objects without cancer and the multiple feature vectors of the one or more training objects with cancer.

[0084] In some embodiments, for each CpG site in a reference genome, the methods include the steps of quantifying a number of hypomethylated training segments overlapping the CpG site and a number of hypermethylated training segments overlapping the CpG site, further including the steps of quantifying a number of cancers among the hypomethylated training segments from one or more training subjects with cancer overlapping the CpG site, and quantifying a number of non-cancers among the hypomethylated training segments from one or more training subjects without cancer overlapping the CpG site; and quantifying a number of cancers among the hypermethylated training segments from one or more training subjects with cancer overlapping the CpG site, and quantifying a number of non-cancers among the hypermethylated training segments from one or more training subjects without cancer overlapping the CpG site.

[0085] In some embodiments, for each CpG site in a reference genome, the multiple methods include the steps of: generating a low methylation score and a high methylation score based on the number of multiple low methylation training segments and the number of high methylation training segments; further including the steps of: for generating the low methylation score, calculating a low methylation ratio of the number of cancers in the multiple low methylation training segments to the low methylation sum of the number of cancers in the multiple low methylation training segments and the number of non-cancers in the multiple low methylation training segments; and for generating the high methylation score, calculating a high methylation ratio of the number of cancers in the multiple high methylation training segments to the high methylation sum of the number of cancers in the multiple high methylation training segments and the number of non-cancers in the multiple high methylation training segments.

[0086] In some embodiments, the model comprises one of a kernel logistic regression classifier, a random forest classifier, a hybrid model, a convolutional neural network, and an autoencoder model. In some embodiments, the set of the plurality of sequence reads is obtained using any of a plurality of assay panels as described above.

[0087] The present invention also discloses various methods for designing a test panel for cancer diagnosis, the methods comprising the steps of: identifying a plurality of genomic regions, wherein each of the plurality of genomic regions: (i) comprises at least 30 nucleotides, and (ii) comprises at least 5 methylation sites; selecting a subset of the plurality of genomic regions, wherein the selection is performed when a plurality of cfDNA molecules corresponding to or derived from each of the plurality of genomic regions in a plurality of cancer training samples have an aberrant methylation pattern, wherein the aberrant methylation pattern comprises at least 5 methylation sites that are known or identified to be hypomethylated or hypermethylated; and designing the test panel, the test panel comprising a plurality of probes, wherein each of the plurality of probes is configured to hybridize to a modified fragment obtained from processing a plurality of cfDNA molecules corresponding to or derived from one or more of the subset of the plurality of genomic regions. In some embodiments, the processing of the plurality of cfDNA molecules comprises converting unmethylated C (cytosine) in the plurality of cfDNA molecules to U (uracil).

[0088] The present invention also discloses a plurality of bait sets for hybridization capture, wherein the bait sets comprise at least 50 different polynucleotide-containing probes, wherein each of the plurality of polynucleotide-containing probes has a nucleic acid sequence that is: (1) identical in sequence to a sequence in a genomic region selected from any genomic region listed in any one of Tables 1 to 24, or (2) varies relative to a sequence in the genomic region only by one or more transitions, wherein each of the one or more transitions occurs at a nucleotide corresponding to a CpG site in the genomic region.

[0089] In some embodiments, each of the plurality of polynucleotide-containing probes has a nucleic acid sequence of at least 45 base pairs in length. In some embodiments, the plurality of polynucleotide-containing probes has a nucleic acid sequence of no more than 200 base pairs in length. In some embodiments, the at least 50 different polynucleotide-containing probes are organized into at least 25 pairs of polynucleotide-containing probes, wherein each pair of the plurality of probes comprises a first probe and a second probe, the second probe being different from the first probe, wherein the first probe overlaps with the second probe in sequence by at least 30 nucleotides. In some embodiments, the first probe overlaps with the second probe in sequence by at least 40, 50, 75, or 100 nucleotides.

[0090] In some embodiments, the plurality of polynucleotide-containing probes are organized into at least 50, 60, 70, 80, 90, 100, 120, 150, or 200 pairs of polynucleotide-containing probes. In some embodiments, the plurality of polynucleotide-containing probes are organized into at least 1000, 2000, 2500, 5000, 6000, 7500, 10,000, 15,000, 20,000, or 25,000 pairs of polynucleotide-containing probes. In some embodiments, a uracil or a thymine is located at the transition. In some embodiments, each of the plurality of polynucleotide-containing probes is bound to an affinity moiety, wherein the affinity moiety is not a nucleic acid affinity moiety. In some embodiments, the affinity moiety comprises biotin.

[0091] In some embodiments, each of the plurality of polynucleotide-containing probes comprises fewer than 20, 15, 10, 8, or 6 CpG detection sites. In some embodiments, the bait set does not have probes that are homologous or complementary in sequence to more than 8, 10, 15, or 20 off-target genomic regions.

[0092] In some embodiments, the bait set has at least 50, 60, 70, 80, 90, 100, 120, 150 or 200 polynucleotides containing probes. In some embodiments, the bait set has at least 1000, 2000, 2500, 5000, 10000, 12000, 15000, 20000 or 25000 polynucleotides containing probes. In some embodiments, at least 3%, 5%, 10%, 15%, 20%, 30% or 40% of all the multiple polynucleotide-containing probes in the bait set lack G (guanine). In some embodiments, the multiple polynucleotide-containing probes collectively comprise at least 10,000, 200,000, 50,000, 200,000, 400,000, 600,000, 800,000, 1 million, 2 million or 4 million nucleotides. In some embodiments, each of the multiple polynucleotide probes comprises at least 50, 75, 100 or 120 nucleotides. In some embodiments, each of the plurality of polynucleotide probes has less than 300, 250, 200, or 150 nucleotides. In some embodiments, each of the plurality of polynucleotide probes has 100 to 150 nucleotides.

[0093] In some embodiments, at least 80%, 85%, 90%, 92%, 95% or 98% of the plurality of polynucleotide-containing probes have CpG or CpA only at a plurality of CpG detection sites. In some embodiments, the plurality of polynucleotide-containing probes of the bait set correspond to a total number of genomic regions selected from the plurality of genomic regions of any one of Tables 1 to 24, wherein at least 30% of the plurality of genomic regions are exons or introns. In some embodiments, the plurality of polynucleotide-containing probes of the bait set correspond to a total number of genomic regions, wherein at least 15% of the plurality of genomic regions are exons. In some embodiments, the plurality of polynucleotide-containing probes of the bait set correspond to a total number of genomic regions, wherein at least 20% of the plurality of genomic regions are exons. In some embodiments, the plurality of polynucleotide-containing probes of the bait set correspond to a total number of genomic regions, wherein less than 10% of the plurality of genomic regions are intergenic regions.

[0094] In some embodiments, the bait set further comprises: a plurality of virus-specific probes, each of the plurality of virus-specific probes being configured to hybridize with a viral genomic fragment. In some embodiments, the viral genomic fragment is from MCV, EBV, HBV, HCMV, HCV, HHV5, HPV16 or HPV18. In some embodiments, the plurality of virus-specific probes comprises at least 50, 100, 200, 500, 1000, 2000 or 3000 virus-specific probes. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 2, or (2) varies only by one or more conversions relative to a sequence in a genomic region selected from Table 2, wherein each of the one or more conversions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0095] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 3, or (2) varies from a sequence in a genomic region selected from Table 3 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 4, or (2) varies from a sequence in a genomic region selected from Table 4 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0096] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 5, or (2) varies from a sequence in a genomic region selected from Table 5 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 6, or (2) varies from a sequence in a genomic region selected from Table 6 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0097] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 7, or (2) varies from a sequence in a genomic region selected from Table 7 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 8, or (2) varies from a sequence in a genomic region selected from Table 8 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0098] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 9, or (2) varies from a sequence in a genomic region selected from Table 9 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 10, or (2) varies from a sequence in a genomic region selected from Table 10 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0099] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from any one of Tables 2 to 10, or (2) varies from a sequence in a genomic region selected from any one of Tables 2 to 10 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 11, or (2) varies from a sequence in a genomic region selected from Table 11 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0100] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 12, or (2) varies from a sequence in a genomic region selected from Table 12 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 13, or (2) varies from a sequence in a genomic region selected from Table 13 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0101] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 14, or (2) varies from a sequence in a genomic region selected from Table 14 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 15, or (2) varies from a sequence in a genomic region selected from Table 15 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0102] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 16, or (2) varies from a sequence in a genomic region selected from Table 16 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 17, or (2) varies from a sequence in a genomic region selected from Table 17 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 18, or (2) varies from a sequence in a genomic region selected from Table 18 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0103] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 19, or (2) varies from a sequence in a genomic region selected from Table 19 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 20, or (2) varies from a sequence in a genomic region selected from Table 20 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0104] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 21, or (2) varies from a sequence in a genomic region selected from Table 21 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 22, or (2) varies from a sequence in a genomic region selected from Table 22 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0105] In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 23, or (2) varies from a sequence in a genomic region selected from Table 23 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region. In some embodiments, each of the plurality of polynucleotide-containing probes is: (1) identical in sequence to a sequence in a genomic region selected from Table 24, or (2) varies from a sequence in a genomic region selected from Table 24 only by one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide corresponding to a CpG site in the genomic region.

[0106] In some embodiments, an entirety of the plurality of polynucleotide probes in the bait set is configured to hybridize to a plurality of fragments obtained from a plurality of cfDNA molecules corresponding to at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions selected from any one of Tables 1 to 24. In some embodiments, an entirety of the plurality of polynucleotide-containing probes in the bait set is configured to hybridize to a plurality of fragments obtained from a plurality of cfDNA molecules corresponding to at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions selected from any one of Tables 2 to 10 or 16 to 24. In some embodiments, an entirety of the plurality of polynucleotide-containing probes in the bait set is configured to hybridize to a plurality of fragments obtained from a plurality of cfDNA molecules corresponding to at least 500, 1000, 5000, 10,000, or 15,000 genomic regions selected from any one of Tables 1 to 24. In some embodiments, an entirety of the plurality of polynucleotide-containing probes in the bait set is configured to hybridize to a plurality of fragments obtained from cfDNA molecules corresponding to at least 50, 60, 70, 80, 90, 100, 120, 150, or 200 genomic regions in any one of Tables 2 to 10 or 16 to 24. In some embodiments, the nucleic acid sequence of each of the plurality of polynucleotide-containing probes varies from a sequence in the genomic region by only one or more transitions, wherein each of the one or more transitions corresponds to a transition occurring at a nucleotide at a CpG site in the genomic region.

[0107] The present invention also discloses a plurality of mixtures comprising: bisulfite-converted cell-free DNA; and a plurality of bait sets as described above.

[0108] Furthermore, the present invention discloses various methods for amplifying a bisulfite-converted cell-free DNA sample, the methods comprising the steps of: contacting the bisulfite-converted cell-free DNA sample with the bait set of any one of claims 122 to 177 to form a mixture; and amplifying a first set of multiple genomic regions of the sample by hybridization capture.

[0109] The present invention also discloses methods for providing sequence information indicating the presence or absence of a cancer, a stage of a cancer, or a type of cancer, the methods comprising the steps of: treating cell-free DNA from a biological sample with a deaminating agent to produce a cell-free DNA sample, the cell-free DNA sample comprising a plurality of deaminated nucleotides; amplifying the cell-free DNA sample to obtain information about a plurality of cell-free DNA molecules, wherein amplifying the cell-free DNA sample to obtain information about the plurality of cell-free DNA molecules comprises: contacting the cell-free DNA with a plurality of probes configured to hybridize to a plurality of cell-free DNA molecules corresponding to a plurality of regions identified in any one of Tables 1 to 24; and sequencing the amplified plurality of cell-free DNA molecules to obtain a set of sequence reads to provide information about the presence or absence of a cancer, a stage of a cancer, or a type of cancer.

[0110] In some embodiments, the plurality of probes comprises a plurality of primers, and amplifying the cell-free DNA comprises: amplifying (e.g., by polymerase chain reaction (PCR)) the plurality of cell-free DNA fragments using the plurality of primers (optionally, without hybridization capture). In some embodiments, the cell-free DNA sample is amplified by any suitable method as described herein, and the plurality of probes comprises the plurality of probes comprising polynucleotides.

[0111] In some embodiments, the methods further comprise the step of determining a cancer classification by evaluating the set of the plurality of sequence reads, wherein the cancer classification is: the presence or absence of cancer; a stage of cancer; the presence or absence of a type of cancer; or the presence or absence of at least 1, 2, 3, 4, or 5 different types of cancer.

[0112] In some embodiments, the plurality of probes are configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from one of Tables 1, 11, 12, 13, 14, or 15, wherein the cancer classification is the presence or absence of cancer, or a stage of cancer. In some embodiments, the plurality of probes are configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from one of Tables 1, 11, 12, 13, 14, or 15.

[0113] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 2, wherein the cancer classification is the presence or absence of a blood cancer or a stage of a blood cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 2.

[0114] In some embodiments, the plurality of probes are configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 3, wherein the cancer classification is the presence or absence of breast cancer or a stage of breast cancer. In some embodiments, the plurality of probes are configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 3.

[0115] In some embodiments, the plurality of probes are configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 4, wherein the cancer classification is the presence or absence of colorectal cancer or a stage of colorectal cancer. In some embodiments, the plurality of probes are configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 4.

[0116] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 5, wherein the cancer classification is the presence or absence of esophageal cancer or a stage of esophageal cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 5.

[0117] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 6, wherein the cancer classification is the presence or absence of head and neck cancer or a stage of head and neck cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 6.

[0118] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 7, wherein the cancer classification is the presence or absence of hepatobiliary cancer or a stage of hepatobiliary cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 7.

[0119] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 8, wherein the cancer classification is the presence or absence of lung cancer or a stage of lung cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 8.

[0120] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 9, wherein the cancer classification is the presence or absence of ovarian cancer or a stage of ovarian cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 9.

[0121] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 10, wherein the cancer classification is the presence or absence of pancreatic cancer or a stage of pancreatic cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in one of Table 10.

[0122] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 16, wherein the cancer classification is the presence or absence of a blood cancer or a stage of a blood cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 16.

[0123] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 17, wherein the cancer classification is the presence or absence of breast cancer or a stage of breast cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 17.

[0124] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 18, wherein the cancer classification is the presence or absence of colorectal cancer or a stage of colorectal cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 18.

[0125] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 19, wherein the cancer classification is the presence or absence of esophageal cancer or a stage of esophageal cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 19.

[0126] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 20, wherein the cancer classification is the presence or absence of head and neck cancer or a stage of head and neck cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 20.

[0127] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 21, wherein the cancer classification is the presence or absence of hepatobiliary cancer or a stage of hepatobiliary cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 21.

[0128] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 22, wherein the cancer classification is the presence or absence of lung cancer or a stage of lung cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 22.

[0129] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 23, wherein the cancer classification is the presence or absence of ovarian cancer or a stage of ovarian cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 23.

[0130] In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions selected from Table 24, wherein the cancer classification is the presence or absence of pancreatic cancer or a stage of pancreatic cancer. In some embodiments, the plurality of probes is configured to hybridize to a plurality of bisulfite-converted fragments obtained from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of cfDNA molecules corresponding to or derived from the plurality of genomic regions in Table 24.

[0131] In some embodiments, the step of determining a cancer classification comprises: generating a test feature vector based on the set of the plurality of sequence reads; and applying the test feature vector to a model obtained through a training procedure, the model having a cancer set of multiple fragments from one or more training subjects having cancer and a non-cancer set of multiple fragments from one or more training subjects not having cancer, wherein the cancer set of multiple fragments and the non-cancer set of multiple fragments each comprise multiple training fragments.

[0132] In some embodiments, the training procedure comprises: obtaining sequence information of a plurality of training fragments from a plurality of training subjects; for each training fragment, determining whether the training fragment is hypomethylated or hypermethylated, wherein each of the plurality of hypomethylated and hypermethylated training fragments comprises a plurality of CpG sites having at least a threshold number, and the plurality of CpG sites respectively have at least a threshold percentage of being unmethylated or methylated; for each training subject, generating a training feature vector based on the plurality of hypomethylated training fragments and the plurality of hypermethylated training fragments; and training the model using the plurality of training feature vectors from the one or more training subjects without cancer and the plurality of training feature vectors from the one or more training subjects with cancer.

[0133] In some embodiments, the training program includes: obtaining sequence information of multiple training fragments from multiple training subjects; for each training fragment, determining whether the training fragment is hypomethylated or hypermethylated, wherein each of the multiple hypomethylated and hypermethylated training fragments contains at least a threshold number of multiple CpG sites, and the multiple CpG sites have at least a threshold percentage of being unmethylated or methylated; for each of multiple CpG sites in a reference genome: quantifying a number of multiple hypomethylated training fragments overlapping with the CpG site and a number of multiple hypermethylated training fragments overlapping with the CpG site; and generating a hypomethylation score and a hypermethylation score based on the number of multiple hypomethylated training fragments and multiple hypermethylated training fragments; for each training fragment, Generate a total low methylation score based on the low methylation scores of the multiple CpG sites in the training fragments, and generate a total high methylation score based on the high methylation scores of the multiple CpG sites in the training fragments; for each training object: rank the multiple training fragments based on the total low methylation score, and rank the multiple training fragments based on the total high methylation score; and generate a feature vector based on the ranking of the multiple training fragments; obtain multiple training feature vectors of one or more training objects without cancer, and multiple training feature vectors of one or more training objects with cancer; and train the model using the multiple feature vectors of the one or more training objects without cancer and the multiple feature vectors of the one or more training objects with cancer.

[0134] In some embodiments, the model comprises one of a kernel logistic regression classifier, a random forest classifier, a hybrid model, a convolutional neural network, and an autoencoder model. In some embodiments, the method further comprises: obtaining a probability of cancer for the test sample based on the model; and comparing the probability of cancer to a threshold probability to determine whether the test sample is from a patient with cancer or a patient without cancer. In some embodiments, the method further comprises: administering an anticancer agent to the subject.

[0135] The present invention also discloses methods of treating a cancer patient, comprising administering an anticancer agent to a subject identified as having cancer by the methods described herein. In some embodiments, the anticancer agent is a chemotherapeutic agent selected from the group consisting of an alkylating agent, an antimetabolite, an anthracycline, an antitumor antibiotic, a cytoskeletal disruptor (paclitaxel), a topoisomerase inhibitor, a mitotic inhibitor, a corticosteroid, a kinase inhibitor, a nucleotide analog, and a platinum-based drug.

[0136] Furthermore, the present invention discloses various methods for providing sequence information indicating the presence or absence of cancer, the methods comprising the steps of: contacting cell-free DNA from a biological sample with a deaminating agent to generate a cell-free DNA sample, the cell-free DNA sample comprising a plurality of deaminated nucleotides; amplifying the cell-free DNA to obtain a plurality of DNA fragments, the plurality of DNA fragments collectively corresponding to at least 100, 200, 500, or 1000 genomic regions selected from a plurality of genomic regions identified in any one of Tables 1 to 24; and sequencing the amplified cell-free DNA molecules to obtain a set of a plurality of sequence reads.

[0137] In some embodiments, amplifying the cell-free DNA does not involve hybridization capture. In some embodiments, amplifying the cell-free DNA comprises amplifying the plurality of DNA fragments. In some embodiments, amplifying the cell-free DNA molecules comprises contacting the cell-free DNA with a plurality of primer sets, wherein each primer set comprises a forward primer and a reverse primer, and amplifying the plurality of cell-free DNA molecules by polymerase chain reaction.

[0138] The present invention further discloses multiple assay panels for amplifying multiple cfDNA molecules for cancer diagnosis, the assay panels comprising: at least 50 different pairs of polynucleotide probes, wherein each of the at least 50 pairs of probes: (i) comprises two different probes configured to overlap with each other through an overlapping sequence of 30 or more nucleotides, and (ii) is configured to hybridize with a modified fragment obtained from processing the multiple cfDNA molecules, wherein each of the multiple cfDNA molecules corresponds to or is derived from one or more genomic regions, and wherein each of the one or more genomic regions comprises at least 5 methylation sites and has an aberrant methylation pattern in multiple cancerous training samples.

[0139] In some embodiments, the overlapping sequences comprise at least 40, 50, 75, or 100 nucleotides. In some embodiments, the plurality of assay panels disclosed herein comprise at least 50, 60, 70, 80, 90, 100, 120, 150, or 200 pairs of probes.

[0140] The present invention also discloses multiple assay panels for amplifying multiple cfDNA molecules for cancer diagnosis, the assay panels comprising: at least 100 polynucleotide probes, wherein each of the at least 100 probes is configured to hybridize with a modified polynucleotide, the modified polynucleotide hybridization being obtained from processing the multiple cfDNA molecules, wherein the multiple cfDNA molecules correspond to or are derived from one or more genomic regions, and wherein each of the one or more genomic regions comprises at least 5 methylation sites and has an abnormal methylation pattern in multiple cancerous training samples.

[0141] In some embodiments, the processing of the plurality of cfDNA molecules comprises converting unmethylated C (cytosine) in the plurality of cfDNA molecules into U (uracil). In some embodiments, each of the plurality of probes on the assay plate is conjugated to a biotin moiety. In some embodiments, the plurality of training samples are a plurality of samples from a plurality of subjects determined to have cancer. In some embodiments, when a methylation state vector representing the genomic region in the plurality of cancerous training samples appears at a frequency lower than a threshold in a plurality of reference samples, a genomic region in the plurality of cancerous training samples has an abnormal methylation pattern. In some embodiments, the threshold is 0.1, 0.01, 0.001, or 0.0001.

[0142] In some embodiments, each of the one or more genomic regions is hypermethylated or hypomethylated in the multiple cancerous training samples. In some embodiments, at least 80%, 85%, 90%, 92%, 95% or 98% of the at least 5 methylation sites are methylated or unmethylated in the multiple cancerous training samples. In some embodiments, at least 3%, 5%, 10%, 15%, 20%, 30% or 40% of the multiple probes on the assay plate do not contain G (guanine). In some embodiments, at least 80%, 85%, 90%, 92%, 95%, 98% of the multiple probes on the assay plate have only CpG or only CpA at multiple CpG detection sites. In some embodiments, each of the multiple probes on the assay plate contains less than 20, 15, 10, 8 or 6 CpG detection sites.

[0143] In some embodiments, each of the plurality of probes on the assay plate is designed to have sequence homology or sequence complementarity with fewer than 20, 15, 10, or 8 off-target genomic regions. In some embodiments, the fewer than 20 off-target genomic regions are identified using a k-mer seeding strategy. In some embodiments, the fewer than 20 off-target genomic regions are identified using a k-mer seeding strategy in combination with local alignment at multiple seeding positions.

[0144] In some embodiments, the assay plate comprises at least 100, 200, 300, or 400 probes. In some embodiments, the at least 500 pairs of probes or the at least 100 probes simultaneously comprise at least 10,000, 20,000, or 50,000 nucleotides. In some embodiments, each of the plurality of probes on the assay plate comprises at least 50, 75, 100, or 120 nucleotides. In some embodiments, each of the plurality of probes on the assay plate comprises less than 300, 250, 200, or 150 nucleotides. In some embodiments, each of the plurality of probes on the assay plate comprises between 100 and 150 nucleotides.

