Single-molecule genome-wide mutation and fragmentation profiling of cell-free DNA

Abstract

Methods for non-invasive cancer detection use a combination of genome-wide mutations and fragmentation characteristics of cfDNA to facilitate cancer screening. TIFF2025503433000052.tif30159

Description

[Technical field]

[0001] STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH This invention was made with United States Government support under Grants CA006973, CA121113, and CA233259 awarded by the National Institutes of Health. The United States Government retains certain rights in the invention.

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of U.S. Provisional Application No. 63 / 290,017, filed December 15, 2021. The entire contents of this application are incorporated herein by reference in their entirety.

[0003] Field Embodiments relate to methods for determining the frequency of somatic mutations in a subject, particularly in the diagnosis and treatment of cancer. [Background technology]

[0004] background Most human cancer deaths are the result of late diagnosis when treatments no longer work 1 Early cancer screening has proven clinical benefit in multiple cancer types, but implementation of screening approaches remains a challenge. 2 For example, currently in the United States, lung cancer screening using low-dose computed tomography (LDCT) is recommended for adults aged 50-80 years who have smoked at least 20 pack-years, are current smokers, or have quit smoking within the last 15 years. 3 Screening with LDCT has been shown to reduce mortality, 4,5 Adherence to this test among high-risk individuals is low (<6%), due in part to concerns about its low specificity, radiation exposure, and potential harm from unnecessary diagnostic procedures. 6 Liquid biopsy overcomes some of these challenges and may represent an attractive approach for the noninvasive detection of lung cancer and other neoplasms. Summary of the Invention

[0005] overview Provided herein is a non-invasive and ultra-sensitive analysis of single cell-free DNA (cfDNA) molecule to detect the frequency of somatic mutations throughout the genome.It is found that cancer patients have altered chromatin organization-related mutation profiles compared with healthy individuals.

[0006] Thus, in certain aspects, a method of determining the frequency of somatic mutations in a subject includes the steps of extracting cell-free DNA (cfDNA) from a biological sample of the subject; generating a genomic library from the extracted cfDNA; sequencing individual cfDNA molecules to obtain a mutation profile; determining multiregional differences in the mutation profile; and determining the frequency of somatic mutations in the subject.

[0007] In certain embodiments, determining the genome-wide mutation and fragmentation profile comprises identifying mutations and fragment length changes in the sequence of individual cfDNA molecules.

[0008] In certain embodiments, the mutation profile comprises mutation frequencies and mutation types across the subject's genome.

[0009] In certain embodiments, the subject's genome-wide mutation profile is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 20 million bases.

[0010] In certain embodiments, the subject's genome-wide mutation profile is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 10 million bases.

[0011] In certain embodiments, the subject's genome-wide mutation profile is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 5 million bases.

[0012] In certain embodiments, mutations in each sequenced molecule are determined after removal of common germline variants and non-scalable regions.

[0013] In certain embodiments, the frequency and mutation type of single molecule somatic mutations across the genome of a subject compared to the frequency and mutation type of single molecule somatic mutations across the genome of a normal subject is diagnostic of cancer.

[0014] In certain aspects, a method of treating cancer in a subject includes extracting cell-free DNA (cfDNA) from a biological sample of the subject; generating a genomic library from the extracted cfDNA; sequencing individual cfDNA molecules to obtain a mutation profile; determining multiregional differences in the mutation profile and determining the frequency of somatic mutations in the subject; and administering a cancer treatment to the subject based thereon.

[0015] In certain embodiments, the cancer treatment comprises surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy, targeted therapy, and combinations thereof.

[0016] In certain aspects, the cancer comprises colorectal cancer, lung cancer, breast cancer, gastric cancer, pancreatic cancer, bile duct cancer, brain cancer, or ovarian cancer.

[0017] In certain embodiments, the lung cancer is small cell lung cancer (SCLC).

[0018] In certain embodiments, the lung cancer is non-small cell lung cancer (NSCLC).

[0019] In certain aspects, subjects with cancer comprise an altered chromatin organization-associated mutation profile compared to healthy individuals.

[0020] In certain embodiments, the genome-wide mutation and fragmentation profile comprises identifying mutations and fragment length changes in the sequence of individual cfDNA molecules.

[0021] In certain aspects, the mutation profile comprises mutation frequency and mutation type across the subject's genome.

[0022] In certain embodiments, the subject's genome-wide mutation profile is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 20 million bases.

[0023] In certain embodiments, the subject's genome-wide mutation profile is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 10 million bases.

[0024] In certain embodiments, the subject's genome-wide mutation profile is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 5 million bases.

[0025] In certain embodiments, after removal of common germline variants and non-evaluable regions, genome-wide variation of each sequenced molecule is determined.