[0145] In some embodiments, the assay panel further comprises: a plurality of virus-specific probes, wherein each of the plurality of virus-specific probes is configured to hybridize to a viral genomic fragment from cfDNA. In some embodiments, the viral genomic fragment is from MCV, EBV, HBV, HCMV, HCV, HHV5, HPV16, or HPV18. In some embodiments, the assay panel comprises at least 50, 100, 200, 500, 1000, 2000, or 3000 virus-specific probes.

[0146] In some embodiments, each of the one or more genomic regions is selected from one or more of the plurality of genomic regions in Tables 2 to 10 or Tables 16 to 24. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in one or more of Tables 2 to 10 or Tables 16 to 24. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 50, 60, 70, 80, 90, 100, 120, 150, or 200 genomic regions in one or more of Tables 2 to 10 or 16 to 24. BRIEF DESCRIPTION OF THE DRAWINGS

[0147] Figure 1A A design of 2× tiled probes is shown, according to one embodiment, with each base in a target area (framed within the dashed rectangle) being covered by exactly two probes.

[0148] Figure 1B Probe designs targeting multiple hypomethylated and / or hypermethylated fragments in multiple genomic regions are shown according to one embodiment.

[0149] Figure 2 is a schematic diagram illustrating a process for generating a cancer test panel according to one embodiment.

[0150] Figure 3A is a flowchart depicting a process for creating a data structure for a control group, according to one embodiment.

[0151] Figure 3B According to one embodiment, the verification Figure 3A A flowchart of an additional step for the data structure of the control group.

[0152] Figure 4 A flow chart depicts a process for selecting genomic regions for designing probes for use in a cancer assay panel, according to one embodiment.

[0153] Figure 5 is a diagram of an example p-value score calculation, according to one embodiment.

[0154] Figure 6Ais a flow chart describing a process for training a classifier based on multiple segments representing hypomethylation and hypermethylation of cancer, according to one embodiment.

[0155] Figure 6B is a flow chart describing a process for identifying multiple segments indicative of cancer determined by multiple probabilistic models, according to one embodiment.

[0156] Figure 7A is a flow chart describing a process for sequencing a fragment of cell-free DNA (cfDNA), according to one embodiment.

[0157] Figure 7B According to one embodiment, Figure 7A Schematic diagram of a process for sequencing a fraction of cell-free DNA (cfDNA) to obtain a methylation state vector.

[0158] Figure 8A is a flow chart of multiple apparatuses for sequencing multiple nucleic acid samples according to one embodiment.

[0159] Figure 8B According to one embodiment, an analysis system for analyzing the methylation status of cfDNA is provided.

[0160] Figure 9 A graph is provided showing the amount of DNA fragments hybridized to the probes depending on the size of overlap between the DNA fragments and the probes.

[0161] Figure 10 The numbers of high-quality (high Q), low-quality (low Q), and poor-quality (poor Q) probes were compared among probes targeting multiple fragments with hypermethylation (Hyper) or multiple fragments with hypomethylation (Hypo).

[0162] Figure 11A 、 11B and 11C contain graphs showing the cancer log-odds ratios determined for various cancers at different cancer stages and cancer types.

[0163] Figure 12 The frequencies of multiple genomic annotations are summarized for multiple targeted genomic regions (black) and multiple randomly selected regions (grey).

[0164] Figure 13A and 13B Three graphs are shown to verify the consistency of data from a control group sequencing.

[0165] Figure 14A Shown are the sensitivities for cancer classification at 95% specificity for three different sources of data - Mscore.testV1, Mscore.testV1.cv, and Mscore.testV1.cv.panel.

[0166] Figure 14B Receiver operating characteristic (ROC) curves for cancer classifier performance analysis on three different datasets are shown.

[0167] The various embodiments of the present specification are described in the drawings for illustration only. Those skilled in the art will readily recognize from the following discussion that alternative embodiments of the structures and methods shown herein may be used without departing from the principles described herein. DETAILED DESCRIPTION

[0168] definition

[0169] Unless otherwise defined, all technical and scientific terms used herein have the meaning commonly understood by those skilled in the art to which this specification belongs. As used herein, the following terms have the meanings described below.

[0170] The term "methylation" as used herein refers to the process of adding a methyl group to a DNA molecule. Two of the four bases of DNA, cytosine ("C") and adenine ("A"), can be methylated. For example, a hydrogen atom on the pyrimidine ring of the cytosine base can be converted into a methyl group to form 5-methylcytosine. Methylation tends to occur at dinucleotides of cytosine and guanine, referred to herein as "CpG sites." In other cases, methylation may occur at dinucleotides where cytosine is not a CpG site or at another nucleotide that is not cytosine; however, these cases are relatively rare. In this disclosure, for the sake of clarity, methylation is discussed with respect to multiple CpG sites. However, the principles described herein are equally applicable to detecting methylation in non-CpG contexts, including non-cytosine methylation. For example, adenine methylation has been observed in the DNA of bacteria, plants, and mammals, although it has received much less attention.

[0171] In these embodiments, the wet laboratory assays used to detect methylation may differ from those known in the art as described herein. Furthermore, the methylation state vector may contain multiple elements, typically vectors of multiple sites where methylation has or has not occurred (even if these sites are not specific CpG sites). Using these substitutions, the remaining procedures described herein are the same, and thus the inventive concepts described herein are applicable to these other forms of methylation.

[0172] As used herein, the term "methylation site" refers to a site on a DNA molecule where a methyl group can be added. The "CpG" site is the most common methylation site, but methylation sites are not limited to CpG sites.

[0173] For example, DNA methylation can occur in cytosine at CHG and CHH, where H is adenine, cytosine, or thymine. Cytosine methylation in the form of 5-hydroxymethylcytosine can also be assessed using the methods and procedures disclosed herein (e.g., see PCT Patent Application Publication Nos. WO 2010 / 037001 and WO 2011 / 127136, incorporated herein by reference) and its characteristics.

[0174] As used herein, the term "CpG site" refers to a region in a DNA molecule where a cytosine nucleotide is followed by a guanine nucleotide in a linear sequence of bases in the 5' to 3' direction. "CpG" is short for 5'-C-phosphate-G-3', which is a cytosine and guanine separated by only one phosphate group; the phosphate links any two nucleotides in DNA together. Multiple cytosines in a CpG dinucleotide can be methylated to form 5-methylcytosine.

[0175] As used herein, the term "CpG detection site" refers to a region of a probe configured to hybridize to a CpG site of a target DNA molecule. The CpG site on the target DNA molecule may comprise cytosine and guanine separated by a phosphate group, wherein the cytosine is methylated or unmethylated. The CpG site on the target DNA molecule may comprise uracil and guanine separated by a phosphate group, wherein uracil is generated by conversion of unmethylated cytosine.

[0176] The term "UpG" is short for 5'-U-phosphate-G-3', meaning uracil and guanine are separated by only a phosphate group. UpG can be generated from DNA by bisulfite treatment, which converts unmethylated cytosine to uracil. Cytosine can also be converted to uracil by other methods known in the art, such as chemical modification or synthesis.

[0177] As used herein, the terms "hypomethylated" or "hypermethylated" refer to the methylation state of a DNA molecule containing multiple CpG sites (e.g., greater than 3, 4, 5, 6, 7, 8, 9, 10, etc.), wherein a high percentage (e.g., greater than 80%, 85%, 90% or 95%, or any other percentage within the range of 50% to 100%) of the multiple CpG sites are unmethylated or methylated, respectively.

[0178] As used herein, the term "methylation state vector" refers to a vector comprising a plurality of elements, wherein each element represents the methylation status of a methylation site in a DNA molecule comprising the plurality of methylation sites, in the order from 5' to 3' in the DNA molecule. For example, <M x ,M x+1 ,M x+2 >、 <M x ,M x+1 ,U x+2 >、…、 x ,U x+1 ,U x+2 > can be multiple methylation vectors of multiple DNA molecules containing three methylation sites, where M represents methylated sites and U represents unmethylated sites.

[0179] ​As used herein, the term "abnormal methylation pattern" refers to a methylation state vector or a methylation state of a DNA molecule having the methylation state vector that is expected to be found in a sample at a frequency lower than a threshold. In specific embodiments provided herein, the expected value of finding a particular methylation state vector in a healthy control group comprising multiple healthy individuals is represented by a p-value. Thus, a low p-value score typically corresponds to a methylation state vector that is relatively unexpected compared to other methylation state vectors in multiple samples from multiple healthy individuals in the healthy control group. A high p-value score typically corresponds to a methylation state vector that is relatively more expected compared to other methylation state vectors found in multiple samples from multiple healthy individuals in the healthy control group. A methylation state vector having a p-value lower than a threshold (e.g., 0.1, 0.01, 0.001, 0.0001, etc.) can be defined as an abnormal methylation pattern. Various methods known in the art can be used to calculate the p-value or expected value of a methylation pattern or methylation state vector. Exemplary methods provided herein involve the use of Markov chain probability, which assumes that the methylation states of multiple CpG sites depend on the methylation states of multiple adjacent CpG sites. Alternative methods provided herein calculate the expected value of observing a particular methylation state vector in multiple healthy individuals by using a mixture model comprising multiple mixture components, wherein each mixture component is an independent site model, wherein the methylation at each CpG site is assumed to be independent of the methylation state at other CpG sites.

[0180] The methods provided herein use multiple genomic regions with an abnormal methylation pattern. When multiple cfDNA fragments corresponding to or originating from a genomic region have multiple methylation state vectors whose frequency of occurrence is lower than a threshold value in multiple reference samples, the genomic region can be determined to have an abnormal methylation pattern. The multiple reference samples can be samples from multiple control subjects or multiple healthy subjects. The frequency of occurrence of a methylation state vector in the multiple reference samples can be represented by a p-value score. When the multiple cfDNA fragments corresponding to or originating from the genomic region do not have a single, consistent methylation state vector, the genomic region can have multiple p-value scores for multiple methylation state vectors. In this case, the multiple p-value scores can be summed or averaged before being compared with the threshold. Various methods known in the art can be used to compare the multiple p-value scores corresponding to the genomic region and the threshold, including but not limited to the arithmetic mean, geometric mean, harmonic mean, median, mode, etc.

[0181] The term "cancerous sample" as used herein refers to a sample comprising multiple genomic DNAs from an individual diagnosed with cancer. The multiple genomic DNAs may be, but are not limited to, multiple cfDNA fragments or multiple chromosomal DNAs from an object with cancer. The multiple genomic DNAs may be sequenced and their methylation status may be assessed by methods known in the art (e.g., bisulfite sequencing). When multiple genomic sequences are obtained from a public database (e.g., The Cancer Genome Atlas (TCGA)) or by sequencing a genome of an individual diagnosed with cancer, the cancerous sample may refer to multiple genomic DNAs or multiple cfDNA fragments having the multiple genomic sequences. The term "multiple cancerous samples" as a plural number refers to multiple samples comprising multiple genomic DNAs from multiple individuals, each individual being diagnosed with cancer. In various embodiments, multiple cancerous samples from 100, 300, 500, 1000, 10,000, 20,000, 40,000, 50,000 or more individuals diagnosed with cancer are used.

[0182] As used herein, the term "non-cancerous sample" refers to a sample comprising multiple genomic DNAs from an individual who has not been diagnosed with cancer. The multiple genomic DNAs may be, but are not limited to, multiple cfDNA fragments or multiple chromosomal DNAs from a subject who does not have cancer. The multiple genomic DNAs may be sequenced and their methylation status may be assessed by methods known in the art (e.g., bisulfite sequencing). When multiple genomic sequences are obtained from a public database (e.g., The Cancer Genome Atlas (TCGA)) or obtained by sequencing a genome of an individual who does not have cancer, the non-cancerous sample may refer to multiple genomic DNAs or multiple cfDNA fragments having the multiple genomic sequences. The term "multiple non-cancerous samples" as a plural number refers to multiple samples comprising multiple genomic DNAs from multiple individuals, each individual not being diagnosed with cancer. In various embodiments, cancerous samples from 100, 300, 500, 1000, 10,000, 20,000, 40,000, 50,000 or more individuals who do not have cancer are used.

[0183] The term "training sample" as used herein refers to a sample used to train a classifier described herein and / or to select one or more genomic regions for cancer diagnosis. The multiple training samples may include one or more healthy subjects and one or more subjects with a disease condition for diagnosis (e.g., cancer, a specific type of cancer, a specific stage of cancer, etc.). The multiple genomic DNAs may be, but are not limited to, multiple cfDNA fragments or multiple chromosomal DNAs. The multiple genomic DNAs may be sequenced and their methylation status may be assessed by methods known in the art (e.g., bisulfite sequencing). When multiple genomic sequences are obtained from a public database (e.g., The Cancer Genome Atlas (TCGA)) or obtained by sequencing a genome of an individual, a training sample may refer to multiple genomic DNAs or multiple cfDNA fragments having the multiple genomic sequences.

[0184] As used herein, a "test sample" refers to a sample from a subject whose health condition has been, has been, or will be tested using a classifier and / or an assay panel described herein. The test sample may comprise a plurality of genomic DNA fragments or modifications thereof. The plurality of genomic DNA fragments may be, but are not limited to, a plurality of cfDNA fragments or a plurality of chromosomal DNA fragments.

[0185] As used herein, the term "target genomic region" refers to a region of a genome selected for designing a probe to be included in an assay panel. The probe can be designed to hybridize to (and optionally pull down) a nucleic acid fragment corresponding to or derived from the target genomic region or a fragment thereof. A nucleic acid fragment corresponding to or derived from the target genomic region refers to a nucleic acid fragment produced by degradation, cleavage, or other biological treatment of the target genomic region, or a nucleic acid fragment having a sequence homologous or complementary to the target genomic region.

[0186] As used herein, the term "off-target genomic region" refers to a region in a genome that was not selected for designing a probe to be included in an assay panel, but that has sufficient homology to a target genomic region to be bound and pulled down by a probe designed to target the target genomic region. In one embodiment, the off-target genomic region is a genomic region that has at least a 90% match with a probe along at least 45 base pairs (bp).

[0187] The terms "cell-free nucleic acid," "cell-free DNA," or "cfDNA" refer to nucleic acid fragments that circulate in an individual's body (e.g., bloodstream) and originate from one or more healthy cells and / or one or more cancer cells. In addition, cfDNA may originate from other sources, such as viruses and fetal DNA.

[0188] The terms "converted DNA molecules," "converted cfDNA molecules," or "modified fragments obtained from the treatment of multiple cfDNA molecules" refer to multiple DNA molecules obtained by treating multiple DNA or cfDNA molecules in a sample in a chemical reaction to distinguish between a methylated nucleotide and an unmethylated nucleotide in the multiple DNA or cfDNA molecules. For example, in one embodiment, the sample can be treated with sodium bisulfite ions (e.g., using sodium bisulfite) as known in the art to convert unmethylated cytosine ("C") to uracil ("U"). In another embodiment, the conversion of unmethylated cytosine to uracil is accomplished using an enzymatic conversion reaction, such as using a cytidine deaminase (e.g., APOBEC). After treatment, the converted DNA molecules or cfDNA molecules contain additional uracil that was not present in the original cfDNA sample.

[0189] The term "circulating tumor DNA" or "ctDNA" refers to nucleic acid fragments originating from tumor cells or other cell types that are released into an individual's blood as a result of biological processes, such as apoptosis or necrosis of dying cells, or active release by living tumor cells.

[0190] The term "individual" refers to a human individual. The term "healthy individual" refers to an individual who does not have cancer or disease.

[0191] The term "subject" refers to an individual whose DNA is being analyzed. A subject can be a test subject whose DNA can be evaluated using the targeted assay panels described herein to assess whether that person has cancer or another disease. A subject can also be part of a control group of individuals known to be free of cancer or another disease. A subject can also be part of a cancer or other disease group of individuals known to have cancer or another disease. Control groups and cancer / disease groups can be used to assist in the design or validation of targeted assay panels.

[0192] The term "multiple sequence reads" as used herein refers to multiple nucleotide sequences read from a sample. Multiple sequence reads can be obtained by various methods provided herein or methods known in the art.

[0193] As used herein, the term "sequencing depth" refers to the number of counts of a given target nucleic acid that have been sequenced in a sample (e.g., the number of sequence reads in a given target region). Increasing sequencing depth can reduce the number of target nucleic acids required to assess a disease state (e.g., cancer or tissue of origin).

[0194] As used herein, the term "tissue of origin" or "TOO" refers to the organ, group of organs, body region, or cell type from which a cancer arises or originates. Identification of the tissue of origin or cancer cell type often allows identification of the most appropriate next step in the cancer care continuum to further diagnose, stage, and determine treatment.

[0195] Reference to "an entirety of the plurality of probes" of an assay plate or bait set, or "an entirety of the plurality of polynucleotide-containing probes" of an assay plate or bait set, generally refers to all probes provided with a particular assay plate or bait set. For example, in some embodiments, an assay plate or bait set may simultaneously include: (1) a plurality of probes having a plurality of features specified herein (e.g., a plurality of probes for binding to a plurality of cell-free DNA fragments corresponding to or derived from a plurality of genomic regions in one or more tables set forth herein); and (2) an additional plurality of probes that do not include such feature(s). Reference to an entirety of the plurality of probes of an assay plate generally refers to all probes provided with the assay plate or bait set, including probes that do not include the specified feature(s).

[0196] Other interpretation conventions

[0197] 47, 48, 49, and 50.

[0198] Cancer Panel

[0199] In a first aspect, the present specification provides a cancer assay panel (e.g., a bait panel) comprising a plurality of probes or a plurality of probe pairs. The plurality of probes can be a plurality of polynucleotide-containing probes specifically designed to target one or more nucleic acid molecules corresponding to or derived from a plurality of genomic regions that are differentially methylated between cancer and non-cancerous samples, between different types of cancer tissues of origin, between different types of cancer cells, or between samples of different cancer stages, as identified by the methods provided herein. In some embodiments, the plurality of probes target a plurality of genomic regions (or a plurality of nucleic acid molecules derived therefrom) having a plurality of methylation patterns specific to a cancer type, for example, (1) blood cancer, (2) breast cancer, (3) colorectal cancer, (4) esophageal cancer, (5) head and neck cancer, (6) hepatobiliary cancer, (7) lung cancer, (8) ovarian cancer, or (9) pancreatic cancer. In some embodiments, the assay panel comprises a plurality of probes that target a plurality of genomic regions specific to a single cancer type. In some embodiments, the assay panel comprises multiple probes specific for 2, 3, 4, 5, 6, 7, 8, or 9 or more cancer types. In some embodiments, multiple target genomic regions are selected to maximize classification accuracy, subject to size constraints (which can be determined by the sequencing budget and desired sequencing depth).

[0200] Design features and potential utility of an exemplary cancer assay panel include: Figure 2As shown. In order to design the cancer test panel, an analysis system can collect multiple samples corresponding to various results under consideration, such as multiple samples known to have cancer, multiple samples considered healthy, multiple samples from known tissues of origin, etc. These samples can be processed with whole genome bisulfite sequencing (WGBS) or obtained from public databases (such as TCGA). The analysis system can be any general-purpose computing system having a computer processor and a computer-readable storage medium, with multiple instructions for executing the computer processor to perform any or all operations described in the present disclosure. With respect to the multiple samples, the analysis system determines the methylation status at one or more CpG sites of each nucleic acid fragment in the sample. The analysis system can then select multiple target genomic regions based on the multiple methylation patterns of the multiple nucleic acid fragments. One method considers the pairwise resolution between paired results of multiple regions or more specifically one or more CpG sites. Another method considers the resolution of multiple regions or more specifically one or more CpG sites when considering each result relative to the remaining results. From the selected multiple target genomic regions with high resolution capabilities, the analysis system can design multiple probes to target multiple nucleic acid fragments containing the selected multiple genomic regions. The analysis system can generate cancer assay panels of different sizes, for example, a small-sized cancer assay panel comprising multiple probes targeting the most informative genomic regions, a medium-sized cancer assay panel comprising multiple probes from the small-sized cancer assay panel and multiple additional probes targeting genomic regions with secondary information, and a large-sized cancer assay panel comprising multiple probes from the small-sized and medium-sized cancer assay panels and multiple additional probes targeting genomic regions with tertiary information. Using such cancer assay panels, the analysis system can train multiple classifiers with various classification techniques to predict the likelihood that a sample will have a specific outcome (e.g., cancer, a specific cancer type, other diseases, etc.).

[0201] Specifically, in some embodiments, the cancer assay panel comprises at least 50 pairs of probes, wherein each of the at least 50 pairs of probes comprises two probes configured to overlap with each other via an overlapping sequence, wherein the overlapping sequence comprises a sequence of 30 nucleotides configured to hybridize to a modified fragment obtained from a plurality of processed cfDNA molecules corresponding to one or more genomic regions, wherein each of the plurality of genomic regions comprises at least 5 methylation sites, wherein the at least 5 methylation sites have an aberrant methylation pattern in a plurality of training samples. In other words, when analyzing a plurality of cfDNA molecules in a plurality of training samples corresponding to the genomic regions, they have a plurality of methylation state vectors having a frequency of occurrence that is lower than a threshold value in a plurality of reference samples.

[0202] In other embodiments, the cancer assay panel comprises at least 500 probe pairs, wherein each of the at least 500 probe pairs comprises two probes configured to overlap with each other via an overlapping sequence, wherein the overlapping sequence comprises a sequence of 30 nucleotides configured to hybridize to a modified fragment obtained from a plurality of processed cfDNA molecules corresponding to one or more genomic regions, wherein each of the plurality of genomic regions comprises at least 5 methylation sites, wherein the at least 5 methylation sites have an aberrant methylation pattern in a plurality of training samples. Similarly, when analyzing a plurality of cfDNA molecules in a plurality of training samples corresponding to the genomic regions, they have a plurality of methylation state vectors having a frequency of occurrence below a threshold value in a plurality of reference samples.

[0203] In preferred embodiments, the at least five methylation sites are differentially methylated between cancerous and non-cancerous samples or between one or more pairs of samples from different cancer types. In some embodiments, the converted multiple cfDNA molecules comprise multiple cfDNA molecules that are treated (e.g., by bisulfite treatment) to convert unmethylated C (cytosine) to U (uracil). In some cases, uracil is further converted to thymine (e.g., by PCR amplification).

[0204] Because the plurality of probes are configured to hybridize to a converted DNA or cfDNA molecule corresponding to or derived from one or more genomic regions, the plurality of probes may have a sequence different from that of the target genomic region. For example, a DNA molecule containing an unmethylated CpG site will be converted to contain UpG because unmethylated cytosines are converted to uracils by a conversion reaction (e.g., bisulfite treatment). As a result, a probe is configured to hybridize to a sequence containing UpG rather than naturally occurring unmethylated CpG. Thus, a complementary site in the probe to the unmethylated site may contain CpA instead of CpG, and some probes targeting a hypomethylated site where all methylated sites are unmethylated may lack a guanine (G) base. In some embodiments, at least 3%, 5%, 10%, 15%, 20%, 30%, or 40% of the plurality of probes lack G (guanine). In some embodiments, at least 80, 85, 90, 92, 95, 98% of the plurality of probes on the assay plate have only CpG or only CpA at the plurality of CpG detection positions. Thus, in some embodiments, the plurality of polynucleotide-containing probes have a nucleic acid sequence that is (1) identical in sequence to a sequence in a target genomic region (e.g., a plurality of target genomic regions listed in Tables 1-24), or (2) varies by only one or more transitions (e.g., a change in base composition at a locus due to bisulfite conversion or other conversion techniques) relative to a sequence in the genomic region, wherein each of the one or more transitions occurs at a nucleotide corresponding to a CpG site in the genomic region.

[0205] In some embodiments, the plurality of probes on the assay plate comprises fewer than 20, 15, 10, 8, or 6 CpG detection positions. In some embodiments, the plurality of probes on the assay plate comprises more than 5, 6, 7, 8, 9, or 10 CpG detection positions.

[0206] In some embodiments, the plurality of probes are conjugated to a tag (eg, a non-nucleic acid affinity moiety), such as a biotin moiety.

[0207] The cancer assay panel can be used to generally detect the presence or absence of cancer, and / or provide a classification of cancer (e.g., cancer type, or cancer stage (e.g., I, II, III, or IV), or the origin of the cancer). The assay panel can include a plurality of probes that target a plurality of nucleic acids derived from a plurality of genomic regions that are differentially methylated between a plurality of general cancerous (pan-cancer) samples and a plurality of non-cancerous samples, or a plurality of cancerous samples having only a particular cancer type (e.g., a plurality of lung cancer-specific targets). For example, in some embodiments, a cancer assay panel is designed to amplify (enrich) a plurality of nucleic acids derived from a plurality of differentially methylated genomic regions in a plurality of cancerous samples identified based on bisulfite sequencing data generated from cfDNA of cancerous and non-cancerous individuals.

[0208] Each of the plurality of probes (or probe pairs) can be designed to target a plurality of nucleic acids derived from one or more target genomic regions. The plurality of target genomic regions are selected based on several criteria designed to increase the selective amplification of multiple informative cfDNA fragments while reducing noise and nonspecific binding.

[0209] In one example, an assay panel may include multiple probes that are capable of selectively hybridizing (i.e., binding) and amplifying multiple cfDNA fragments that are differentially methylated in multiple cancerous samples. In this case, sequencing the amplified multiple fragments can provide information relevant to cancer diagnosis. In addition, the multiple probes are designed to target multiple genomic regions that are determined to have an abnormal methylation pattern in multiple cancerous samples, or in multiple samples of certain types of tissues or cell types. In one embodiment, multiple probes are designed to target multiple genomic regions that are determined to be hypermethylated or hypomethylated in certain cancers or cancer-origin tissues to provide additional selectivity and specificity for detection. In some embodiments, an assay panel includes multiple probes that target multiple hypomethylated fragments. In some embodiments, an assay panel includes multiple probes that target multiple hypermethylated fragments. In some embodiments, an assay panel includes a first group of multiple probes that target multiple hypermethylated fragments and a second group of multiple probes that target multiple hypomethylated fragments ( Figure 1B In some embodiments, the ratio between the plurality of probes of the first group targeting the plurality of hypermethylated fragments and the plurality of probes of the second group targeting the plurality of hypomethylated fragments (Hyper:Hypo ratio) ranges from 0.4 to 2, from 0.5 to 1.8, from 0.5 to 1.6, from 1.4 to 1.6, from 1.2 to 1.4, from 1 to 1.2, from 0.8 to 1, from 0.6 to 0.8, or from 0.4 to 0.6.

[0210] Various methods for identifying multiple genomic regions that produce aberrantly methylated DNA molecules or differentially methylated DNA molecules (i.e., between cancer and non-cancerous samples, between different types of cancer-origin tissues, between different types of cancer cells, or between multiple samples at different cancer stages) are provided in detail in the section entitled "Methods for Selecting Targeted Genomic Regions", and methods for identifying aberrantly methylated multiple DNA molecules or multiple fragments, wherein the aberrantly methylated multiple DNA molecules or multiple fragments are identified as indicating cancer, are provided in detail in the section entitled "Aberrantly Methylated Fragments" and the section entitled "Filtering of Aberrantly Methylated Fragments".

[0211] In a second example, multiple genomic regions can be selected when they produce aberrantly methylated DNA molecules in multiple cancer samples or multiple samples of a known type of cancer. For example, as described herein, a Markov model trained on a set of multiple reference samples (e.g., multiple samples from healthy subjects) can be used to identify multiple genomic regions that produce aberrantly methylated DNA molecules (i.e., multiple DNA molecules with a methylation pattern below a p-value threshold).

[0212] Each of the plurality of probes can target a genomic region comprising at least 30 bp, 35 bp, 40 bp, 45 bp, 50 bp, 60 bp, 70 bp, 80 bp, 90 bp, 100 bp or more. In some embodiments, the plurality of genomic regions can be selected to have fewer than 30, 25, 20, 15, 12, 10, 8, or 6 methylation sites.

[0213] The plurality of genomic regions may be selected when at least 80%, 85%, 90%, 92%, 95% or 98% of the at least 5 methylation (e.g., CpG) sites within the region are methylated or unmethylated in a plurality of non-cancerous or cancerous samples, or a plurality of cancer samples of a particular cancer type.

[0214] Multiple genomic regions can be further filtered to select only those that are potentially informative based on their methylation patterns, for example, multiple CpG sites that are differentially methylated between multiple samples of cancer and non-cancerous tissue (e.g., aberrantly methylated or unmethylated in cancer versus non-cancerous tissue), multiple cancerous samples of one source tissue and multiple cancerous samples of a different source tissue, or multiple CpG sites that are differentially methylated only in multiple cancerous samples of a particular type. For selection, calculations can be performed relative to each CpG site or multiple CpG sites. For example, a first count can be determined as the number of multiple samples containing cancer (cancer_count) that include a fragment overlapping the CpG, and a second count is determined as the number of total samples containing multiple fragments overlapping the CpG site (total). The plurality of genomic regions can be selected based on a criterion that is positively correlated with the number of samples containing cancer (cancer count) that contain a fragment indicative of cancer that overlaps the CpG site, and negatively correlated with the number of total samples containing fragments indicative of cancer that overlap the CpG site (total). In one embodiment, the number of non-cancerous samples (n non-cancer ) and the number of cancerous samples with a fragment overlapping a CpG site (n cancer ) are counted. Then the probability that a sample is cancer is estimated, for example, as (n cancer +1) / (n cancer +n non-cancer +2).

[0215] Multiple CpG sites calculated by this metric can be ranked and greedily added to an assay panel until the assay panel size budget is exhausted. The procedure for selecting multiple genomic regions representing cancer is further detailed in the section entitled "Multiple Genomic Regions Representing Cancer and Classifiers."

[0216] The selection of which CpG sites comprise the assay panel depends on whether the assay is intended to be a pan-cancer assay or a single cancer assay, or on the desired flexibility. An assay panel for diagnosing a specific cancer type can be designed using a similar procedure. In this example, for each cancer type and each CpG site, information gain is calculated to determine whether a probe targeting that CpG site should be included. Information gain can be calculated for multiple samples of tissue origin of a given cancer type compared to all other samples. For example, consider two random variables, "AF" and "CT." "AF" is a binary variable that indicates whether an aberrant fragment overlaps with a specific CpG site in a particular sample (yes or no). "CT" is a binary random variable that indicates whether the cancer is of a specific type (e.g., lung cancer or non-lung cancer). Given "AF," mutual information about "CT" can be calculated. That is, knowing whether an aberrant fragment overlaps with a specific CpG site tells us how much information can be gained about the cancer type (lung cancer vs. non-lung cancer in this example). This can be used to rank CpGs based on lung specificity. This process is repeated for multiple cancer types. If a particular region is commonly differentially methylated only in lung cancer (and not other cancer types or non-cancer), then the CpGs in that region will tend to have high information gain for lung cancer. For each cancer type, multiple CpG sites are ranked by this information gain measure and then greedily added to an assay panel until the size budget for that cancer type is exhausted.

[0217] Further filtering can be performed to select multiple probes with high specificity for the amplification (i.e., high binding efficiency) of multiple nucleic acids derived from multiple genomic regions of the target. Multiple probes can be filtered to reduce the non-specific binding (or off-target (off-target) binding) of multiple nucleic acids derived from multiple genomic regions of non-target. For example, multiple probes can be filtered to select only those probes with a set threshold value that is less than multiple off-target binding events. In one embodiment, multiple probes can be aligned with a reference genome (e.g., a human reference genome) to select multiple probes aligned with the probes in multiple regions that are less than a set threshold value in the genome. For example, multiple probes aligned with 25, 24, 23, 22, 21, 20, 19, 18, 17, 16, 15, 14, 13, 12, 11, 10, 9 or 8 off-target regions in the reference genome can be selected. In other cases, filtering is performed to remove multiple genomic regions when a sequence of the multiple target genomic regions occurs more than 5, 10, 15, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, or 35 times in a genome. Further filtering can be performed to select multiple target genomic regions when a sequence or a group of sequences is 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, or 99% homologous to the multiple target genomic regions but occurs less than 25, 24, 23, 22, 21, 20, 19, 18, 17, 16, 15, 14, 13, 12, 11, 10, 9, or 8 times in a reference genome, or when the multiple target genomic regions are present in a genome that is 10, 15, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, or 35 times. When the sequence or a group of sequences has 90%, 91%, 92%, 93%, 94%, 95%, 96%, 97%, 98%, or 99% homology to the target genomic regions and occurs more than 5, 10, 15, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, or 35 times in a reference genome, further filtering can be performed to remove the target genomic regions. This is to exclude multiple probes that may pull down multiple off-target fragments, which is undesirable and affects assay efficiency.

[0218] As shown in Example 1, a fragment-probe overlap of at least 45 bp can produce a non-negligible amount of pull-down (although this number may vary depending on the details of the assay). Therefore, in some embodiments, the multiple probes are at least 45 base pairs in length. In some embodiments, a mismatch rate of more than 10% between the probe and the multiple fragment sequences in the overlapping region is sufficient to greatly disrupt binding, thereby reducing efficiency. Therefore, sequences that can be aligned with the probe along at least 45 bp and have a match rate of at least 90% are ideal for off-target pull-down. Therefore, in one embodiment, the number of these regions is scored. The best probes score 1, meaning that they match only in one place (the expected target region). Probes with lower scores (e.g., less than 5 or 10) are accepted, but any probes above this score will be discarded. Other cutoff values ​​can be used for specific samples.

[0219] Once the multiple probes are hybridized and captured with multiple DNA fragments corresponding to or derived from a target genomic region, the hybridized probe-DNA fragment intermediate is pulled down (or separated), and the target DNA is amplified and sequenced. Sequence readings provide information relevant to cancer diagnosis. To this end, an assay plate is designed to include multiple probes that can capture multiple fragments that can collectively provide information relevant to cancer diagnosis. In some embodiments, an assay plate includes at least 50, 60, 70, 80, 90, 100, 120, 150, or 200 pairs of different probes. In other embodiments, an assay plate includes at least 500, 1000, 2000, 2500, 5000, 6000, 7500, 10000, 15000, 20000, 25000, or 50000 pairs of different probes. In some embodiments, an assay plate comprises at least 100, 120, 140, 160, 180, 200, 240, 300, or 400 different probes. In other embodiments, an assay plate comprises at least 1,000, 2,000, 5,000, 10,000, 12,000, 15,000, 20,000, 30,000, 40,000, 50,000, or 100,000 different probes. The plurality of probes may collectively comprise at least 10,000, 20,000, 30,000, 40,000, 50,000, 100,000, 200,000, 400,000, 600,000, 800,000, 1,000,000, 2,000,000, 3,000,000, 4,000,000, 5,000,000, or 10,000,000 nucleotides.

[0220] The selected multiple target genomic regions can be located at different locations in a genome, including but not limited to exons, introns, intergenic regions, and other parts. In some embodiments, multiple probes targeting multiple non-human genomic regions (e.g., target viral genomic regions) can be added.

[0221] probe

[0222] The various cancer assay panels provided herein (e.g., various bait sets) can include a set of multiple hybridization probes (also referred to herein as "multiple probes") designed to amplify, target, and pull down (e.g., by hybridization capture) multiple nucleic acid fragments of interest for assay. In some embodiments, the multiple probes are designed to hybridize and amplify a modified fragment obtained from the processing of multiple molecules of DNA or cfDNA from multiple samples of a subject having cancer or a subject having a specific cancer type. The processing step can convert unmethylated cytosine (C) into uracil (U). The multiple probes can be designed to anneal (or hybridize) to a target (complementary) strand of the modified fragment (e.g., DNA or RNA). The target strand can be a "positive" strand (e.g., a strand that is transcribed into mRNA and subsequently converted into protein) or a complementary "negative" strand. In a specific embodiment, a cancer assay panel comprises two probe sets, one probe targeting the positive strand and the other probe targeting the negative strand of a target genomic region.

[0223] For each target genomic region, four possible probe sequences can be designed. The multiple DNA molecules corresponding to or derived from each target region are double-stranded, and therefore, a probe or probe set can target the "positive strand" or forward strand or its complementary reverse strand ("negative" strand). In addition, in some embodiments, the multiple probes or multiple probe sets are designed to amplify multiple DNA molecules or multiple fragments that have been treated to convert unmethylated cytosine (C) to uracil (U). Because the multiple probes or multiple probe sets are designed to amplify multiple DNA molecules corresponding to or derived from the multiple target regions after conversion, the sequences of the probes can be designed to amplify multiple DNA molecules of multiple fragments, in which the unmethylated C has been converted to U (by replacing G with A at the site of unmethylated cytosine in the multiple DNA molecules or multiple fragments corresponding to or derived from the multiple target regions). In one embodiment, multiple probes are designed to bind or hybridize to multiple DNA molecules or multiple fragments of multiple genomic regions known to contain cancer-specific methylation patterns (e.g., multiple DNA molecules that are hypermethylated or hypomethylated), thereby amplifying multiple DNA molecules or multiple fragments that are cancer-specific. Targeting multiple genomic regions, or multiple methylation patterns that are cancer-specific, can be advantageous, allowing specific amplification of multiple DNA molecules or multiple fragments that are identified as cancer or cancer-origin tissue information, thereby reducing sequencing requirements and sequencing costs. In other embodiments, two probe sequences (one for each DNA chain) can be designed for each target genomic region. In other cases, multiple probes are designed to amplify all DNA molecules or fragments corresponding to or derived from a target region (i.e., regardless of chain or methylation status). This may be because the cancer methylation state is not highly methylated or unmethylated, or because the multiple probes are designed to target small mutations or other variations, rather than methylation changes, that similarly indicate the presence or absence of a cancer, or the presence or absence of one or more tissues of origin of a cancer. In this case, all four possible probe sequences can be included in each target genomic region.

[0224] For example, a plurality of cancer assay panels (e.g., a plurality of bait sets) for hybridization capture can include a plurality of polynucleotide-containing probes, each of the plurality of polynucleotide-containing probes comprising a nucleic acid sequence that is (1) identical in sequence to a sequence within a genomic region (e.g., a genomic region listed in any one of Tables 1 to 24) or (2) varies from a sequence within the genomic region only by one or more transitions, wherein each of the one or more transitions occurs at a nucleotide corresponding to a CpG site within the genomic region. A plurality of probes that are identical in sequence to a sequence within a genomic region can be used to bind to "fully methylated" cell-free DNA molecules, in which no cytosine bases are converted to uracil. Conversely, a plurality of probes having a nucleic acid sequence that varies from a sequence within the genomic region only by one or more transitions (e.g., at multiple CpG sites) can be used to bind to a plurality of partially or fully methylated probes, in which one or more (e.g., all) cytosines are unmethylated and subsequently converted to uracil by a deaminating agent (e.g., sodium bisulfite).

[0225] The length range of the multiple probes is 10s, 100s, 200s or 300s base pairs. The multiple probes may comprise at least 45, 50, 75, 100 or 120 nucleotides. The multiple probes may comprise less than 300, 250, 200 or 150 nucleotides. In one embodiment, the multiple probes comprise 45 to 200 or 100 to 150 nucleotides. In a specific embodiment, the multiple probes comprise 120 nucleotides.

[0226] The plurality of probes are designed to analyze the methylation status of a plurality of target genomic regions (e.g., of a human or other organism) suspected of being associated with the presence or absence of cancer, the presence or absence of certain types of cancer, the stage of cancer, or the presence or absence of other types of disease.

[0227] Furthermore, the plurality of probes can be designed to efficiently hybridize to (or bind to) and pull down a plurality of cfDNA fragments containing a target genomic region. In some embodiments, the plurality of probes are designed to cover multiple overlapping portions of a target genomic region, such that each probe is "tiled" in the coverage range, such that each probe at least partially overlaps with another probe in the library in the coverage range. Figure 1A). In these embodiments, the assay plate comprises a plurality of pairs of probes, wherein each pair comprises at least two overlapping probes, wherein the at least two overlapping probes overlap with each other via an overlapping sequence of at least 25, 30, 35, 40, 45, 50, 60, 70, 75, or 100 nucleotides. In some embodiments, the overlapping sequence can be designed to have homology or complementarity to a target genomic region (or a transformed version of the target genomic region), such that a nucleotide fragment corresponding to, derived from, or containing the target genomic region can be bound and pulled down by at least one of the plurality of probes.

[0228] In one embodiment, Figure 1A A 2× tiling design was used, where one target area ( Figure 1A Each base in the dashed rectangle (in the figure) is overlapped by two probes. For example, each pair of multiple probes can include a first probe and a second probe that is different from the first probe and overlaps with the first probe in sequence (for example, by overlapping by at least 30 nucleotides). This is done to ensure that even relatively short DNA fragments (for example, 100 bp) corresponding to or derived from a target region are guaranteed to have a substantial overlap (or sequence complementarity) with at least one probe, thereby enabling efficient capture of the relatively short DNA fragment. For example, the overlap of a 100 bp DNA fragment with a 30 bp target region is that at least one of the two probes overlaps by at least 75 bp. Other levels of tiling can also be used. For example, to increase target size and capture efficiency, more probes can be tiled (tilted) on a given target region. To increase capture of any DNA fragments overlapping the target region, the multiple probes can be designed to extend beyond the end of the target region on one or both sides. For example, multiple probes can be designed to extend beyond the end of a 30 bp target region by at least 50 bp, 60 bp, 70 bp, 80 bp, 90 bp, or 100 bp.

[0229] In one embodiment, the smallest target genomic region is 30bp. When a new target region is added to the assay plate (based on the greedy selection described above), 30bp of the new target region can be centered on a specific CpG position of interest. Each edge of this new target is then checked to see if it is close enough to other targets so that they can be merged. This is based on a "merge distance" parameter, which can be 200bp by default, but can be adjusted. This allows multiple target regions that are close but different to be amplified with overlapping probes. Depending on whether there are multiple targets close enough to the left or right of the new target, the new target can be merged with nothing (increasing the number of assay plate targets by one), only one target can be merged to the left or right (not changing the number of assay plate targets), or both sides can be merged with existing targets (reducing the number of assay plate targets by one).

[0230] An assay plate provided herein comprises a plurality of polynucleotide probes configured to hybridize to a modified fragment obtained from the processing of a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions. In other words, a plurality of polynucleotide-containing probes of a bait set can (as a group) correspond to a number of a plurality of genomic regions. In some embodiments, at least 15%, 20%, 30% or 40% of the plurality of genomic regions are exons or introns. In some embodiments, at least 5%, 10%, 15%, 20%, 30% or 40% of the plurality of genomic regions are exons. In some embodiments, less than 5%, 10%, 15%, 20%, 25% or 30% of the plurality of genomic regions are intergenic regions.

[0231] In some embodiments, each of the one or more genomic regions is selected from one or more of Tables 1 to 24. In some embodiments, each of the one or more genomic regions is selected from one or more of Tables 2 to 10 or 16 to 24. In some embodiments, each of the one or more genomic regions is selected from Table 1. In some embodiments, each of the one or more genomic regions is selected from Table 2. In some embodiments, each of the one or more genomic regions is selected from Table 3. In some embodiments, each of the one or more genomic regions is selected from Table 4. In some embodiments, each of the one or more genomic regions is selected from Table 5. In some embodiments, each of the one or more genomic regions is selected from Table 6. In some embodiments, each of the one or more genomic regions is selected from Table 7. In some embodiments, each of the one or more genomic regions is selected from Table 8. In some embodiments, each of the one or more genomic regions is selected from Table 9. In some embodiments, each of the one or more genomic regions is selected from Table 10. In some embodiments, each of the one or more genomic regions is selected from Table 11. In some embodiments, each of the one or more genomic regions is selected from Table 12. In some embodiments, each of the one or more genomic regions is selected from Table 13. In some embodiments, each of the one or more genomic regions is selected from Table 14. In some embodiments, each of the one or more genomic regions is selected from Table 15. In some embodiments, each of the one or more genomic regions is selected from Table 16. In some embodiments, each of the one or more genomic regions is selected from Table 17. In some embodiments, each of the one or more genomic regions is selected from Table 18. In some embodiments, each of the one or more genomic regions is selected from Table 19. In some embodiments, each of the one or more genomic regions is selected from Table 20. In some embodiments, each of the one or more genomic regions is selected from Table 21. In some embodiments, each of the one or more genomic regions is selected from Table 22. In some embodiments, each of the one or more genomic regions is selected from Table 23. In some embodiments, each of the one or more genomic regions is selected from Table 24.

[0232] In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions of one or more of Tables 1 to 24. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions of one or more of Tables 2 to 10 or 16 to 24. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 1. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 2. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 3. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 4. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 5. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 6.In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 7. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 8. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 9. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 10. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 11. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 12. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 13. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 14.In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 15. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 16. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 17. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 18. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 19. In some embodiments, an entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 20. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 21. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 22.In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 23. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from the plurality of cfDNA molecules corresponding to or derived from at least 30%, 40%, 50%, 60%, 70%, 80%, 90%, or 95% of the plurality of genomic regions in Table 24.

[0233] In some embodiments, the entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from cfDNA molecules corresponding to or derived from at least 500, 1000, 5000, 10,000, or 15,000 genomic regions of Table 1 or one or more of Tables 11 to 15. In some embodiments, the entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from cfDNA molecules corresponding to or derived from at least 50, 60, 70, 80, 90, 100, 120, 150, or 200 genomic regions of Tables 2 to 10 or one or more of Tables 16 to 24. In some embodiments, the entirety of the plurality of probes on the assay plate is configured to hybridize to a plurality of modified fragments obtained from cfDNA molecules corresponding to or derived from at least 500, 1000, 5000, 10,000, or 15,000 genomic regions of Table 13. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from cfDNA molecules corresponding to or derived from at least 500, 1000, 5000, 10,000, or 15,000 genomic regions in Table 14. In some embodiments, an entirety of the plurality of probes on the assay plate are configured together to hybridize to a plurality of modified fragments obtained from cfDNA molecules corresponding to or derived from at least 500, 1000, 5000, 10,000, or 15,000 genomic regions in Table 15.

[0234] In some embodiments, an assay plate further comprises a plurality of virus-specific probes, wherein each of the plurality of virus-specific probes is configured to hybridize with a viral genome fragment. The plurality of probes can be configured to hybridize with a viral genome fragment from a virus strain associated with cancer. In some embodiments, the viral genome fragment is from MCV, EBV, HBV, HCMV, HCV, HHV5, HPV16 or HPV18. In some embodiments, the assay plate comprises at least 50, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 1500, 2000, 2500 or 3000 virus-specific probes.

[0235] Methods for selecting target genomic regions

[0236] In another aspect, methods are provided for selecting multiple target genomic regions for diagnosing cancer and / or tissue of origin. The multiple target genomic regions can be used to design and manufacture multiple probes for a cancer assay panel. The methylation status of DNA or cfDNA molecules corresponding to or derived from the multiple target genomic regions can be screened by targeted sequencing using the cancer assay panel. Alternative methods, such as WGBS or other methods known in the art, can also be implemented to detect the methylation status of multiple DNA molecules or fragments corresponding to or derived from the multiple target genomic regions.

[0237] Sample processing

[0238] For the selection of multiple target genomic regions, a nucleic acid sample (DNA or RNA) extracted from one or more subjects is used. In the disclosure of the present invention, unless otherwise stated, DNA and RNA can be used interchangeably. That is, the embodiments described herein can be applied to both DNA and RNA types of nucleic acid sequences. However, for the purpose of clarity and explanation, the examples described herein can focus on DNA. The sample comprises any subset of the human genome, any subset of the human genome comprising the entire genome. The sample may comprise blood, plasma, serum, urine, feces, saliva, other types of body fluids, or any combination thereof. In some embodiments, the method for extracting a blood sample (e.g., a syringe or finger prick) may be less invasive than the procedure for obtaining a tissue biopsy (which may require surgery). The extracted sample may comprise cfDNA and / or ctDNA. For healthy individuals, the human body may naturally clear cfDNA and other cellular debris. If a subject has a cancer or disease, the ctDNA in the extracted sample may be at a detectable level for diagnosis.

[0239] The plurality of cfDNA fragments are treated to convert unmethylated cytosine to uracil. In one embodiment, the method uses bisulfite treatment of DNA to convert unmethylated cytosine to uracil without converting methylated cytosine. For example, commercial kits for bisulfite conversion, such as EZ-DNA Methylation TM -Gold, EZ-DNA Methylation TM -Direct or EZ-DNA Methylation TM -Lightning kit (available from Zymo Research, Irvine, CA). In another embodiment, the conversion of unmethylated cytosine to uracil is accomplished using an enzymatic reaction. For example, the conversion can be performed using a commercially available kit for converting unmethylated cytosine to uracil, such as APOBEC-Seq (NEBiolabs, Ipswich, MA).

[0240] A sequencing library is prepared using the converted cfDNA fragments. In a first step, an ssDNA adapter is added to the 3′-OH end of a bisulfite-converted ssDNA molecule using an ssDNA ligation reaction. In one embodiment, the ssDNA ligation reaction uses CircLigase II (Epicentre) to ligate the ssDNA adapter to the 3′-OH end of a bisulfite-converted ssDNA molecule, wherein the 5′-end of the adapter is phosphorylated and the bisulfite-converted ssDNA has been dephosphorylated (i.e., the 3′ end has a hydroxyl group). In another embodiment, the ssDNA ligation reaction uses Thermostable 5′AppDNA / RNA ligase (available from New England BioLabs, Ipswich, Massachusetts) to ligate the ssDNA adapter to the 3′-OH end of a bisulfite-converted ssDNA molecule. In this example, the first UMI adapter is adenylated at the 5′ end and blocked at the 3′ end. In another embodiment, the ssDNA ligation reaction uses T4 RNA ligase (available from New England BioLabs) to ligate the ssDNA adapter to the 3′-OH end of a bisulfite-converted ssDNA molecule. In a second step, a second strand of DNA is synthesized in an extension reaction. For example, an extension primer that hybridizes to a primer sequence contained in the ssDNA adapter is used in a primer extension reaction to form a double-stranded bisulfite-converted DNA molecule. Alternatively, in one embodiment, the extension reaction uses an enzyme that can read multiple uracil residues in the bisulfite-converted template strand. Optionally, in a third step, a dsDNA adapter is added to the double-stranded bisulfite-converted DNA molecule. Finally, the double-stranded bisulfite-converted DNA is amplified to add multiple sequencing adapters. For example, PCR amplification using a forward primer containing a P5 sequence and a reverse primer containing a P7 sequence is used to add P5 and P7 sequences to the bisulfite-converted DNA. Alternatively, during library preparation, multiple unique molecular identifiers (UMIs) can be added to multiple nucleic acid molecules (e.g., multiple DNA molecules) through adapter ligation. The multiple unique molecular identifiers (UMIs) are multiple short nucleic acid sequences (e.g., 4-10 base pairs) that are added to the ends of multiple DNA fragments during adapter ligation.In some embodiments, UMIs are degenerate base pairs that serve as unique tags to identify sequence reads originating from a specific DNA fragment. During adapter ligation after PCR amplification, these unique molecular identifiers are replicated along with the attached DNA fragments, providing a method for identifying sequence reads originating from the same original fragment in downstream analyses.

[0241] Multiple sequence reads are generated from the DNA sequence. The method may comprise a next generation sequencing (NGS) technology, including synthesis technology (Illumina), pyrosequencing (454 Life Sciences), ion semiconductor technology (IonTorrent sequencing), single molecule real-time sequencing (Pacific Biosciences), sequencing by ligation (SOLiD sequencing), nanopore sequencing (Oxford Nanopore Technologies), or paired-end sequencing. In some embodiments, massively parallel sequencing is performed by using sequencing-by-synthesis with multiple reversible dye terminators.

[0242] A plurality of methylation state vectors are then generated from the plurality of sequence reads. To do this, a sequence read is aligned with a reference genome. The reference genome helps provide a position in a human genome from which cfDNA fragments are derived. In a simplified example, the sequence reads are aligned so that three CpG sites are associated with CpG sites 23, 24, and 25 (arbitrary reference identifiers used for ease of description). After alignment, there is information about the methylation status of all CpG sites on the cfDNA fragment and the position of the plurality of CpG sites in the human genome. Based on the methylation status and position, a methylation state vector of the cfDNA fragment can be generated.

[0243] Data structure generation

[0244] Figure 3A1 is a flow chart illustrating a process 300 for generating a data structure for a healthy control group (e.g., multiple reference samples), according to one embodiment. To create the healthy control group data structure, the analysis system obtains information related to the methylation status of multiple CpG sites on multiple sequence reads derived from multiple DNA molecules or fragments from multiple healthy subjects. The methods provided herein for creating a healthy control group data structure can be similarly performed for multiple subjects with cancer, multiple subjects with a source tissue of cancer, multiple subjects with a known cancer type, or multiple subjects with another known disease state. A methylation state vector is generated for each DNA molecule or fragment (e.g., via process 100).

[0245] Using the methylation state vector for each fragment, the analysis system subdivides 310 the methylation state vector into a plurality of strings of CpG sites. In one embodiment, the analysis system subdivides 310 the methylation state vector such that the resulting plurality of strings are all less than a given length. For example, a methylation state vector of length 11 can be subdivided into a plurality of strings of length less than or equal to 3, resulting in 9 strings of length 3, 10 strings of length 2, and 11 strings of length 1. In another example, a methylation state vector of length 7 can be subdivided into a plurality of strings of length less than or equal to 4, resulting in 4 strings of length 4, 5 strings of length 3, 6 strings of length 2, and 7 strings of length 1. If a methylation state vector is less than or equal to a specified string length, the methylation state vector can be converted into a single string containing all CpG sites of the vector.

[0246] The analysis system tallies 320 the plurality of strings by calculating, for each possible CpG site and the likelihood of multiple methylation states in the vector, the number of strings in the control group that have a particular CpG site as the first CpG site in the string and have a methylation state likelihood. For example, at a given CpG site, given a string length of 3, there are 2^3 or 8 possible string configurations. At that given CpG site, for each of the 8 possible string configurations, the analysis system tallies 320 the number of occurrences of each methylation state vector likelihood that appears in the control group. Continuing with this example, this may involve tallying the following quantities: <M x 、M x+1 、M x+2 >、 <M x 、M x+1 、U x+2 >、…、 x 、U x+1 、U x+2 ​>(for each starting CpG site x in the reference genome). The analysis system creates 330 a data structure that stores the calculated counts of each starting CpG site and string likelihood.

[0247] Setting an upper limit on string length has several advantages. First, the size of the data structures created by the analysis system can increase dramatically in size depending on the maximum length of a string. For example, a maximum string length of 4 means that each CpG site has at least 2 4 digits are used to operate on strings of length 4. Increasing the maximum string length to 5 means that each CpG site has an additional 2 4 Or 16 numbers to operate on, which doubles the number of numbers to operate on (and the computer memory required) compared to the previous string length. Reducing the string size helps maintain reasonable data structure creation and performance (e.g., for later access as described below), both in terms of computation and storage. A second statistical consideration for limiting the maximum string length is to avoid overfitting downstream models that use string counts. If the number of CpG sites in a long string does not (biologically) have a strong impact on the outcome (e.g., predicting abnormality versus the presence of cancer), basing probabilities on the number of CpG sites in a large string can be problematic because it requires a large amount of data that may not be available and, therefore, is too sparse for a model to perform properly. For example, calculating a probability of abnormality / cancer conditional on the first 100 CpG sites would require multiple string counts in a data structure of length 100, ideally some of which exactly match the first 100 methylation states. If only sparse counts for strings of length 100 are available, there is insufficient data to determine whether a given string of length 100 in a test sample is abnormal.

[0248] Validation of data structures

[0249] Once the data structure is created, the analysis system may attempt to validate 340 the data structure and / or any downstream models that use the data structure. One type of validation checks for consistency within the data structure of the control group. For example, if there are any outlier subjects, samples, and / or fragments in a control group, then the analysis system may perform various calculations to determine whether to exclude any fragments from those categories. In a representative example, a healthy control group may contain a sample that is not diagnosed but is cancerous, such that the sample contains multiple fragments that are abnormally methylated. This first type of validation ensures that potentially cancerous samples are removed from the healthy control group, thereby not affecting the purity of the control group.

[0250] A second type of validation check is used to calculate multiple p-values ​​for the probability model using the data structure itself (i.e., counts from the healthy control group). Figure 5 A procedure is described for calculating a p-value. Once the analysis system generates a p-value for multiple methylation state vectors in the validation group, the analysis system uses the multiple p-values ​​to construct a cumulative density function (CDF). Using the CDF, the analysis system can perform various calculations on the CDF to verify the data structure of the control group. One test uses the fact that the CDF should ideally be at or below an identity function, such that CDF(x)≤x. Conversely, being above the identity function indicates that there are some flaws in the probability model used for the control group data structure. For example, if 1 / 100 of the fragments have a p-value score of 1 / 1000, meaning CDF(1 / 1000)=1 / 100>1 / 1000, then the second type of verification fails, indicating that there is a problem with the probability model.

[0251] A third type of validation uses a healthy set of validation samples separated from those used to build the data structure to test whether the data structure is built correctly and whether the model is valid. Figure 3B An example procedure for performing this type of validation is described. The third type of validation can quantify how well the healthy control group generalizes the distribution of multiple healthy samples. If the third type of validation fails, then the healthy control group does not generalize the healthy distribution well.

[0252] A fourth type of validation test involves multiple samples from an unhealthy validation group. The analysis system calculates multiple p-values ​​and constructs a CDF for the unhealthy validation group. For an unhealthy validation group, the analysis system expects to see CDF(x) > x for at least some samples. Alternatively, the analysis system expects opposite results for the second and third types of validation for the healthy control and healthy validation groups. If the fourth type of validation fails, this indicates that the model is not appropriately identifying the anomaly it was designed to identify.

[0253] Figure 3B Verification is described according to an embodiment Figure 3AFlowchart of the additional step 340 of verifying the data structure for the control group. In this embodiment of step 340 of validating the data structure, the analysis system performs the fourth type of validation test described above, utilizing a validation group having a plurality of subjects, samples, and / or fragments that are hypothetically similar in composition to the control group. For example, if the analysis system selects a plurality of healthy subjects without cancer for the control group, the analysis system also uses a plurality of healthy subjects without cancer in the validation group.

[0254] The analysis system uses the validation set and generates a set of more than 100 methylation state vectors, such as Figure 3A The analysis system performs a p-value calculation on each methylation state vector from the validation set. The p-value calculation program combines Figure 4 and Figure 5 Further described. For each likelihood of a methylation state vector, the analysis system calculates a probability from the data structure of the control group. Once the probabilities are calculated for the multiple possibilities of the methylation state vector, the analysis system calculates 350 a p-value score for the methylation state vector based on the calculated multiple probabilities. The p-value score represents the expected value of finding a particular methylation state vector in the control group with a lower probability than other possible methylation state vectors. Therefore, a low p-value score generally corresponds to a methylation state vector that is relatively undesirable compared to other methylation state vectors in the control group, wherein a high p-value score generally corresponds to a methylation state vector that is relatively more desirable than other methylation state vectors found in the control group. Once the analysis system generates a p-value score for the multiple methylation state vectors in the validation group, the analysis system constructs 360 a cumulative density function (CDF) using the p-value scores from the validation group. The analysis system verifies 370 the consistency of the CDF in the fourth type of validation test described above.

[0255] Abnormally methylated fragments

[0256] Figure 4 is a flow chart describing a process 400 for identifying a plurality of abnormally methylated segments from a subject, according to one embodiment. An example of the process 400 is Figure 5 It is shown intuitively in Figure 4 The description of is further described below. In process 400, the analysis system generates more than 100 methylation state vectors from multiple cfDNA fragments of the subject. The analysis system processes each methylation state vector as follows.

[0257] In some embodiments, the analysis system filters multiple fragments with uncertain states at one or more CpG sites. In these embodiments, the analysis system executes a prediction model to identify multiple fragments that are unlikely to have an abnormal methylation pattern for filtering. For a sample fragment, the prediction model calculates a sample probability of the methylation state vector of the sample fragment occurring compared to the data structure of the healthy control group. The prediction model randomly samples a subset of multiple possible methylation state vectors, and the subset of the multiple possible methylation state vectors includes multiple CpG sites in the methylation state vector of the sample fragment. The prediction model calculates a probability corresponding to each of the multiple sampled possible methylation state vectors. The multiple probability calculations for the methylation state vector of the fragment and the multiple sampled possible methylation state vectors can be calculated according to a Markov chain model, as described in the following section "P-value score calculation". The prediction model calculates a ratio of the multiple sampled possible methylation state vectors corresponding to multiple probabilities that are less than or equal to the sample probability. The prediction model generates an estimated p-value score for the fragment based on the calculated ratio. The prediction model may filter segments corresponding to p-value scores above a threshold, and retain segments corresponding to p-value scores below the threshold.

[0258] In other embodiments, the prediction model can calculate a confidence probability, which is used by the prediction model to determine when to continue or when to terminate sampling. The confidence probability describes the likelihood that the true p-score of the fragment (the calculation of the true p-score will be further described in the subsection entitled "P-score Calculation" below) is lower than a threshold value based on the estimated p-score and the probability of multiple sampled possible methylation state vectors. The prediction model can sample one or more additional possible methylation state vectors while iteratively calculating the estimated p-score and the confidence probability. When the confidence probability is higher than a confidence threshold, the prediction model can terminate sampling.

[0259] For a given methylation state vector, the analysis system enumerates 410 all possibilities of multiple methylation state vectors having the same starting CpG site and the same length (i.e., a set of multiple CpG sites) in the methylation state vector. Since each observed methylation state may be methylated or unmethylated, and each CpG site has only two possible states, the count of different possibilities of the methylation state vector depends on a power of 2, so a methylation state vector of length n will have the same value as 2 of the methylation state vectors. nUsing the fact that the plurality of methylation state vectors include a plurality of uncertain states for one or more CpG sites, the analysis system may enumerate 410 possibilities of the plurality of methylation state vectors considering only the plurality of CpG sites having a plurality of observed states.

[0260] The analysis system calculates 420 the probability of observing each possibility of the methylation state vector of the identified starting CpG site / methylation state vector length by accessing the healthy control group data structure. In one embodiment, calculating the probability of observing a given possibility uses Markov chain probability to model joint probability calculation. Figure 5 The joint probability calculation is described in more detail. In other embodiments, calculation methods other than Markov chain probability are used to determine the probability of each possibility of observing the methylation state vector.

[0261] The analysis system calculates 430 a p-value score for the methylation state vector using the calculated probabilities for each possibility. In one embodiment, this involves identifying the calculated probability corresponding to the likelihood of matching the methylation state vector in question. Specifically, this is the likelihood of having the same set of CpG sites, or similarly, the same starting CpG site and length, as the methylation state vector. The analysis system sums the calculated probabilities for any possibilities with a probability less than or equal to the identified probability to generate the p-value score.

[0262] This p-value represents the probability that the methylation state vector of the observed fragment, or other methylation state vector, is even less likely in the healthy control group. Therefore, a low p-value score generally corresponds to a methylation state vector that is rare in a healthy subject and, relative to the healthy control group, causes the fragment to be labeled as abnormally methylated. A high p-value score is generally associated with a methylation state vector that is expected to be present (in a relative sense) in a healthy subject. If the healthy control group is a non-cancer group, for example, a low p-value indicates that the fragment is abnormally methylated relative to the non-cancer group and, therefore, may indicate the presence of cancer in the test subject.

[0263] As described above, the analysis system calculates a p-value score for each of a plurality of methylation state vectors, each representing a cfDNA fragment in the test sample. To identify which fragments are aberrantly methylated, the analysis system can filter 440 the set of methylation state vectors based on their p-value scores. In one embodiment, filtering is performed by comparing the p-value scores to a threshold and retaining only those fragments below the threshold. This threshold p-value score can be 0.1, 0.01, 0.001, 0.0001, or a similar value.

[0264] Based on the example results of procedure 400, the analysis system can generate a median (range) of 2800 (1500-12000) segments with abnormal methylation patterns for participants who did not have cancer during training, and a median (range) of 3000 (1200-220000) segments with abnormal methylation patterns for participants who had cancer during training. These multiple sets of filtered multiple segments with multiple abnormal methylation patterns can be used for downstream analysis as described in the section entitled "Filtering of Abnormally Methylated Segments" below.

[0265] P-value score calculation

[0266] Figure 5 is a diagram 500 of an example p-value score calculation according to one embodiment. To calculate a p-value score given a test methylation state vector 505, the analysis system takes the test methylation state vector 505 and enumerates 410 possibilities of the methylation state vectors. In this illustrative example, the test methylation state vector 505 is<M23,M24,M25,U26> Since the length of the test methylation state vector 505 is 4, the multiple possibilities of multiple methylation state vectors containing CpG sites 23-26 are 2 4 In a general example, the number of possibilities for multiple methylation state vectors is 2 n , where n is the length of the test methylation state vector, or alternatively the length of a sliding window (described further below).

[0267] The analysis system calculates 420 the probability of the plurality of possibilities listed in the plurality of methylation state vectors 515. Since methylation is conditionally dependent on the methylation status of a plurality of adjacent CpG sites, one method of calculating the probability of observing a given methylation state vector is to use a Markov chain model. Typically, a methylation state vector, such as <S1,S2,…,S n >, where S represents the methylation status, whether it is methylated (denoted as M), unmethylated (denoted as U), or uncertain (denoted as I), with a joint probability, which can be expanded using the chain rule of probability as:

[0268] P( <S1,S2,…,S n >)=P(S n |S1,…,S n-1 )*P(S n-1 |S1,…,S n-2 )*…*P(S2|S1)*P(S1) (1)

[0269] A Markov chain model can be used to make the calculation of the conditional probability of each possibility more efficient. In one embodiment, the analysis system selects a Markov chain order k that corresponds to the number of previous CpG sites in the vector (or window) that need to be considered in the conditional probability calculation, so that the conditional probability is modeled as P(S n |S1,…,S n-1 )~P(S n |S n-k-2 ,…,S n-1 ).

[0270] To calculate the probability of each Markov model of the likelihood of the methylation state vector, the analysis system accesses the control data structure, specifically the counts of CpG sites and states for different strings. To calculate P(M n |S n-k-2 ,…,S n-1 ), the analysis system matches the data structure n-k-2 ,…,S n-1 ,M n >The stored count of the number of strings matched from the data structure is divided by n-k-2 ,…,S n-1 ,M n > and n-k-2 ,…,S n-1 ,U n > is the sum of the stored counts of the number of strings. Therefore, P(M n |S n-k-2 ,…,S n-1 ) is the calculated ratio, which is given by the following formula:

[0271]

[0272] The calculation can also be smoothed by applying a prior distribution. In one embodiment, the prior distribution is a uniform prior, such as Laplace smoothing. For example, a constant is added to the numerator of the above equation and another constant is added to the denominator of the above equation (e.g., twice the constant in the numerator). In other embodiments, an algorithmic technique such as Knesser-Ney smoothing is used.

[0273] ​​​In the illustration, the formula shown above is applied to a test methylation state vector 505 covering positions 23 to 26. Once the calculated probabilities 515 are complete, the analysis system calculates 430 a p-value score 525 that sums to a probability less than or equal to the likelihood that the methylation state vector matches the test methylation state vector 505.

[0274] In embodiments with multiple uncertain states, the analysis system can calculate a p-value score for multiple CpG sites with multiple uncertain states in a methylation state vector of a segment. The analysis system identifies all possibilities that are consistent with all methylation states of the methylation state vector, excluding the uncertain states. The analysis system can assign a probability to the methylation state vector as a sum of the probabilities of the multiple identified possibilities. As an example, the analysis system calculates a methylation state vector<M1,I2,U3> A probability, as the methylation state vector<M1,M2,U3> and<M1,U2,U3> The sum of the probabilities is a sum of the probabilities because the methylation status of CpG sites 1 and 3 is observed and is consistent with the methylation status of the fragment at CpG sites 1 and 3. This method of summing up the CpG sites with uncertain status uses up to 2 i , where i represents the number of uncertain states in the methylation state vector. In another embodiment, a dynamic programming algorithm can be implemented to calculate the probability of a methylation state vector having one or more uncertain states. Advantageously, the dynamic programming algorithm operates in linear computation time.

[0275] In one embodiment, the computational burden of calculating multiple probabilities and / or multiple p-value scores can be further reduced by caching at least some of the calculations. For example, the analysis system can cache probability calculations for multiple possibilities for multiple methylation state vectors (or their windows) in temporary or persistent memory. If other fragments have the same multiple CpG sites, caching multiple possibility probabilities can efficiently calculate p-value scores without recalculating potential possibility probabilities. Similarly, the analysis system can calculate multiple p-value scores for each of the multiple possibilities of multiple methylation state vectors associated with a set of multiple CpG sites by vector (or its window). The analysis system can cache the multiple p-value scores for use in determining p-value scores for other fragments containing the same CpG site. Typically, the p-value scores for the likelihood of multiple methylation state vectors having the same CpG site can be used to determine the p-value scores for the likelihood of different CpG sites from multiple CpG sites in the same set.

[0276] Sliding Window

[0277] In one embodiment, the analysis system uses 435 a sliding window to determine the likelihood of multiple methylation state vectors and calculate a p-value. Rather than enumerating the likelihood and calculating the p-value for the entire methylation state vector, the analysis system enumerates the likelihood and calculates the p-value for a window of sequential CpG sites, where the length of the window (the length of the CpG site) is shorter than the length of at least some of the fragments (otherwise, the window would be meaningless). The window length can be static, user-defined, dynamic, or selected in some other manner.

[0278] When calculating the p-value of a methylation state vector that is larger than the window, the window starts at the first CpG site in the vector and identifies a sequence set of multiple CpG sites from the vector within the window. The analysis system calculates a p-value score for the window (including the first CpG site). The analysis system then "slides" the window to the second CpG site in the vector and calculates another p-value score for the second window. Therefore, for a window size of l and a methylation vector length of m, each methylation state vector will generate m–l+1 p-value scores. After completing the p-value calculation for each part of the vector, the lowest p-value score in all sliding windows is used as the overall p-value score of the methylation state vector. In another embodiment, the analysis system sums up multiple p-value scores of multiple methylation state vectors to generate an overall p-value score.

[0279] Using the sliding window helps reduce the number of possibilities listed in the methylation state vector and their corresponding probability calculations that would otherwise need to be performed. An example probability calculation is Figure 5 As shown, the number of possibilities for the methylation state vector increases exponentially by a factor of 2 with the size of the methylation state vector. In a realistic example, there may be more than 54 CpG sites for multiple fragments. Instead of calculating probabilities for 2^54 (~1.8×10^16) possibilities to generate a single p-value score, the analysis system can instead use a window of size 5 (for example), thereby performing 50 p-value calculations for each of the 50 windows of the methylation state vector for the fragment. Each of the 50 calculations enumerates 2^5 (32) possibilities for the methylation state vector, resulting in a total of 50×2^5 (1.6×10^3) ​​probability calculations. This results in a significant reduction in the number of calculations to be performed, without any significant improvement in the accurate identification of abnormal fragments. This additional step can also be applied when the methylation state vectors of the validation group are used to validate the 240 control group.

[0280] Filtering of abnormally methylated fragments

[0281] In some embodiments, additional filtering steps are performed to identify genomic regions that are diagnostic of cancer or a type or stage of cancer.

[0282] Hypomethylated and hypermethylated fragments

[0283] An additional analysis identified more than 450 hypomethylated fragments or hypermethylated fragments from the filtered set. Hypomethylated or hypermethylated fragments can be defined as fragments of multiple CpG sites of a certain length (e.g., more than 3, 4, 5, 6, 7, 8, 9, 10, etc.), having a high percentage of methylated CpG sites (e.g., more than 80%, 85%, 90% or 95%, or any other percentage in the range of 50% to 100%) or a high percentage of unmethylated CpG sites (e.g., more than 80%, 85%, 90% or 95% or any other percentage in the range of 50% to 100%), respectively. Described below Figures 6A to 6B An example procedure for identifying these hypomethylated or hypermethylated portions of a genome based on a set of aberrantly methylated fragments is shown.

[0284] Probabilistic Model

[0285] According to the second method, multiple abnormal fragments are further filtered using multiple probability models suitable for multiple methylation patterns of cancer types or non-cancer types. It calculates the log-odds ratio of the number of abnormal fragments from a subject that generally represents cancer or a specific type of cancer. The log-odds ratio can be calculated by taking the logarithm of the ratio of the probability of being cancerous to the probability of being non-cancerous (i.e., 1 minus the probability of being cancerous), both of which are determined by the applied 460 classification model.

[0286] In one embodiment of dividing the genome, the analysis system divides the genome into multiple regions in multiple stages. In a first stage, the analysis system divides the genome into multiple blocks of multiple CpG sites. Each block is defined when the interval between two adjacent CpG sites exceeds a certain threshold (e.g., greater than 200bp, 300bp, 400bp, 500bp, 600bp, 700bp, 800bp, 900bp, or 1000bp). From each block, the analysis system subdivides each block into multiple regions of a certain length in a second stage, such as 500bp, 600bp, 700bp, 800bp, 900bp, 1000bp, 1100bp, 1200bp, 1300bp, 1400bp, or 1500bp. The analysis system may further overlap adjacent regions by a percentage of length, such as 10%, 20%, 30%, 40%, 50%, or 60%.

[0287] The analysis system analyzes multiple sequence reads derived from multiple DNA fragments for each region. The analysis system can process multiple samples from tissues and / or high-signal cfDNA. High-signal cfDNA samples can be identified by a binary classification model, by cancer stage, or by other indicators.

[0288] For each cancer type and non-cancer type, the analysis system fits a separate probabilistic model to the plurality of fragments. In one example, each probabilistic model is a mixture model comprising a combination of a plurality of mixture components, each mixture component being an independent site model, wherein the methylation at each CpG site is assumed to be independent of the methylation status at other CpG sites.

[0289] In an alternative embodiment, a calculation is performed for each CpG site. Specifically, a first count is determined as the number of cancerous samples containing an aberrantly methylated DNA fragment overlapping the CpG (cancer_count), and a second count is determined as the total number of samples in the set containing multiple fragments overlapping the CpG (total). Genomic regions can be selected based on quantity, for example, based on a criterion that the number of cancerous samples (cancer_count) containing a DNA fragment overlapping the CpG is positively correlated with the total number of samples in the set containing multiple fragments overlapping the CpG (total).

[0290] The analysis system can further calculate log-likelihood ratios ("R") for a segment, the segment representing segment likelihoods representing cancer given a probability model suitable for each of the plurality of cancer types, a non-cancer type, or a cancer tissue of origin. The two probabilities can be obtained from a probability model suitable for each of the plurality of cancer types and a non-cancer type, the probability model being defined to calculate the likelihood of observing a methylation pattern on a segment given each of the plurality of cancer types and a non-cancer type. For example, the probability model can be suitable for each of the plurality of cancer types and a non-cancer type.

[0291] Figures 11A-11C A diagram showing various cancers of various subjects at different stages, plotted based on the above Figure 4The log-odds ratios of multiple abnormal fragments identified by the described procedure. These underlying data were obtained by whole-genome bisulfite sequencing of CCGA subjects (ClinicalTrial.gov identifier: NCT02889978 (https: / / www.clinicaltrials.gov / ct2 / show / NCT02889978; see Example 3). Briefly, data were obtained from more than 1700 clinically evaluable subjects, of which more than 1400 were screened, including nearly 600 subjects without cancer and just over 800 subjects with cancer. Figure 11A The first graph 1100 in FIG shows all cancer cases at three different levels - non-cancer; stage I / II / II; and stage IV. The log odds ratios for cancer at stage IV are significantly greater than those at stage I / II / II and non-cancer. Figure 11A The second graph 1110 in FIG. 5 shows breast cancer cases in all stages of cancer and non-cancer, with similar increases in the log odds ratios at more advanced stages of cancer. Figure 11B The third diagram 1120 in Figure 1 shows multiple subtypes of breast cancer. Notably, the F1ER2+ and TNBC subtypes are more widely distributed, while the F1R+ / F1ER2- is concentrated around ~1. Figure 11C The fourth graph 1130 shows lung cancer cases in all cancer stages and non-cancer cases, with steady progression in the progressive stage of lung cancer. The fifth graph 1140 shows colorectal cancer cases in all cancer stages and non-cancer cases, again showing steady progression in the progressive stage of colorectal cancer. Figure 11C The sixth diagram 1150 in FIG shows a prostate cancer case in all stages of cancer and non-cancer. This example is different from most of the previous ones, with only stage IV being significantly different from the other stages I / II / II and non-cancer.

[0292] Genomic regions representing cancer and classifiers

[0293] The analysis system identifies 460 a plurality of genomic regions that are indicative of cancer. To identify these information regions, the analysis system calculates information gain for each genomic region, or more specifically, each CpG site describes the ability to discriminate between different outcomes.

[0294] A method for identifying genomic regions that can distinguish between cancer and non-cancer types uses a trained classification model that can be applied to a collection of aberrantly methylated DNA molecules or fragments corresponding to or derived from a cancerous or non-cancerous group. The trained classification model can be trained to recognize any condition of interest that can be identified from a plurality of methylation state vectors.

[0295] In one embodiment, the trained classification model is a binary classifier trained based on multiple methylation states of multiple cfDNA fragments or multiple genomic sequences, the multiple cfDNA fragments or the multiple genomic sequences being from a cohort of subjects with cancer or a cancer type and a healthy subject group without cancer, and then based on multiple methylation state vectors for classifying a test subject with cancer, a cancer type, or without cancer probability. In other embodiments, different classifiers can be trained using multiple subject groups known to have specific cancers (e.g., breast cancer, lung cancer, prostate cancer, etc.), known to have specific tissues of origin of cancer (where the cancer is believed to originate), or known to have different stages of specific cancers (e.g., breast cancer, lung cancer, prostate cancer, etc.). In these embodiments, different classifiers can be trained using multiple sequence reads obtained from multiple samples amplified from tumor cells of multiple subject groups known to have specific cancers (e.g., breast cancer, lung cancer, prostate cancer, etc.). In the classification model, the ability of each genomic region to distinguish between cancer types and non-cancer types is used to rank genomic regions from the most information-rich to the least information-rich in the classification execution. The analysis system can identify multiple genomic regions based on a ranking of information gain in classification between non-cancer types and cancer types. Multiple probes can be designed 470 to target the identified multiple genomic regions.

[0296] Calculating information gain from hypomethylated and hypermethylated fragments indicative of cancer

[0297] According to one embodiment, using multiple segments representing cancer, the analysis system can Figure 6A The process 600 shown trains a classifier by accessing two sample training sets (a non-cancer set and a cancer set) and obtaining 605 a plurality of methylation state vectors for a non-cancer set and a plurality of methylation state vectors for a cancer set containing a plurality of aberrantly methylated segments, such as by step 440 of the process 400 .

[0298] The program determines 610 (for each methylation state vector) whether the methylation state vector is indicative of cancer. Here, multiple fragments indicative of cancer are defined as hypermethylated or hypomethylated if at least some number of CpG sites have a particular state (methylated or unmethylated, respectively) and / or a threshold percentage of multiple sites have the particular state (again methylated or unmethylated, respectively). In one example, multiple cfDNA fragments are identified as hypomethylated or hypermethylated, respectively, if the fragment overlaps with at least 5 CpG sites and at least 80% of its CpG sites are methylated or at least 80% are unmethylated. In an alternative embodiment, the program considers multiple portions of the methylation state vector and determines whether the portion is hypomethylated or hypermethylated, and can distinguish whether the portion is hypomethylated or hypermethylated. This alternative method addresses missing multiple methylation state vectors that are large in size but contain at least one region of dense hypomethylation or hypermethylation. This procedure for defining hypomethylation and hypermethylation can be applied to Figure 4 In step 450, in another embodiment, a plurality of segments representing cancer may be defined based on likelihoods output from a plurality of trained probability models.

[0299] In one embodiment, the program generates a hypomethylation score (P) of 620 for each CpG site in the genome. hypo ) and a high methylation score (P hyper ). To generate any one score at a given CpG site, the classifier takes four counts at that CpG site: (1) a count of multiple vectors (of methylation status) of the cancer set that overlaps with the CpG site, labeled as hypomethylated; (2) a count of multiple vectors of the cancer set that overlaps with the CpG site, labeled as hypermethylated; (3) a count of multiple vectors of the non-cancer set that overlaps with the CpG site, labeled as hypomethylated; and (4) a count of multiple vectors of the non-cancer set that overlaps with the CpG site, labeled as hypermethylated. In addition, the program can normalize these counts for each group to account for differences in group size between the non-cancer and cancer groups. In alternative embodiments, multiple fragments representing cancer are more generally used, and the multiple scores can be more broadly defined as counts of multiple fragments representing cancer at each genomic region and / or CpG site.

[0300] Specifically, in one embodiment, to generate a hypomethylation score of 620 at a given CpG site, the program takes the ratio of (1) to the sum of (1) and (3). Similarly, a hypermethylation score is calculated by taking (2) divided by the ratio of (2) to (4). In addition, these ratios can be calculated using the additional smoothing techniques discussed above. The hypomethylation and hypermethylation scores are related to an estimate of the probability of cancer, given the presence of hypomethylation or hypermethylation in multiple fragments from a cancer collection.

[0301] The program generates 630 a total low methylation score and a total high methylation score for each abnormal methylation state vector. The total high methylation and low methylation scores are determined based on the high methylation and low methylation scores of the multiple CpG sites in the methylation state vector. In one embodiment, the total high methylation score and low methylation score are respectively assigned as the maximum high methylation and low methylation scores of the multiple sites in each state vector. However, in alternative embodiments, the multiple total scores may be based on an average, a median, or other calculation using the high / low methylation scores of the multiple sites in each vector.

[0302] The program 600 ranks 640 the multiple methylation state vectors for the subject by their aggregated low methylation scores and aggregated high methylation scores, resulting in two ranks for each subject. The program selects the aggregated low methylation scores from the low methylation ranking and the aggregated high methylation scores from the high methylation ranking. Based on the selected multiple scores, the classifier generates 650 a single feature vector for each subject. In one embodiment, the multiple scores selected from any ranking are selected in a fixed order, and each generated feature vector is the same for each subject in each of the multiple training sets. As an example, in one embodiment, the classifier selects the first, second, fourth, and eighth aggregated high methylation scores from each ranking, and similarly for each aggregated low methylation score, and writes these scores into the feature vector for the subject.

[0303] The process trains 660 a binary classifier to distinguish between a plurality of feature vectors of a cancer and non-cancer training set. Generally, any of a number of classification techniques can be used. In one embodiment, the classifier is a nonlinear classifier. In a specific embodiment, the classifier is a nonlinear classifier using an L2-regularized kernel logistic regression with a Gaussian radial basis function (RBF) kernel.

[0304] Specifically, in one embodiment, non-cancerous samples or (multiple) different cancer types (n other ) and the number of cancer samples or cancer types (n cancer The probability that a sample is cancerous is then estimated by a score ("S") that is related to n cancer is positively correlated with n other The score can be calculated using the following formula: (n cancer +1) / (n cancer +n other +2)or(n cancer ) / (n cancer +n other ). The analysis system calculates 670 an information gain for each cancer type and each genomic region or CpG site to determine whether the genomic region or CpG site represents cancer. The information gain of multiple training samples with a given cancer type is calculated compared to all other samples. For example, two random variables "abnormal fragment" ("AF") and "cancer type" ("CT") are used. In one embodiment, AF is a binary variable indicating whether there is an abnormal fragment overlapping a given CpG site in a given sample, as determined for the above-mentioned abnormal score / feature vector. CT is a random variable used to indicate whether the cancer is of a specific type. The analysis system calculates the mutual information of CT associated with a given AF. That is, if you know whether there is an abnormal fragment overlapping at a specific CpG site, how much information about the cancer type can be obtained.

[0305] For a given cancer type, the analysis system ranks CpG sites based on how this information is used by its cancer-specificity. This procedure is repeated for all cancer types considered. If a particular region is often aberrantly methylated in multiple training samples of a given cancer, but not in multiple training samples of other cancer types or multiple healthy training samples, then multiple CpG sites that overlap with those aberrant segments tend to have high information gain for the given cancer type. The ranked CpG sites for each cancer type are greedily added (selected) to a selected set of multiple CpG sites based on their ranking for use in the cancer classifier.

[0306] Calculate pairwise information gain to identify cancer-indicating segments from a probabilistic model

[0307] The analysis can be based on the multiple fragments representing cancer identified by the second method under multiple probability models. Figure 6BProcess 680 in the example identifies multiple genomic regions. The analysis system defines 690 a feature vector for each sample, each region, and each cancer type by counting a number of DNA fragments having a calculated log-likelihood ratio that indicates cancer above multiple thresholds, wherein each count is a value in the feature vector. In one embodiment, the analysis system counts the number of fragments present in a sample at a region for each cancer type having a log-likelihood ratio above one or more possible thresholds. The analysis system defines a feature vector for each sample by counting a number of DNA fragments for each genomic region for each cancer type, providing a calculated log-likelihood ratio for fragments above multiple thresholds, wherein each count is a value in the feature vector. The analysis system uses the defined multiple feature vectors to calculate an information score for each genomic region that describes the ability of the genomic region to distinguish between each pair of multiple cancer types. For each pair of multiple cancer types, the analysis system ranks the multiple regions based on the multiple information scores. The analysis system can select multiple regions based on the ranking according to the multiple information scores.

[0308] The analysis system calculates 695 an information score for each region that describes the ability to distinguish between each pair of multiple cancer types. For each pair of different cancer types, the analysis system can designate one type as a positive type and the other type as a negative type. In one embodiment, the ability to distinguish a region between the positive type and the negative type is based on mutual information, calculated using the estimated scores of multiple cfDNA samples of the positive type and the negative type, wherein the feature will not be zero in the final assay, i.e., at least one fragment of this level is sequenced in a target methylation assay. These scores are estimated using the ratios of features observed in healthy cfDNA, high signal cfDNA, and / or tumor samples of each cancer type. For example, if a feature occurs frequently in healthy cfDNA, it will also be estimated to occur frequently in cfDNA of any cancer type and may result in a low information score. The analysis system can select a certain number of regions, such as 1024, from the ranking for each pair of multiple cancer types.

[0309] In other embodiments, the analysis system further identifies multiple regions that are primarily hypermethylated or hypomethylated from the rankings of the multiple regions. The analysis system can load a set of multiple fragments of (multiple) positive types for a region identified as information. The analysis system evaluates whether the loaded multiple fragments are primarily hypermethylated or hypomethylated from the loaded multiple fragments. If the loaded multiple fragments are primarily hypermethylated or hypomethylated, the analysis system can select multiple probes corresponding to the primary methylation pattern. If the loaded multiple fragments are not primarily hypermethylated or hypomethylated, the analysis system can use a mixture of multiple probes to target hypermethylation and hypomethylation. The analysis system can further identify a minimum set of multiple CpG sites of multiple fragments that overlap by more than a certain percentage.

[0310] In other embodiments, the analysis system, after ranking the plurality of regions based on the plurality of information scores, labels each region with the lowest informative ranking across all cancer type pairs. For example, if a region is the tenth most informative region for distinguishing breast cancer from lung cancer, and the fifth most informative region for distinguishing breast cancer from colorectal cancer, the region is given an overall label of "5." The analysis system can design multiple probes starting with the plurality of regions with the lowest labels and simultaneously add multiple regions to the assay panel, for example, until the assay panel size budget is exhausted.

[0311] Off-target genomic regions

[0312] In some embodiments, the plurality of probes targeting the selected plurality of genomic regions are further filtered based on the number of off-target regions 475. This is to screen for probes that pull down too many cfDNA fragments corresponding to or derived from the plurality of off-target genomic regions. Excluding probes with many off-target regions can be valuable because it reduces the off-target rate and increases target coverage for a given amount of sequencing.

[0313] An off-target genomic region is a genomic region that has significant homology to a target genomic region, such that multiple DNA molecules or multiple fragments derived from multiple off-target genomic regions are designed to hybridize and pull down by a probe designed to hybridize to a target genomic region. A non-target genomic region can be a genomic region that aligns with a probe along at least 35bp, 40bp, 45bp, 50bp, 60bp, 70bp or 80bp with a match rate of at least 80%, 85%, 90%, 95% or 97%. In one embodiment, an off-target genomic region is a genomic region that aligns with a probe along at least 45bp with a match rate of at least 90%. Various methods known in the art can be used to screen out multiple target genomic regions.

[0314] Thoroughly searching the genome to find all off-target genomic regions may be computationally challenging. In one embodiment, a k-mer (substring of k bases) inoculation strategy (one or more mismatches may be allowed) is combined into a local alignment at multiple seed positions. In this case, a thorough search for good alignment can be guaranteed based on the k-mer length, the number of mismatches allowed, and the k-mer seed hit count of a specific position. This requires dynamic programming of local alignments at a large number of positions, so this method is highly optimized to use multiple vector CPU instructions (e.g., AVX2, AVX512), and can be parallelized on multiple cores within a machine, or can be parallelized on multiple machines via a network connection. Those of ordinary skill in the art will recognize that multiple modifications and variations can be made to this method to identify multiple genomic regions that have missed their target.

[0315] In some embodiments, multiple off-target genomic regions or multiple DNA molecules corresponding to or derived from multiple off-target genomic regions having sequence homology with multiple probes that are greater than a threshold number are excluded (or filtered). For example, multiple off-target genomic regions or multiple off-target regions corresponding to or derived from multiple off-target genomic regions having sequence homology with multiple probes that are greater than 30, greater than 25, greater than 20, greater than 18, greater than 15, greater than 12, greater than 10, or greater than 5 off-target regions are excluded.

[0316] In some embodiments, probes are divided into 2, 3, 4, 5, 6 or more separate groups depending on the number of off-target regions. For example, probes with no off-target regions or no DNA molecules corresponding to or derived from multiple off-target regions with sequence homology to multiple probes are assigned to a high-quality group, probes with 1 to 18 off-target regions or DNA molecules corresponding to or derived from 1 to 18 off-target regions with sequence homology to multiple probes are assigned to a low-quality group, and probes with more than 19 off-target regions or DNA molecules corresponding to or derived from 19 target regions with sequence homology to multiple probes are assigned to a poor-quality group. Other cutoff values ​​can be used for grouping.

[0317] In some embodiments, multiple probes in the lowest quality group are excluded. In some embodiments, multiple probes in groups other than the highest quality group are excluded. In some embodiments, separate assay plates are made for multiple probes in each group. In some embodiments, all probes are used on the same assay plate, but separate analyses are performed based on the assigned groups.

[0318] In some embodiments, an assay plate comprises a greater number of high-quality probes than probes in a lower group. In some embodiments, an assay plate comprises a smaller number of low-quality probes than the number in the other groups. In some embodiments, more than 95%, 90%, 85%, 80%, 75%, or 70% of the probes in an assay plate are high-quality probes. In some embodiments, less than 35%, 30%, 20%, 10%, 5%, 4%, 3%, 2%, or 1% of the probes in an assay plate are low-quality probes. In some embodiments, less than 5%, 4%, 3%, 2%, or 1% of the probes in an assay plate are low-quality probes. In some embodiments, an assay plate comprises no low-quality probes.

[0319] In some embodiments, probes with less than 50%, less than 40%, less than 30%, less than 20%, less than 10%, or less than 5% are excluded. In some embodiments, an assay plate selectively includes probes with greater than 30%, greater than 40%, greater than 50%, greater than 60%, greater than 70%, greater than 80%, or greater than 90%.

[0320] How to use the cancer panel

[0321] On the other hand, various methods of using a cancer assay plate are provided. The various methods may include the steps of treating multiple DNA molecules or multiple fragments to convert unmethylated cytosine to uracil (e.g., using bisulfite treatment), applying a cancer assay plate (as described herein) to the converted multiple DNA molecules or multiple fragments, amplifying a subset of the converted multiple DNA molecules or multiple fragments that hybridize (or bind) to multiple probes in the assay plate, and sequencing the amplified multiple cfDNA fragments. The step of applying the cancer assay plate to the converted multiple DNA molecules or multiple fragments is performed when the converted multiple DNA molecules or multiple fragments can bind to the multiple probes of the cancer assay plate. Therefore, the converted multiple DNA molecules or multiple fragments bound to the multiple probes can be selectively separated. In some embodiments, the multiple sequence reads can be compared with a reference genome (e.g., a human reference genome), allowing identification of multiple methylation states at multiple CpG sites within the multiple DNA molecules or multiple fragments, thereby providing information relevant to cancer diagnosis.

[0322] Sample processing

[0323] Figure 7A Flowchart of a method for preparing a nucleic acid sample for analysis according to one embodiment. The method includes, but is not limited to, the following steps. For example, any step of the method may include a quantification substep for quality control or other laboratory assay procedures known to those skilled in the art.

[0324] In step 105, a nucleic acid sample (DNA or RNA) is extracted from a subject. In the disclosure of the present invention, unless otherwise indicated, DNA and RNA can be used interchangeably. That is, the embodiments described herein can be applicable to both DNA and RNA types of nucleic acid sequences. However, for the purpose of clarity and explanation, the examples described herein may focus on DNA. The sample may comprise any subset of the human genome, including the entire genome. The sample may comprise blood, plasma, serum, urine, feces, saliva, other types of body fluids, or any combination thereof. In some embodiments, various methods for extracting a blood sample (e.g., a syringe or finger prick) may be less invasive than procedures for obtaining a tissue biopsy (which may require surgery). The extracted sample may comprise cfDNA and / or ctDNA. For many healthy individuals, the human body may naturally clear cfDNA and other cell fragments. If a subject has a cancer or disease, the ctDNA in the extracted sample may be at a detectable level for diagnosis.

[0325] In step 110, the plurality of cfDNA fragments are treated to convert unmethylated cytosine to uracil. In one embodiment, the method uses bisulfite treatment of DNA to convert unmethylated cytosine to uracil without converting methylated cytosine. For example, commercial kits for bisulfite conversion, such as EZ-DNA Methylation TM -Gold, EZ-DNA Methylation TM -Direct or EZ-DNA Methylation TM -Lightning kit (available from Zymo Research Corp., Irvine, CA). In another embodiment, the conversion of unmethylated cytosine to uracil is accomplished using an enzymatic reaction. For example, the conversion can be performed using a commercially available kit for converting unmethylated cytosine to uracil, such as APOBEC-Seq (NEBiolabs, Ipswich, MA).

[0326] In step 115, a sequencing library is prepared. In a first step, a ssDNA adapter is added to the 3'-OH end of a bisulfite-converted ssDNA molecule using an ssDNA ligation reaction. In one embodiment, the ssDNA ligation reaction uses CircLigase II (Epicentre) to ligate the ssDNA adapter to the 3'-OH end of a bisulfite-converted ssDNA molecule, wherein the 5'-end of the adapter is phosphorylated and the bisulfite-converted ssDNA has been dephosphorylated (i.e., the 3'-end has a hydroxyl group). In another embodiment, the ssDNA ligation reaction uses Thermostable 5'App DNA / RNA ligase (available from New England BioLabs, Ipswich, MA) to ligate the ssDNA adapter to the 3'-OH end of a bisulfite-converted ssDNA molecule. In this example, the first UMI adapter is adenylated at the 5' end and blocked at the 3' end. In another embodiment, the ssDNA ligation reaction uses T4 RNA ligase (available from New England BioLabs) to ligate the ssDNA adapter to the 3'-OH end of a bisulfite-converted ssDNA molecule. In a second step, a second-strand DNA is synthesized in an extension reaction. For example, an extension primer that hybridizes to a primer sequence contained in the ssDNA adapter is used in a primer extension reaction to form a double-stranded bisulfite-converted DNA molecule. Alternatively, in one embodiment, the extension reaction uses an enzyme capable of reading multiple uracil residues in the bisulfite-converted template strand. Optionally, in a third step, a dsDNA adapter is added to the double-stranded bisulfite-converted DNA molecule. Finally, the double-stranded bisulfite-converted DNA is amplified to add multiple sequencing adapters. For example, PCR amplification using a forward primer containing a P5 sequence and a reverse primer containing a P7 sequence is used to add P5 and P7 sequences to the bisulfite-converted DNA. Alternatively, during library preparation, multiple unique molecular identifiers (UMIs) can be added to multiple nucleic acid molecules (e.g., multiple DNA molecules) through adapter ligation. The multiple unique molecular identifiers (UMIs) are multiple short nucleic acid sequences (e.g., 4-10 base pairs) that are added to the ends of multiple DNA fragments during adapter ligation.In some embodiments, UMIs are degenerate base pairs that serve as unique tags to identify sequence reads originating from a specific DNA fragment. During adapter ligation after PCR amplification, these unique molecular identifiers are replicated along with the attached DNA fragments, providing a method for identifying sequence reads originating from the same original fragment in downstream analyses.

[0327] In step 120, multiple target DNA sequences can be amplified from the library (e.g., by hybridization). Any suitable amplification method can be used. For example, in some embodiments, a target assay plate assay is performed on (e.g., contacted with) multiple samples. During amplification, multiple hybridization probes (also referred to herein as "multiple probes") can be used to target and pull down multiple nucleic acid fragments that can provide information about the presence or absence of cancer (or disease), cancer status, or cancer classification (e.g., cancer type or tissue of origin). For a given workflow, the multiple probes can be designed to anneal (or hybridize) to a target (complementary) strand of DNA or RNA. The target strand can be a "positive" strand (e.g., a strand that is transcribed into mRNA and subsequently converted into protein) or a complementary "negative" strand. The multiple probes can be 10s, 100s, or 1000s of base pairs in length. In addition, the multiple probes can cover multiple overlapping portions of a target region.

[0328] In some examples, multiple primers can be used to specifically amplify multiple targets / biomarkers of interest (e.g., by PCR), thereby amplifying multiple targets / biomarkers required for the sample (hybridization capture may be selected without the need for such amplification). For example, forward and reverse primers can be prepared for each genomic region of interest and used to amplify multiple fragments corresponding to or derived from the desired genomic region. Therefore, although the present invention pays particular attention to a variety of cancer assay panels and a variety of bait groups, the scope of the present invention is broad enough to include other methods for amplifying cell-free DNA. Therefore, those skilled in the art will recognize, with the benefit of the present invention, that methods similar to those described herein in connection with hybridization capture can be achieved by replacing hybridization capture with some other amplification strategies, such as PCR amplification of multiple fragments of cell-free DNA corresponding to multiple genomic regions of interest. In some embodiments, bisulfite padlock probe capture is used to amplify multiple regions of interest, as described in Zhang et al. (U.S. Patent Application Publication No. US2016 / 0340740). In some embodiments, additional or alternative methods are used for amplification (e.g., non-target amplification), such as reduced bisulfite sequencing, methylated restriction enzyme sequencing, methylated DNA immunoprecipitation sequencing, methylated CpG binding domain protein sequencing, methylated DNA capture sequencing, or droplet PCR.

[0329] After pull-down and / or hybridization (see step 120), the hybridized multiple nucleic acid fragments can also be optionally amplified using PCR (amplification 125). For example, multiple target sequences can be amplified to obtain multiple amplified sequences that can then be sequenced. In general, any method known in the art can be used to separate and amplify multiple target nucleic acids hybridized to the probe. For example, as is known in the art, a biotin moiety can be added to the 5' end of the probe (i.e., biotinylation) to facilitate the separation of multiple target nucleic acids hybridized to the probe using a streptavidin-coated surface (e.g., streptavidin-coated beads). The multiple nucleic acid fragments are applied to an assay plate containing multiple probes under conditions that allow the multiple nucleic acid fragments to specifically bind to the multiple complementary probes. Therefore, it is possible to selectively separate and amplify multiple nucleic acid fragments that have a high affinity for the multiple probes.

[0330] In step 130, a plurality of sequence reads are generated from the amplified plurality of DNA sequences (e.g., a plurality of amplified sequences). Sequencing data can be obtained from the amplified plurality of DNA sequences by means known in the art. For example, the method can include next-generation sequencing (NGS) technology, including synthesis technology (Illumina), pyrosequencing (454 Life Sciences), ion semiconductor technology (Ion Torrent sequencing), single-molecule real-time sequencing (Pacific Biosciences), sequencing by ligation (SOLiD sequencing), nanopore sequencing (Oxford Nanopore Technologies), or paired-end sequencing. In some embodiments, massively parallel sequencing is performed by using sequencing-by-synthesis with a plurality of reversible dye terminators.

[0331] Sequence read analysis

[0332] In some embodiments, the plurality of sequence reads can be aligned with a reference genome using methods known in the art to determine alignment position information. The alignment position information can represent a starting position and an ending position in the reference genome, wherein the starting position and the ending position in the reference genome correspond to a starting nucleotide base and an ending nucleotide base of a given sequence read. The alignment position information can also include a sequence read length, which can be determined from the starting position and the ending position. A region in the reference genome may be associated with a gene or a fragment of a gene.

[0333] In various embodiments, a sequence read comprises a read pair represented as R1 and R2. For example, a first read R1 can be sequenced from a first end of a nucleic acid fragment, and a second read R2 can be sequenced from the second end of the nucleic acid fragment. Therefore, the nucleotide base pairs of the first read R1 and the second read R2 can be aligned consistently (e.g., in opposite directions) with a plurality of nucleotide bases of the reference genome. The alignment position information derived from the read pairs R1 and R2 can include a starting position (e.g., R1) corresponding to the end of a first read in the reference genome and an end position (e.g., R2) corresponding to the end of a second read in the reference genome. In other words, the starting position and the end position in the reference genome represent possible positions corresponding to the nucleic acid fragments in the reference genome. Output files with SAM (sequence alignment mapping) format or BAM (binary alignment mapping) format can be generated and output for further analysis.

[0334] According to the sequence readings, the position and methylation state of each CpG site can be determined based on alignment with a reference genome. In addition, a methylation state vector for each fragment can be generated to specify a position of the fragment in the reference genome (e.g., specified by the position of the first CpG site in each fragment or another similar metric), a number of multiple CpG sites in the fragment, and whether the methylation state of each CpG site in the fragment is methylated (e.g., expressed as M), unmethylated (e.g., expressed as U) or uncertain (e.g., expressed as I). The multiple methylation state vectors can be stored in a temporary or persistent computer memory for subsequent use and processing. In addition, multiple repeated readings or multiple repeated methylation state vectors from a single object can be removed. In another embodiment, it can be determined that a certain fragment has one or more CpG sites with uncertain methylation states. Such fragments can be excluded from subsequent processing or selectively included in downstream data models to illustrate such uncertain methylation states.

[0335] Figure 7B According to one embodiment Figure 7A 1 illustrates a process 100 for sequencing a cfDNA fragment to obtain a methylation state vector. As an example, the analysis system uses a cfDNA fragment 112. In this example, the cfDNA fragment 112 contains three CpG sites. As shown, the first and third CpG sites of the cfDNA fragment 112 are methylated 114. During processing step 120, the cfDNA fragment 112 is converted to produce a converted cfDNA fragment 122. During processing 120, the second unmethylated CpG site has its cytosine converted to uracil. However, the first and third CpG sites are not converted.

[0336] After conversion, a sequencing library 130 is prepared and sequenced 140 to generate a sequence read 142. The analysis system aligns 150 the sequence read 142 with a reference genome 144. The reference genome 144 provides the position in the human genome where the cfDNA fragment originates. In this simplified example, the analysis system aligns 150 the sequence reads so that three CpG sites are associated with CpG sites 23, 24, and 25 (arbitrary reference identifiers used for ease of description). The analysis system therefore generates information on the methylation status of all CpG sites on the cfDNA fragment 112, as well as information on the location of the CpG site map in the human genome. As shown in the figure, the methylated CpG sites on the sequence read 142 are read as cytosine. In this example, cytosine appears in the sequence read 142, only at the first and third CpG sites, which allows inference that the first and third CpG sites in the original cfDNA fragment are methylated. However, the second CpG site is read as thymine (U is converted to T in the sequencing program), so it can be inferred that the second CpG site is unmethylated in the original cfDNA fragment. Using these two pieces of information, methylation status and position, the analysis system generates 160 a methylation state vector 152 for the fragmented cfDNA 112. In this example, the resulting methylation state vector 152 is <M 23 , U 24 , M 25 >, where M corresponds to a methylated CpG site, U corresponds to an unmethylated CpG site, and the subscript number corresponds to the position of each CpG site in the reference genome.

[0337] Figure 13A and 13B Three diagrams are shown verifying the sequencing consistency of data from a control group. The first diagram 1300 shows the conversion accuracy of unmethylated cytosine to uracil (step 120) on cfDNA fragments obtained from a test sample of multiple subjects at different cancer stages (stage 0, stage I, stage II, stage III, stage IV, and non-cancer). As shown in the figure, there is uniform consistency in converting unmethylated cytosine on multiple cfDNA fragments to uracil. The overall conversion accuracy is 99.47%, with a precision of ±0.024%. The second diagram 1310 shows the average coverage of different stages of cancer. The average coverage for all groups is about 34 times (~34X) the genome coverage of multiple DNA fragments, and only those that are confidently mapped to the genome are counted. The third diagram 1320 ( Figure 13B ) shows the cfDNA concentration of each sample at different cancer stages.

[0338] Cancer diagnosis

[0339] The multiple sequence reads obtained by the methods provided herein can be further processed by a variety of automated algorithms. For example, the analysis system is used to receive sequencing data from a sequencer and perform the processing of various aspects as described herein. The analysis system can be one of a personal computer (PC), a desktop computer, a laptop computer, a notebook computer, a tablet computer, and a mobile device. A computing device can be communicatively coupled to the sequencer via wireless, wired, or a combination of wireless and wired communication technologies. Typically, the computing device is configured with a processor and a memory for storing a plurality of computer instructions that, when executed by the processor, cause the processor to perform the steps described in the remainder of this document. Generally speaking, genetic data and the amount of data generated therefrom are large enough, and the computing power required is so large that it is impossible to perform on paper or with the human mind alone.

[0340] The clinical interpretation of the methylation status of multiple target genomic regions is a process that includes classifying the clinical effect of each methylation state or combination of methylation states and reporting the results in a manner meaningful to medical professionals. The clinical interpretation can be based on comparison of multiple sequence reads with a database specific to cancer or non-cancer subjects and / or based on the number and type of multiple cfDNA fragments with cancer-specific methylation patterns identified from a sample. In some embodiments, multiple target genomic regions are ranked or classified based on the similarity of differential methylation in multiple cancerous samples, and multiple ranks or multiple classifications are used in the interpretation process. The multiple ranks and the multiple classifications can include (1) the type of clinical effect, (2) the strength of the evidence for effect, and (3) the magnitude of the effect. Various methods of clinical analysis and genomic data interpretation can be used for the analysis of sequence reads. In some other embodiments, the clinical interpretation of multiple methylation states of multiple differentially methylated regions can be based on a machine learning method that interprets a current sample based on a classification or regression method, and the machine learning method is trained using multiple methylation states of multiple differentially methylated regions from patients with known cancer status, cancer type, cancer stage, tissue of origin, etc. from cancerous and non-cancerous samples.

[0341] Clinically significant information can generally include the presence or absence of cancer, the presence or absence of certain types of cancer, the stage of cancer, or the presence or absence of other types of disease. In some embodiments, the information relates to the presence or absence of one or more cancer types selected from the group consisting of: (1) blood cancer; (2) breast cancer; (3) colorectal cancer; (4) esophageal cancer; (5) head and neck cancer; (6) hepatobiliary cancer; (7) lung cancer; (8) ovarian cancer; and (9) pancreatic cancer.

[0342] Cancer Grader

[0343] To train a cancer type classifier, the analysis system obtains a plurality of training samples, each having a set of multiple segments representing hypomethylation and hypermethylation of a cancer, for example, identified by step 450 in process 400, and a label of the cancer type of the training sample. The analysis system determines a feature vector (for each training sample) based on the set of multiple segments representing hypomethylation and hypermethylation of a cancer. The analysis system calculates an anomaly score for each CpG site in a plurality of target genomic regions. In one embodiment, the analysis system defines the anomaly score of the feature vector as a binary score based on whether a hypomethylated or hypermethylated segment is present in the set comprising the CpG site. Once all anomaly scores for a training sample are determined, the analysis system determines the feature vector as a vector of multiple elements, including, for each element, one of the anomaly scores associated with one of the multiple CpG sites. The analysis system can normalize the multiple anomaly scores of the feature vector based on sample coverage (i.e., the median or average sequencing depth of all CpG sites).

[0344] The analysis system can train the cancer classifier using multiple feature vectors from multiple training samples. In one embodiment, the analysis system trains a binary cancer classifier based on the multiple feature vectors from the multiple training samples to distinguish between multiple labels, cancer and non-cancer. In this embodiment, the classifier outputs a prediction score representing the likelihood of the presence or absence of cancer. In another embodiment, the analysis system trains a multiclass cancer classifier to distinguish between multiple cancer types. In this multiclass cancer classifier embodiment, the cancer classifier is trained to determine a cancer prediction that includes a prediction value for each of the multiple cancer types being classified. The multiple prediction values ​​may correspond to a likelihood that a given sample has each of the multiple cancer types. For example, the cancer classifier returns a cancer prediction that includes a prediction value for breast cancer, lung cancer, and non-cancer. For example, the cancer classifier can return a cancer prediction for a test sample that includes a prediction score for breast cancer, lung cancer, and / or no cancer. In any embodiment, the analysis system trains the cancer classifier by inputting multiple sets of multiple training samples and their multiple feature vectors into the cancer classifier and adjusting multiple classification parameters so that a function of the classifier can accurately associate the multiple training feature vectors with their corresponding labels. The analysis system can group the multiple training samples into multiple sets of one or more training samples for iterative batch training of the cancer classifier. After inputting all sets of multiple training samples containing their multiple training feature vectors and adjusting multiple classification parameters, the cancer classifier is fully trained to label multiple test samples according to their feature vectors within a certain error range. The analysis system can train the cancer classifier according to any of a variety of methods. For example, the binary cancer classifier can be an L2-regularized logistic regression classifier trained using a log-loss function. As another example, the multi-cancer classifier can be a multinomial logistic regression. In practice, any type of cancer classifier can be trained using other techniques. There are many such techniques, including a variety of potentially usable kernel methods, a variety of machine learning algorithms (such as multi-layer neural networks, etc.), and in particular the methods described in PCT patent application No. PCT / US2019 / 022122 and U.S. patent application No. 16 / 352602, which are incorporated herein by reference in their entirety and may be used in various embodiments.

[0345] In a specific embodiment, a cancer classifier is trained by a program comprising the following steps: a. obtaining sequence information of multiple training fragments from multiple training subjects; b. for each training fragment, determining whether the training fragment is hypomethylated or hypermethylated, wherein each of the multiple hypomethylated and hypermethylated training fragments comprises at least a threshold number of multiple CpG sites, and the multiple CpG sites each have at least a threshold percentage of being unmethylated or methylated; c. for each training subject, generating a training feature vector based on the multiple hypomethylated training fragments and the multiple hypermethylated training fragments; and d. training the model using the multiple training feature vectors from one or more training subjects without cancer and the multiple training feature vectors from one or more training subjects with cancer. The training method may further comprise the following steps: a. obtaining sequence information of a plurality of training segments from a plurality of training objects; b. for each training segment, determining whether the training segment is hypomethylated or hypermethylated, wherein each of the plurality of hypomethylated and hypermethylated training segments comprises a plurality of CpG sites having at least a threshold number, and the plurality of CpG sites respectively have at least a threshold percentage of being unmethylated or methylated; c. for each of a plurality of CpG sites in a reference genome: quantifying a number of a plurality of hypomethylated training segments overlapping with the CpG site and a number of a plurality of hypermethylated training segments overlapping with the CpG site; and generating a hypomethylation score and a hypermethylation score based on the number of the plurality of hypomethylated training segments and the plurality of hypermethylated training segments; d. for each training segment , generating a total low methylation score based on the low methylation scores of the multiple CpG sites in the training segments, and generating a total high methylation score based on the high methylation scores of the multiple CpG sites in the training segments; e. for each training subject: ranking the multiple training segments based on the total low methylation score, and ranking the multiple training segments based on the total high methylation score; and generating a feature vector based on the ranking of the multiple training segments; f. obtaining multiple training feature vectors of one or more training subjects without cancer, and multiple training feature vectors of one or more training subjects with cancer; and g. training the model using the multiple feature vectors of the one or more training subjects without cancer and the multiple feature vectors of the one or more training subjects with cancer. In some embodiments, the model comprises one of a kernel logistic regression classifier, a random forest classifier, a hybrid model, a convolutional neural network, and an autoencoder model.

[0346] In some embodiments, quantifying a number of multiple hypomethylated training segments overlapping with the CpG site and a number of multiple hypermethylated training segments overlapping with the CpG site further comprises the steps of: a. quantifying a number of cancers from multiple hypomethylated training segments from one or more training subjects with cancer overlapping with the CpG site, and quantifying a number of non-cancers from multiple hypomethylated training segments from one or more training subjects without cancer overlapping with the CpG site; and b. quantifying a number of cancers from multiple hypermethylated training segments from one or more training subjects with cancer overlapping with the CpG site, and quantifying a number of non-cancers from multiple hypermethylated training segments from one or more training subjects without cancer overlapping with the CpG site. In some embodiments, generating a low methylation score and a high methylation score based on the number of multiple low methylation training segments and the number of multiple high methylation training segments further includes the steps of: a. for generating the low methylation score, calculating a low methylation ratio of the number of cancers in the multiple low methylation training segments to the low methylation sum of the number of cancers in the multiple low methylation training segments and the number of non-cancers in the multiple low methylation training segments; and b. for generating the high methylation score, calculating a high methylation ratio of the number of cancers in the multiple high methylation training segments to the high methylation sum of the number of cancers in the multiple high methylation training segments and the number of non-cancers in the multiple high methylation training segments.

[0347] During deployment, the analysis system obtains a plurality of sequence reads from a test sample collected from a subject. Various sequencing methods available in the art can be used to obtain the plurality of sequence reads. In some embodiments, the plurality of sequence reads are obtained from whole genome sequencing or targeted sequencing. In some embodiments, the plurality of sequence reads comprises a set of plurality of sequence reads of a modified plurality of test fragments, wherein the modified plurality of test fragments are obtained by processing a set of a plurality of nucleic acid fragments, wherein each of the plurality of nucleic acid fragments corresponds to or is derived from a plurality of genomic regions selected from one or more of Tables 1 to 24. In some embodiments, the plurality of sequence reads are from a plurality of DNA samples amplified using an assay plate as described herein.

[0348] The analysis system processes the multiple sequence reads to obtain a test feature vector in a similar procedure as described for the multiple training samples. In some embodiments, the test feature vector is obtained by a procedure comprising the steps of: a. for each of the multiple nucleic acid fragments, determining whether the nucleic acid fragment is hypomethylated or hypermethylated, wherein each of the multiple nucleic acid fragments that are hypomethylated and hypermethylated contains at least a threshold number of multiple CpG sites, and the multiple CpG sites have at least a threshold percentage of being unmethylated or methylated, respectively; b. for each of the multiple CpG sites in a reference genome: quantifying a number of multiple nucleic acid fragments that are hypomethylated and overlap with the CpG site and a number of multiple nucleic acid fragments that are hypermethylated and overlap with the CpG site. ; and based on the number of multiple low-methylation nucleic acid fragments and multiple high-methylation nucleic acid fragments, generate a low methylation score and a high methylation score; c. for each nucleic acid fragment, generate a total low methylation score based on the low methylation scores of the multiple CpG sites in the nucleic acid fragment and generate a total high methylation score based on the high methylation scores of the multiple CpG sites in the nucleic acid fragment; d. rank the multiple nucleic acid fragments based on the total low methylation scores, and rank the multiple nucleic acid fragments based on the total high methylation scores; and e. generate the test feature vector based on the ranking of the multiple nucleic acid fragments.

[0349] The analysis system then inputs the test feature vector into a trained cancer classifier to generate a cancer prediction, for example, a binary prediction (cancer or non-cancer) or a multi-class cancer prediction (a prediction score for each of multiple cancer types). In some embodiments, the analysis system outputs a cancer probability for the test sample. The cancer probability can be compared to a threshold probability to determine whether the test sample from a subject has cancer or does not have cancer.

[0350] Exemplary sequencers and analysis systems

[0351] Figure 8A 8 is a flow chart illustrating various systems and devices for sequencing multiple nucleic acid samples according to one embodiment. This illustrative flow chart includes multiple devices, such as a sequencer 820 and an analysis system 800. The sequencer 820 and the analysis system 800 can operate in tandem to perform one or more steps of the procedures described herein.

[0352] In various embodiments, the sequencer 820 receives an amplified nucleic acid sample 810. Figure 8AAs shown, the sequencer 820 can include a graphical user interface 825 that allows a user to interact with a plurality of specific tasks (e.g., start sequencing or terminate sequencing) and one or more loading stations 830 for loading a sequencing cartridge (containing a plurality of amplified fragment samples) and / or for loading a plurality of buffers required to perform a sequencing assay. Thus, once a user of the sequencer 820 has provided the required reagents and sequencing cartridge to the loading station 830 of the sequencer 820, the user can initiate sequencing by interacting with the graphical user interface 825 of the sequencer 820. Once initiated, the sequencer 820 performs sequencing and outputs a plurality of sequence reads of the plurality of amplified fragments from the nucleic acid sample 810.

[0353] In some embodiments, the sequencer 820 is communicatively coupled to the analysis system 800. The analysis system 800 includes some computing devices for processing sequence reads for various applications, such as evaluating the methylation status of one or more CpG sites, identifying variants (variant calling), or quality control. The sequencer 820 can provide a plurality of sequence reads in a BAM file format to the analysis system 800. The analysis system 800 can be communicatively coupled to the sequencer 820 via wireless, wired, or a combination of wireless and wired communication technologies. Typically, the analysis system 800 is configured with

[0354] A processor and a non-transitory computer-readable storage medium storing a plurality of computer instructions that, when executed by the processor, cause the processor to process the plurality of sequence reads or perform one or more steps of any method or procedure disclosed herein.

[0355] In some embodiments, the plurality of sequence reads can be aligned to a reference genome using methods known in the art to determine alignment position information, e.g., Figure 3A as part of step 140 of procedure 100 in the preceding text. The alignment position can generally describe a starting position and an ending position of a region in the reference genome, which corresponds to a starting base and an ending base of a given sequence read. Corresponding to methylation sequencing, the alignment position information can be summarized as representing a first CpG site and a last CpG site contained in the sequence read based on the alignment with the reference genome. The alignment position information can further represent multiple methylation states and the positions of all CpG sites in a given sequence read. A region in the reference genome can be associated with a gene or a fragment of a gene; therefore, the analysis system 800 can use one or more genes to align with the sequence read to mark a sequence read. In one embodiment, the fragment length (or size) is determined from the starting position and the ending position.

[0356] In various embodiments, for example, when a paired end sequencing program is used, a sequence read consists of a pair of reads represented as R_1 and R_2. For example, a first read R_1 can be sequenced from a first end of a double-stranded DNA (dsDNA) molecule, and a second read R_2 can be sequenced from a second end of the double-stranded DNA (dsDNA). Therefore, the nucleotide base pairs of the first read R_1 and the second read R_2 can be aligned consistently (e.g., in opposite directions) with multiple nucleotide bases of the reference genome. The alignment position information derived from the read pair R_1 and R_2 can include a starting position in the reference genome, the starting position corresponding to the end point of the first read (e.g., R_1) and an ending position in the reference genome, the ending position corresponding to the end point of a second read (e.g., R_2). In other words, the starting position and the ending position in the reference genome represent possible positions corresponding to the nucleic acid fragments within the reference genome. An output file having a SAM (sequence alignment map) format or a BAM (binary) format can be generated and output for further analysis.

[0357] Now refer to Figure 8B , Figure 8B is a block diagram of an analysis system 800 for processing multiple DNA samples according to one embodiment. The analysis system implements one or more computing devices for analyzing multiple DNA samples. The analysis system 800 includes a sequence processor 840, a sequence database 845, a model database 855, multiple models 850, a parameter database 865, and a scoring engine 860. In some embodiments, the analysis system 800 executes Figure 3A Program 100, Figure 3B Program 340, Figure 4 Program 400, Figure 5 Program 500, Figure 6A Program 600 or Figure 6B One or more steps of procedure 680 and other procedures described herein.

[0358] The sequence processor 840 generates multiple methylation state vectors from multiple fragments of a sample. At each CpG site on a fragment, the sequence processor 840 generates multiple methylation state vectors by Figure 3AThe program 100 generates a methylation state vector for each fragment, which specifies the position of the fragment in the reference genome, the number of multiple CpG sites in the fragment, and the methylation state of each CpG site in the fragment, whether methylated, unmethylated, or uncertain. The sequence processor 840 can store multiple methylation state vectors for multiple fragments in the sequence database 845. The data in the sequence database 845 can be organized so that multiple methylation state vectors from a sample are associated with each other.

[0359] In addition, multiple different models 850 can be stored in the model database 855 or retrieved for use with multiple test samples. In one example, a model is a trained cancer classifier that uses a feature vector from multiple abnormal segments to determine a cancer prediction for a test sample. The training and use of the cancer classifier will be further discussed in conjunction with the section titled "Genomic Regions Representing Cancer and Classifiers." The analysis system 800 can train one or more models 850 and store various trained parameters in the parameter database 865. The analysis system 800 stores the multiple models 850 along with multiple functions in the model database 855.

[0360] During inference, the scoring engine 860 uses one or more 850 models to return multiple outputs. The scoring engine 860 accesses the multiple models 850 in the model database 855 and the multiple trained parameters in the parameter database 865. For each model, the scoring engine receives an appropriate input for the model and computes an output based on the received input, the multiple parameters, and a function of each model that relates the input and output. In some use cases, the scoring engine 860 further computes multiple metrics related to a confidence level for the multiple outputs computed by the model. In other use cases, the scoring engine 860 is used to compute other intermediate values ​​within the model.

[0361] application

[0362] In some embodiments, the methods, analysis systems and / or classifiers of the present invention can be used to detect the presence of cancer, monitor the progression or recurrence of cancer, monitor treatment response or effectiveness, determine the presence of minimal residual disease (MRD) or monitor minimal residual disease or any combination thereof. For example, as described herein, a classifier can be used to generate a likelihood or probability score (e.g., from 0 to 100) that a sample feature vector is from a cancer patient. In some embodiments, the probability score is compared with a threshold probability to determine whether a subject has cancer. In other embodiments, the likelihood or probability score can be evaluated at different time points (e.g., before or after treatment) to monitor disease progression or monitor treatment effectiveness (e.g., therapeutic efficacy). In other embodiments, the likelihood or probability score can be used to make or influence a clinical decision (e.g., cancer diagnosis, treatment selection, treatment effect evaluation, etc.). For example, in one embodiment, if the likelihood or probability score exceeds a threshold, the doctor can prescribe an appropriate treatment plan.

[0363] Early detection of cancer

[0364] In some embodiments, the methods and / or classifiers of the present invention are used to detect the presence or absence of cancer in a subject suspected of having cancer. For example, a classifier (as described herein) can be used to determine a likelihood or probability score that a sample feature vector is from a subject having cancer.

[0365] In one embodiment, a probability score greater than or equal to 60 may indicate that the subject has cancer. In other embodiments, a probability score greater than or equal to 65, greater than or equal to 70, greater than or equal to 75, greater than or equal to 80, greater than or equal to 85, greater than or equal to 90, or greater than or equal to 95 indicates that the subject has cancer. In other embodiments, a probability score may indicate the severity of the disease. For example, a probability score of 80 may indicate a more severe or advanced stage of the cancer compared to a probability score below 80 (e.g., 70). Similarly, an increase in the probability score over time (e.g., at a second, later time point) may indicate progression of the disease, or a decrease in the probability score over time (e.g., at a second, later time point) may indicate successful treatment.

[0366] In another embodiment, a cancer log probability ratio for a test subject can be calculated by taking the logarithm of the ratio of a probability of cancer to a probability of non-cancerous (i.e., 1 minus the probability of cancer), as described herein. According to this embodiment, a cancer log probability ratio greater than 1 can indicate that the subject has cancer. In other embodiments, a cancer log probability ratio greater than 1.2, greater than 1.3, greater than 1.4, greater than 1.5, greater than 1.7, greater than 2, greater than 2.5, greater than 3, greater than 3.5, or greater than 4 indicates that the subject has cancer. In other embodiments, a cancer log probability ratio can indicate the severity of the disease. For example, a cancer log probability ratio greater than 2 may indicate a more severe or advanced stage of the cancer compared to a score less than 2 (e.g., a score of 1). Similarly, an increase in the cancer log probability ratio over time (e.g., at a second, later time point) can indicate progression of the disease, or a decrease in the cancer log probability ratio over time (e.g., at a second, later time point) can indicate successful treatment.

[0367] According to various aspects of the present invention, the methods and systems of the present invention can be trained to detect or classify multiple cancer indications. For example, the methods, systems, and classifiers of the present invention can be used to detect the presence of one or more, two or more, three or more, or five or more different types of cancer.

[0368] In some embodiments, the cancer is one or more of: (1) blood cancer; (2) breast cancer; (3) colorectal cancer; (4) esophageal cancer; (5) head and neck cancer; (6) hepatobiliary cancer; (7) lung cancer; (8) ovarian cancer; and (9) pancreatic cancer.

[0369] Cancer and Treatment Monitoring

[0370] In some embodiments, likelihood or probability scores can be assessed at different time points (e.g., before or after treatment) to monitor disease progression or monitor treatment efficacy (e.g., treatment efficacy). For example, the methods provided herein involve obtaining a first sample (e.g., a first plasma cfDNA sample) from a cancer patient at a first time point, determining a first likelihood or probability score (as described herein) therefrom, obtaining a second test sample (e.g., a second plasma cfDNA sample) from the cancer patient at a second time point, and determining a second likelihood or probability score (as described herein) therefrom.

[0371] In certain embodiments, the first time point is before a cancer treatment (e.g., before resection surgery or therapeutic intervention), and the second time point is after a cancer treatment (e.g., after resection surgery or therapeutic intervention), and a method for monitoring the effectiveness of the treatment. For example, if the second likelihood or probability score is lower than the first likelihood or probability score, the treatment is considered successful. However, if the second likelihood or probability score is higher than the first likelihood or probability score, the treatment is considered unsuccessful. In other embodiments, both the first and the second time points are before a cancer treatment (e.g., before resection surgery or therapeutic intervention). In other embodiments, both the first and the second time points are after a cancer treatment (e.g., before resection surgery or therapeutic intervention), and a method for monitoring the effectiveness of the treatment or the loss of the effectiveness of the treatment. In other embodiments, multiple cfDNA samples can be obtained and analyzed from a cancer patient at the first and second time points. For example, to monitor cancer progression, determine whether the cancer is in remission (e.g., after treatment), monitor or detect residual disease or disease recurrence, or monitor the efficacy of a treatment (e.g., therapeutic).

[0372] Those skilled in the art will readily appreciate that a test sample can be obtained from a cancer patient at any desired time point and analyzed according to the methods of the present invention to monitor the patient's cancer status. In some embodiments, the first time point and the second time point are separated by an amount of time ranging from about 15 minutes to about 30 years, such as about 30 minutes, such as about 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, or about 24 hours, such as about 1, 2, 3, 4, 5, 10, 15, 20, 25, or about 30 days, or such as about 1, 2, 3, 3, 4, 5, 6, 7, 8, 9, 10, 11, or 12 months, or such as about 1, 1.5, 2, 2.5, 3, 3 .5, 4, 4.5, 5, 5, 5.5, 6.5, 6, 6.5, 7, 7.5, 8, 8.5, 9, 9.5, 10, 10.5, 11, 11.5, 12, 12.5, 13, 13.5, 14, 14.5, 15, 15.5, 16, 16.5, 17, 17.5, 18, 18.5, 19, 19.5, 20, 20.5, 21, 21.5, 22, 22.5, 23, 23.5, 24, 24.5, 25, 25.5, 26, 26.5, 27, 27.5, 28, 28.5, 29, 29.5 years or about 30 years. In other embodiments, multiple test samples may be obtained from the patient at least once every 3 months, at least once every 6 months, at least once a year, at least once every 2 years, at least once every 3 years, at least once every 4 years, or at least once every 5 years.

[0373] treat

[0374] In another embodiment, information obtained from any of the methods described herein (e.g., likelihood or probability scores) can be used to make or influence a clinical decision (e.g., cancer diagnosis, treatment selection, treatment efficacy assessment, etc.). For example, in one embodiment, if the likelihood or probability score exceeds a threshold, a doctor can prescribe an appropriate treatment plan (e.g., resection surgery, radiation therapy, chemotherapy, and / or immunotherapy). In some embodiments, information such as a likelihood or probability score can be provided to a doctor or subject as an output value.

[0375] A classifier (as described herein) can be used to determine a likelihood or probability score for a sample feature vector from a subject having cancer. In one embodiment, when the likelihood or probability exceeds a threshold, an appropriate treatment (e.g., resection surgery or treatment) is prescribed. For example, in one embodiment, if the likelihood or probability score is greater than or equal to 60, one or more appropriate treatments are prescribed. In another embodiment, if the likelihood or probability score is greater than or equal to 65, greater than or equal to 70, greater than or equal to 75, greater than or equal to 80, greater than or equal to 85, greater than or equal to 90, or greater than or equal to 95, one or more appropriate treatments are prescribed. In other embodiments, a cancer log probability ratio can indicate the effectiveness of a cancer treatment. For example, an increase in the cancer log probability ratio over time (e.g., at a second time after treatment) can indicate that the treatment was ineffective. Similarly, a decrease in the cancer log probability ratio over time (e.g., at a second time after treatment) can indicate that the treatment was successful. In another embodiment, if the cancer log odds ratio is greater than 1, greater than 1.5, greater than 2, greater than 2.5, greater than 3, greater than 3.5, or greater than 4, one or more appropriate treatments are prescribed.

[0376] In some embodiments, the treatment is one or more cancer therapeutic agents selected from the group consisting of: a chemotherapeutic agent, a targeted cancer therapeutic agent, a differentiation therapeutic agent, a hormone therapeutic agent, and an immunotherapeutic agent. For example, the treatment can be one or more chemotherapeutic agents selected from the group consisting of: alkylating agents, antimetabolites, anthracyclines, antitumor antibiotics, cytoskeleton disruptors (paclitaxel), topoisomerase inhibitors, mitotic inhibitors, corticosteroids, kinase inhibitors, nucleotide analogs, and platinum-based drugs, and any combination thereof. In some embodiments, the treatment is one or more targeted cancer therapeutic agents selected from the group consisting of: signal transduction inhibitors (e.g., tyrosine kinase and growth factor receptor inhibitors), histone deacetylase (HDAC) inhibitors, retinoic acid receptor agonists, proteome inhibitors, angiogenesis inhibitors, and monoclonal antibody conjugates. In some embodiments, the treatment is one or more differentiation therapeutic agents comprising retinoids, such as tretinoin, alitretinoin, and bexarotene. In some embodiments, the treatment is one or more hormonal therapeutics selected from the group consisting of antiestrogen, aromatase inhibitors, progestins, estrogens, antiandrogens, and GnRH agonists or analogs. In one embodiment, the treatment is one or more immunotherapeutics selected from the group consisting of monoclonal antibody therapies such as rituximab (RITUXAN) and alemtuzumab (CAMPATH), nonspecific immunotherapies and adjuvants such as BCG, interleukin 2 (IL-2), and interferon alpha, immunomodulatory drugs such as thalidomide and lenalidomide (REVLIMID). A skilled physician or oncologist has the ability to select an appropriate cancer therapeutic based on characteristics such as the type of tumor, the stage of the cancer, previous exposure to cancer treatments or therapeutic agents, and other characteristics of the cancer.

[0377] Example

[0378] The following examples are set forth in order to provide those skilled in the art with a complete disclosure and description of how to make and use the present specification and are not intended to limit the scope of what the inventors consider to be the description, nor are they intended to represent that the experiments below are all or the only experiments performed. Efforts have been made to ensure accuracy with respect to numbers used (e.g., amounts, temperatures, etc.), but some experimental errors and deviations should be accounted for.

[0379] Example 1: Probe Quality Analysis

[0380] To test how much overlap is required between a cfDNA fragment and a probe to achieve a non-negligible amount of pull-down, different lengths of overlap were tested using multiple assay plates designed to contain three different types of probes (V1D3, V1D4, V1E2), with different overlaps with the specific 175bp target DNA fragment for each probe. The overlap range tested was between 0bp and 120bp. Multiple samples containing multiple 175bp target DNA fragments were applied to the assay plate and washed, and then the multiple DNA fragments bound to the multiple probes were collected. Multiple quantities of the multiple DNA fragments collected were measured and plotted as density versus overlap size, as shown in Figure 2. Figure 9 shown.

[0381] When the overlap was less than 45 bp, there was no significant binding and pull-down of multiple target DNA fragments. These results indicate that a probe overlap of at least 45 bp is generally required to achieve non-negligible pull-down, although this number may vary depending on the assay conditions.

[0382] Furthermore, it was shown that a mismatch rate greater than 10% in the overlapping region between the probe and multiple fragment sequences was sufficient to significantly disrupt binding, thereby reducing pull-down efficiency. Therefore, sequences that can align with the probe along at least 45 bp with a match rate of at least 90% are candidates for off-target pull-down.

[0383] Therefore, we performed an exhaustive search of all genomic regions with a 45bp alignment and a match rate of more than 90% (i.e., off-target regions) for each probe. Specifically, we combined a k-mer seeding strategy (which allows one or more mismatches) with local alignment at multiple seed positions. This ensures that no good alignments are lost based on the k-mer length, the number of mismatches allowed, and the number of k-mer seed hits at a particular position. This involves performing dynamic programming local alignments at a large number of positions, so the implementation is optimized to use vector CPU instructions (e.g., AVX2, AVX512) and parallelize across many cores within a computer and across many computers connected via a network. This allows for an exhaustive search, which is very valuable when designing high-performance assay panels (i.e., low off-target rates and high on-target coverage for a given amount of sequencing).

[0384] After an exhaustive search, each probe is scored based on the number of off-target regions. The best probes have a score of 1, meaning they match only in one place (high Q). Probes with low scores (low Q) between 2-19 are accepted, but poorly scored probes with scores over 20 (poor Q) are discarded. Other cutoff values ​​can be used for specific samples.

[0385] The number of high-quality, low-quality, and poor-quality probes is then counted in multiple hypermethylated genomic regions or multiple hypomethylated genomic regions targeted by multiple probes. Figure 10 As shown, multiple probes targeting multiple hypermethylated regions tend to have fewer off-target regions.

[0386] Example 2: Annotation of target genomic regions

[0387] pass Figure 4 The multiple target genomic regions identified by the procedure outlined in are analyzed to understand multiple characteristics of the multiple target regions. In particular, the selected multiple target genomic regions are aligned to a reference genome to determine multiple alignment positions. Alignment position information is collected for each selected target genomic region, and the alignment position information includes the chromosome number, start base, end base, and genomic annotation for the given genomic region. The multiple target genomic regions are located in introns, exons, intergenic regions, 5'UTRs, 3'UTRs, or control regions, such as promoters or enhancers. The number of multiple target genomic regions that fall within each genomic annotation is counted and plotted in Figure 12 In the illustration provided in . Figure 12 The number of selected multiple target genomic regions (black bars) or randomly selected multiple genomic regions (grey bars) that fell within each genomic annotation were also compared.

[0388] The analysis showed that compared with randomly selected targets of the same size, the selected multiple target genomic regions were not random in their genomic distribution, with a higher degree of amplification of regulatory and functional elements (such as promoters and 5'UTRs) and a lower representation of intergenic sequences. For example, multiple target genomic regions were found to be located in promoters, 5'UTRs, exons, intron / exon boundaries, introns, 3'UTRs or enhancers, but not in intergenic regions.

[0389] Example 3: Cancer Gene Assay Panel (CCGA)

[0390] A database generated by sequencing multiple cfDNA fragments obtained from more than 1,800 individuals was used to select multiple target genomic regions. The cfDNA sequencing database is referred to herein as the Circulating Cell-Free Genome Atlas (CCGA) study. The CCGA study is described in ClinicalTrials.gov, identifier: NCT02889978 (https: / / www.clinicaltrials.gov / ct2 / show / NCT02889978).

[0391] Specifically, multiple cfDNA sequences in the database were screened based on the p-value using the non-cancer distribution, and only multiple fragments with p<0.001 were retained. The selected multiple cfDNAs were further filtered to retain only those that were at least 90% methylated or 90% unmethylated. Next, for each CpG site in the selected fragments, the number of cancer samples or non-cancer samples containing multiple fragments overlapping the CpG site was counted. Specifically, P(cancer|overlapping fragments) was calculated for each CpG, and multiple genomic sites with high P values ​​were selected as general cancer targets. By design, the selected multiple fragments have very low noise (i.e., there are few overlapping non-cancer fragments).

[0392] To find multiple cancer type-specific targets, a similar selection procedure was performed. Multiple CpG sites were ranked based on their information gain, comparing one cancer type to all other samples (non-cancer plus other cancer types). We only attempted to find cancer type-specific targets for a subset of cancer types in the CCGA where we believed there should be sufficient signal to make this feasible. This intuition was supported by the favorable results obtained when using the same selection method to find multiple features for a cancer type classifier.

[0393] Multiple cancer assay panels are generated containing multiple probes targeting multiple genomic regions of choice. Specifically, the multiple assay panels are designed to detect the presence and / or stage of cancer generally (i.e., compared to non-cancer) or specific cancer types as listed below:

[0394] Table 1: Pan-Cancer #1

[0395] Table 2: Blood Cancer #1

[0396] Table 3: Breast Cancer #1

[0397] Table 4: Colorectal Cancer #1

[0398] Table 5: Esophageal Cancer #1

[0399] Table 6: Head and Neck Cancer #1

[0400] Table 7: Hepatobiliary carcinoma #1

[0401] Table 8: Lung Cancer #1

[0402] Table 9: Ovarian Cancer #1

[0403] Table 10: Pancreatic Cancer #1

[0404] Table 12: Pan-Cancer #2

[0405] Table 13: Pan-Cancer #3

[0406] Table 14: Pan-Cancer #4

[0407] Table 15: Pan-Cancer #5

[0408] Table 16: Blood Cancer #2

[0409] Table 17: Breast Cancer #2

[0410] Table 18: Colorectal Cancer #2

[0411] Table 19: Esophageal Cancer #2

[0412] Table 20: Head and Neck Cancer#

[0413] Table 21: Hepatobiliary carcinoma #2

[0414] Table 22: Lung Cancer #2

[0415] Table 23: Ovarian Cancer #2

[0416] Table 24: Pancreatic Cancer #2

[0417] The plurality of assay panels may comprise a plurality of probes configured to hybridize to a modified fragment obtained from processing a plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions included in the list provided in Tables 1 to 24. The plurality of genomic regions provided in Tables 1 to 10 and 12 to 13 are Figure 4 The program shown uses the CCGA data set for identification. Tables 14 to 24 represent a subset of Tables 1 to 13. Tables 1 to 10 list the multiple genomic regions in the following column format, starting from the leftmost column: the chromosome on which the target genomic region is located, the start and end positions of the target genomic region, whether the target genomic region is hypermethylated or hypomethylated, and the annotations of any genes within 10,000 bp of the target region of the genome (if known). Tables 12 to 24, starting from the leftmost column, have the following column format: the chromosome on which the target genomic region is located and the start and end positions of the target genomic region on the chromosome. The chromosome number and the start and end positions are provided relative to the known human reference genome hg19. The sequence of the human reference genome hg19 is available from the Genome Reference Consortium under the reference number GRCh37 / hg19, and is also available from the genome browser provided by the Santa Cruz Genome Institute.

[0418] Generally, a probe can be designed to overlap any CpG site contained within the start / end range of any target region (eg, aberrant fragments) contained in Tables 1-24.

[0419] Table 1

[0420]

[0421]

[0422]

[0423]

[0424]

[0425]

[0426]

[0427]

[0428]

[0429]

[0430]

[0431]

[0432]

[0433]

[0434]

[0435]

[0436]

[0437]

[0438]

[0439]

[0440]

[0441]

[0442]

[0443]

[0444]

[0445]

[0446]

[0447]

[0448]

[0449]

[0450]

[0451]

[0452]

[0453]

[0454]

[0455]

[0456]

[0457]

[0458]

[0459]

[0460]

[0461]

[0462]

[0463]

[0464]

[0465]

[0466]

[0467]

[0468]

[0469]

[0470]

[0471]

[0472]

[0473]

[0474]

[0475]

[0476]

[0477]

[0478]

[0479]

[0480]

[0481]

[0482]

[0483]

[0484]

[0485]

[0486]

[0487]

[0488]

[0489]

[0490]

[0491]

[0492]

[0493]

[0494]

[0495]

[0496]

[0497]

[0498]

[0499]

[0500]

[0501]

[0502]

[0503]

[0504]

[0505]

[0506]

[0507]

[0508]

[0509]

[0510]

[0511]

[0512]

[0513]

[0514]

[0515]

[0516]

[0517]

[0518]

[0519]

[0520]

[0521]

[0522]

[0523]

[0524]

[0525]

[0526]

[0527]

[0528] Table 2 - Blood cancer

[0529]

[0530]

[0531]

[0532] Table 3 - Breast cancer

[0533]

[0534]

[0535]

[0536] Table 4 - Colorectal cancer

[0537]

[0538]

[0539]

[0540] Table 5 - Esophageal cancer

[0541]

[0542]

[0543]

[0544] Table 6 - Head and Neck Cancer

[0545]

[0546]

[0547]

[0548] Table 7 - Hepatobiliary cancer

[0549]

[0550]

[0551]

[0552] Table 8 - Lung cancer

[0553]

[0554]

[0555]

[0556]

[0557] Table 9 - Ovarian Cancer

[0558]

[0559]

[0560]

[0561] Table 10 - Pancreatic cancer

[0562]

[0563]

[0564]

[0565] Table 11

[0566]

[0567]

[0568]

[0569]

[0570]

[0571]

[0572]

[0573]

[0574]

[0575]

[0576]

[0577]

[0578]

[0579]

[0580]

[0581]

[0582]

[0583]

[0584]

[0585]

[0586]

[0587]

[0588]

[0589]

[0590]

[0591]

[0592]

[0593]

[0594]

[0595]

[0596]

[0597]

[0598]

[0599]

[0600]

[0601]

[0602]

[0603]

[0604]

[0605]

[0606]

[0607]

[0608]

[0609]

[0610]

[0611]

[0612]

[0613]

[0614]

[0615]

[0616]

[0617]

[0618]

[0619]

[0620]

[0621]

[0622]

[0623]

[0624]

[0625]

[0626]

[0627]

[0628]

[0629]

[0630]

[0631]

[0632]

[0633]

[0634]

[0635]

[0636]

[0637]

[0638]

[0639]

[0640]

[0641]

[0642]

[0643]

[0644]

[0645]

[0646]

[0647]

[0648]

[0649]

[0650]

[0651]

[0652]

[0653]

[0654]

[0655]

[0656]

[0657]

[0658]

[0659]

[0660]

[0661]

[0662]

[0663]

[0664]

[0665]

[0666]

[0667]

[0668]

[0669]

[0670]

[0671]

[0672]

[0673]

[0674]

[0675]

[0676]

[0677]

[0678]

[0679]

[0680]

[0681]

[0682]

[0683]

[0684]

[0685]

[0686]

[0687]

[0688]

[0689]

[0690]

[0691]

[0692]

[0693]

[0694]

[0695]

[0696]

[0697]

[0698]

[0699]

[0700]

[0701]

[0702]

[0703]

[0704]

[0705]

[0706]

[0707]

[0708]

[0709]

[0710]

[0711]

[0712]

[0713]

[0714]

[0715]

[0716]

[0717]

[0718]

[0719]

[0720]

[0721]

[0722]

[0723]

[0724]

[0725]

[0726]

[0727]

[0728]

[0729]

[0730]

[0731]

[0732]

[0733]

[0734]

[0735]

[0736]

[0737]

[0738]

[0739]

[0740]

[0741]

[0742]

[0743]

[0744]

[0745]

[0746]

[0747]

[0748]

[0749]

[0750]

[0751]

[0752]

[0753]

[0754]

[0755]

[0756] Table 12-Pan-Cancer #2

[0757]

[0758]

[0759]

[0760]

[0761]

[0762]

[0763]

[0764]

[0765]

[0766]

[0767]

[0768]

[0769]

[0770]

[0771]

[0772]

[0773]

[0774]

[0775]

[0776]

[0777]

[0778]

[0779]

[0780]

[0781]

[0782]

[0783]

[0784]

[0785]

[0786]

[0787]

[0788]

[0789]

[0790]

[0791]

[0792]

[0793]

[0794]

[0795]

[0796]

[0797]

[0798]

[0799]

[0800]

[0801]

[0802]

[0803]

[0804]

[0805]

[0806]

[0807]

[0808]

[0809]

[0810]

[0811]

[0812]

[0813]

[0814]

[0815]

[0816]

[0817]

[0818]

[0819]

[0820]

[0821]

[0822]

[0823]

[0824]

[0825]

[0826]

[0827]

[0828]

[0829]

[0830]

[0831]

[0832]

[0833]

[0834]

[0835]

[0836]

[0837]

[0838]

[0839]

[0840]

[0841]

[0842]

[0843]

[0844]

[0845]

[0846]

[0847]

[0848]

[0849]

[0850]

[0851]

[0852]

[0853] Table 13-Pan-Cancer #3

[0854]

[0855]

[0856]

[0857]

[0858]

[0859]

[0860]

[0861]

[0862]

[0863]

[0864]

[0865]

[0866]

[0867]

[0868]

[0869]

[0870]

[0871]

[0872]

[0873]

[0874]

[0875]

[0876]

[0877]

[0878]

[0879]

[0880]

[0881]

[0882]

[0883]

[0884]

[0885]

[0886]

[0887]

[0888]

[0889]

[0890]

[0891]

[0892]

[0893]

[0894]

[0895]

[0896]

[0897]

[0898]

[0899]

[0900]

[0901]

[0902]

[0903]

[0904]

[0905]

[0906]

[0907]

[0908]

[0909]

[0910]

[0911]

[0912]

[0913]

[0914]

[0915]

[0916]

[0917]

[0918]

[0919]

[0920]

[0921]

[0922]

[0923]

[0924]

[0925]

[0926]

[0927]

[0928]

[0929]

[0930]

[0931]

[0932]

[0933] Table 14 - Pan-Cancer #4

[0934]

[0935]

[0936]

[0937]

[0938]

[0939]

[0940]

[0941]

[0942]

[0943]

[0944]

[0945]

[0946]

[0947]

[0948]

[0949]

[0950]

[0951]

[0952]

[0953]

[0954]

[0955]

[0956]

[0957]

[0958]

[0959]

[0960]

[0961]

[0962]

[0963]

[0964]

[0965]

[0966]

[0967]

[0968]

[0969]

[0970]

[0971]

[0972]

[0973]

[0974]

[0975]

[0976]

[0977]

[0978]

[0979]

[0980]

[0981]

[0982]

[0983]

[0984]

[0985]

[0986]

[0987]

[0988]

[0989]

[0990]

[0991]

[0992]

[0993]

[0994]

[0995] Table 15-Pan-Cancer #5

[0996]

[0997]

[0998]

[0999]

[1000]

[1001]

[1002]

[1003]

[1004]

[1005]

[1006]

[1007]

[1008]

[1009]

[1010]

[1011]

[1012]

[1013]

[1014]

[1015]

[1016]

[1017]

[1018]

[1019]

[1020]

[1021]

[1022]

[1023]

[1024]

[1025]

[1026]

[1027]

[1028]

[1029]

[1030]

[1031]

[1032]

[1033]

[1034]

[1035]

[1036]

[1037]

[1038]

[1039]

[1040]

[1041]

[1042]

[1043]

[1044]

[1045]

[1046]

[1047]

[1048]

[1049]

[1050]

[1051]

[1052]

[1053]

[1054]

[1055]

[1056]

[1057]

[1058]

[1059]

[1060]

[1061]

[1062]

[1063]

[1064]

[1065]

[1066]

[1067]

[1068]

[1069]

[1070]

[1071]

[1072]

[1073]

[1074] Table 16 - Blood cancer

[1075] chr Start Finish chr Start Finish chr Start Finish chr1 15251121 15251211 chr1 15480854 15480892 chr1 64240031 64240118 chr1 64240618 64240673 chr1 183774245 183774363 chr1 202183372 202183401 chr1 214724532 214724561 chr1 232765226 232765301 chr1 233750126 233750302 chr2 14772762 14772823 chr2 14774475 14774567 chr2 46526303 46526331 chr2 75427370 75427399 chr2 101436638 101436708 chr2 103236166 103236277 chr2 151342979 151343218 chr2 171571265 171571315 chr2 171571890 171571997 chr2 189157513 189157617 chr2 235860803 235860808 chr2 236402772 236402901 chr2 236403271 236403419 chr2 238395907 238395961 chr3 37901952 37901953 chr3 45187297 45187328 chr3 126373521 126373619 chr3 126373669 126373704 chr3 133748141 133748206 chr3 133748552 133748576 chr3 153838819 153838870 chr3 153839519 153839559 chr3 153839641 153839775 chr3 171527953 171527971 chr4 24914639 24914668 chr4 152246133 152246237 chr4 170947288 170947325 chr4 184019693 184019736 chr4 184020107 184020179 chr5 34656933 34657034 chr5 72416247 72416262 chr5 72733094 72733185 chr5 107005984 107006186 chr5 121413538 121413590 chr6 1312001 1312095 chr6 1312680 1312708 chr6 1314089 1314101 chr6 26987968 26988166 chr6 42928322 42928454 chr7 27275514 27275532 chr7 28995658 28995978 chr7 28996458 28996495 chr7 32997125 32997454 chr7 50860227 50860393 chr7 50860980 50861103 chr7 51384328 51384440 chr7 51384916 51384951 chr7 55086481 55086601 chr7 55086984 55087533 chr7 121945823 121945920 chr7 155602752 155602805 chr8 25041747 25041864 chr8 95651539 95651599 chr8 95651637 95651655 chr8 102505798 102505934 chr8 120220429 120220592 chr9 14312995 14313096 chr9 21559295 21559381 chr9 21559678 21559702 chr9 38620642 38620725 chr9 110251389 110251418 chr9 110252364 110252455 chr9 134421818 134421835 chr10 21462534 21462607 chr10 30026077 30026090 chr10 33624167 33624230 chr10 33624493 33624550 chr10 72973131 72973180 chr10 116164249 116164341 chr11 12132525 12132559 chr11 12399041 12399145 chr11 12399181 12399222 chr11 12695482 12695496 chr11 12695573 12695611 chr11 12696612 12696746 chr11 16628820 16628933 chr11 33037468 33037556 chr11 66790622 66790655 chr11 120039834 120039865 chr11 129245747 129245810 chr11 130318961 130318997 chr11 134201503 134201543 chr11 134201842 134202084 chr12 16500577 16500621 chr12 56882365 56882380 chr12 107486551 107486672 chr12 107487195 107487855 chr12 107712274 107712303 chr13 100634315 100634382 chr14 34420251 34420288 chr14 61747389 61747528 chr14 61747583 61747816 chr14 61748002 61748033 chr15 62456923 62456952 chr15 71055770 71055815 chr15 96874363 96874416 chr15 98504115 98504144 chr15 99193207 99193345 chr15 99193350 99193465 chr16 54964949 54965114 chr16 68771167 68771298 chr16 80966400 80966431 chr16 84402245 84402319 chr16 84853289 84853376 chr17 42061337 42061381 chr17 72427854 72427963 chr17 72428345 72428381 chr17 75207840 75207944 chr17 80693343 80693554 chr18 19750309 19750346 chr18 21269350 21269390 chr18 21269660 21269740 chr18 78005004 78005051 chr19 462182 462235 chr19 33792412 33792524 chr20 1206856 1207034 chr20 6748926 6749036 chr20 18039824 18039897 chr20 22564236 22564265 chr20 50384768 50384896 chr21 38070706 38070765 chr22 31198493 31198637

[1076] Table 17 - Breast Cancer

[1077] chr Start Finish chr Start Finish chr Start Finish chr1 2336398 2336427 chr1 2521025 2521062 chr1 6507964 6508126 chr1 21573736 21574203 chr1 23885071 23885088 chr1 155043332 155043657 chr1 167823371 167823461 chr1 185073819 185073966 chr2 44497709 44497842 chr2 61135116 61135137 chr2 127863602 127863725 chr3 12977068 12977144 chr3 183728814 183728926 chr5 43007937 43007966 chr5 176764101 176764169 chr6 41773521 41773844 chr6 43748464 43748616 chr7 907657 907709 chr7 6188652 6188831 chr7 6188926 6189061 chr7 55410020 55410126 chr7 127371130 127371234 chr7 129800244 129800434 chr7 131041516 131041596 chr7 134918504 134918637 chr8 61777576 61777622 chr8 142367673 142367790 chr8 144668567 144668667 chr8 144668910 144668972 chr9 34224349 34224474 chr9 34372806 34372983 chr9 129401098 129401195 chr9 139888946 139888980 chr10 6003403 6003625 chr10 22047362 22047601 chr11 232864 233062 chr11 63641073 63641104 chr12 110353415 110353451 chr13 28239910 28240164 chr14 102564465 102564502 chr16 3802982 3803074 chr16 85699690 85699921 chr17 26961771 26961833 chr17 42092191 42092220 chr17 70026544 70026667 chr18 74755509 74755577 chr19 14181306 14181682 chr19 33468019 33468055 chr19 38782560 38782589 chr19 40829794 40830032 chr19 45570402 45570450 chr19 45574774 45574782 chr19 45574837 45574888 chr20 6022813 6023045 chr20 32301800 32301953 chr20 60620233 60620412 chr20 60772886 60773878 chr21 37775035 37775141 chr21 46935740 46935936 chr22 21977315 21977347 chr22 23801460 23801567 chr22 24560376 24560522 chr22 39830356 39830457 chr22 41657234 41657350

[1078] Table 18 - Colorectal cancer

[1079] chr Start Finish chr Start Finish chr Start Finish chr1 3659551 3659643 chr1 3659672 3659716 chr1 3663533 3663562 chr1 12123244 12123276 chr1 38511662 38511757 chr2 12858453 12858499 chr2 29338159 29338748 chr2 29338810 29338969 chr2 31360307 31360590 chr2 31360631 31360693 chr2 31360695 31360756 chr2 31360804 31360831 chr2 31456893 31457039 chr2 100937837 100938164 chr2 100938481 100938545 chr2 100938575 100938799 chr2 100938801 100938810 chr2 100938985 100939155 chr2 144694753 144695135 chr2 172367022 172367125 chr2 241542045 241542344 chr3 142791152 142791173 chr3 142839563 142839578 chr3 142839580 142839607 chr3 179168977 179169016 chr4 718082 718112 chr4 79689652 79689732 chr4 156297417 156297556 chr4 156297980 156298073 chr5 38845676 38845705 chr5 82769015 82769061 chr5 111987788 111987818 chr5 146257500 146257602 chr6 73331516 73331851 chr6 73331876 73332169 chr6 73332392 73332674 chr6 73332987 73333099 chr6 127440332 127440510 chr6 127440512 127440524 chr6 151815056 151815089 chr6 152957954 152957995 chr6 163834315 163834383 chr6 163834406 163834533 chr6 163836569 163836900 chr7 2728069 2728108 chr7 28449277 28449291 chr7 44364839 44364903 chr7 69064591 69064772 chr7 69064834 69064858 chr7 76033251 76033289 chr7 90226290 90226363 chr7 106797775 106797804 chr7 107483695 107483918 chr7 134143808 134143908 chr7 140027009 140027043 chr7 149411542 149411728 chr7 149411835 149412304 chr7 150069099 150069346 chr7 150070022 150070058 chr8 53853998 53854027 chr8 80803674 80803831 chr8 97507150 97507246 chr8 143533745 143533774 chr9 37026964 37026993 chr9 93698030 93698051 chr9 140024843 140024919 chr9 140024957 140025023 chr10 3641379 3641396 chr10 7450525 7450567 chr10 7452350 7452550 chr10 7453492 7453521 chr10 49731643 49731749 chr10 64578319 64578355 chr10 101089410 101089439 chr10 125851518 125851645 chr10 125852300 125852498 chr10 125852754 125853191 chr10 133795401 133795430 chr11 2040108 2040148 chr11 3169689 3169835 chr11 94275795 94275813 chr11 94473683 94473769 chr11 94473803 94473984 chr11 94502453 94502489 chr12 104850506 104850537 chr12 104850578 104850592 chr12 104851078 104851186 chr13 26625302 26625502 chr13 28366066 28366122 chr13 36920350 36920379 chr13 36920629 36920769 chr13 73619661 73619698 chr13 95364499 95364528 chr13 95364771 95364800 chr13 95620022 95620057 chr13 110959797 110959860 chr15 45670503 45670839 chr15 48937059 48937095 chr15 48937428 48937646 chr15 48937710 48937987 chr15 79383948 79383977 chr15 83776497 83776596 chr16 10276758 10276799 chr16 10276801 10276841 chr16 71715780 71715809 chr17 32908287 32908371 chr17 46125007 46125061 chr17 47574091 47574149 chr17 80535383 80535469 chr19 3578139 3578223 chr19 10823679 10823708 chr19 50316245 50316330 chr19 57862640 57862783 chr20 4803922 4804008 chr20 33547579 33547585 chr20 36531800 36531910 chr20 37434553 37434722 chr20 37434737 37434744 chr20 39317088 39317196 chr21 27012374 27012431 chr21 45508618 45508647 chr22 39853522 39853590 chr22 39853592 39853592

[1080] Table 19 - Esophageal cancer

[1081] chr Start Finish chr Start Finish chr Start Finish chr1 3663875 3663921 chr1 9712075 9712104 chr1 11538796 11538821 chr1 11539176 11539205 chr1 11539411 11539440 chr1 29450492 29450543 chr1 38512386 38512415 chr1 53068387 53068425 chr1 91869989 91870018 chr1 170633608 170633637 chr1 202679216 202679327 chr1 209381133 209381165 chr1 230561780 230561824 chr1 244014222 244014376 chr2 31456683 31456712 chr2 56410918 56410...

Claims

1. An assay panel for amplifying cfDNA molecules associated with cancer, characterized in that: The assay plate comprises: At least 500 different pairs of polynucleotide probes, wherein each of the at least 500 pairs of probes: (i) comprises two different probes configured to overlap with each other through an overlapping sequence of 30 or more nucleotides, and (ii) is configured to hybridize to a modified fragment obtained from processing of the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions, Each of the one or more genomic regions comprises at least 5 methylation sites and has an abnormal methylation pattern in a plurality of cancerous training samples.

2. The assay plate of claim 1, wherein the overlapping sequences comprise at least 40, 50, 75 or 100 nucleotides.

3. An assay plate as claimed in any preceding claim comprising at least 1000, 2000, 2500, 5000, 6000, 7500, 10000, 15000, 20000 or 25000 pairs of probes.

4. An assay panel for amplifying cfDNA molecules associated with cancer, characterized in that: The assay plate comprises: at least 1000 polynucleotide probes, wherein each of the at least 1000 probes is configured to hybridize to a modified polynucleotide obtained from processing the plurality of cfDNA molecules, wherein each of the plurality of cfDNA molecules corresponds to or is derived from one or more genomic regions, Each of the one or more genomic regions comprises at least 5 methylation sites and has an abnormal methylation pattern in a plurality of cancerous training samples.

5. An assay plate according to any one of the preceding claims, characterised in that: The processing of the plurality of cfDNA molecules comprises: converting unmethylated C (cytosine) in the plurality of cfDNA molecules into U (uracil).

6. An assay plate according to any preceding claim, characterised in that: Each of the plurality of polynucleotide probes on the assay plate is conjugated to an affinity moiety.

7. The assay plate of claim 6, wherein: The affinity moiety is a biotin moiety.

8. An assay plate according to any preceding claim, characterised in that: The plurality of training samples are a plurality of samples derived from a plurality of subjects determined to have cancer.

9. An assay plate according to any preceding claim, characterised in that: When a methylation state vector representing the genomic region in the plurality of cancerous training samples appears at a frequency lower than a threshold in the plurality of reference samples, a genomic region in the plurality of cancerous training samples has an abnormal methylation pattern.

10. The assay plate of claim 9, wherein: The threshold value is 0.1, 0.01, 0.001 or 0.0001.

Images