[0026] In certain embodiments, the method of determining the regional frequency of mutations across the genome includes sequencing individual cfDNA molecules isolated from a subject, estimating the mutation frequency and mutation type across the genome, and determining the mutation type and frequency in genomic regions that are altered in cancer relative to the mutation profile and regions that are altered in normal cfDNA to determine multi-regional differences in the mutation profile, thereby determining the regional frequency of mutations across the genome. In certain embodiments, the estimation of the mutation frequency and mutation type across the genome includes using non-overlapping bins that range in size from thousands to millions of bases. In certain embodiments, the tumor-specific changes are quantified by one or more assays. In certain embodiments, the one or more assays include in silico dilution and / or downsampling assays. In certain embodiments, after removal of common germline variants and / or non-evaluable regions, each sequenced molecule is scanned for single nucleotide changes. In certain embodiments, the genomic region is characterized by slow replication timing, low gene expression, B compartmentalization, high H3K9me3 abundance, low GC content, or combinations thereof.In certain embodiments, the estimated mutation frequency is defined as the number of variants per million positions evaluated across all sequenced DNA molecules.In certain embodiments, the method further comprises combining the mutation profile with the genome-wide fragmentation profile.In certain embodiments, the method further comprises executing a machine learning model to determine the change in genome-wide mutation profile, which classifies or excludes the subject as having or at risk of having cancer based on the genome-wide mutation profile identified for the subject.

[0027] In certain embodiments, the method of determining whether a subject responds to treatment comprises any one or more of the methods embodied herein.In certain embodiments, the treatment is selected from surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormone therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy, targeted therapy, and combinations thereof.

[0028] definition Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by those skilled in the art to which the present invention belongs. Furthermore, terms, for example, terms defined in commonly used dictionaries, should be interpreted to have a meaning consistent with the meaning in the context of the relevant art, and unless expressly defined herein, they should not be interpreted in an idealized or overly formal sense.

[0029] As used herein, the singular forms "a," "an," and "the" are intended to include the plural forms unless the context clearly dictates otherwise. Furthermore, to the extent the terms "including," "includes," "having," "has," "with," or variations thereof are used in the detailed description and / or claims, such terms are intended to be inclusive, similar to the term "comprising."

[0030] The term "about" or "approximately" means within an acceptable error range for a particular value as determined by one of ordinary skill in the art. This depends in part on how the value is measured or determined, i.e., on the limitations of the measurement system. For example, "about" may mean within 1 or more standard deviations as is customary in the art. Alternatively, "about" may mean within 20%, 10%, 5%, or 1% of a given value or range. Alternatively, particularly with respect to biological systems or processes, the term may mean within an order of magnitude of within 5-fold of a value, and may also mean within 2-fold of a value. When a particular value is described in this application and claims, the term "about" should be assumed to mean within an acceptable error range for the particular value, unless otherwise indicated.

[0031] The terms "aligned", "alignment", "mapped", or "aligning" or "mapping" refer to one or more sequences that are found to match, in terms of the order of nucleic acid molecules, with a known sequence from a reference genome. Such alignments may be performed manually or by computer algorithms. Examples include the Efficient Local Alignment of Nucleotide Data (ELAND) computer program distributed as part of the Illumina Genomics Analysts pipeline. The matching of sequence reads in an alignment may be a 100% sequence match or may be less than 100% (non-perfect match).

[0032] The term "cancer" as used herein means a disease, condition, trait, genotype, or phenotype characterized by unregulated cell growth or replication, as known in the art, and includes, but is not limited to, lung cancer (including non-small cell lung cancer), gastric cancer, colorectal cancer, and, for example, leukemias, such as acute myeloid leukemia (AML), chronic myeloid leukemia (CML), acute lymphocytic leukemia (ALL), and chronic lymphocytic leukemia, AIDS-related cancers, such as Kaposi's sarcoma; breast cancer; bone cancers, such as osteosarcoma, chondrosarcoma, Ewing's sarcoma, fibrosarcoma, giant cell tumor, adamantinoma, and chordoma; brain cancers, such as meningioma, glioblastoma, low-grade astrocytoma, oligodendroma, and glioma. Oligodendrocytoma, pituitary tumor, Schwannoma, and metastatic brain cancer; various lymphomas, e.g., mantle cell lymphoma, non-Hodgkin's lymphoma, adenoma, squamous cell carcinoma, head and neck cancer including laryngeal cancer, gallbladder and bile duct cancer, retinal cancer, e.g., retinoblastoma, esophageal cancer, gastric cancer, multiple myeloma, ovarian cancer, uterine cancer, thyroid cancer, testicular cancer, endometrial cancer, melanoma, bladder cancer, prostate cancer, pancreatic cancer, sarcoma, Wilms' tumor, cervical cancer, head and neck cancer, skin cancer, nasopharyngeal cancer, liposarcoma, epithelial carcinoma, renal cell carcinoma, gallbladder adenocarcinoma, parotid adenocarcinoma, endometrial sarcoma, multidrug resistant cancer; and proliferative diseases and conditions, e.g., angiogenesis associated with tumor angiogenesis.

[0033] The terms "cell-free nucleic acid," "cell-free DNA," or "cfDNA" refer to nucleic acid fragments circulating in an individual's body (e.g., in the bloodstream) and derived from one or more healthy cells and / or derived from one or more cancer cells. Additionally, cfDNA may originate from other sources, such as viruses, fetuses, etc.

[0034] The term "circulating tumor DNA" or "ctDNA" refers to nucleic acid fragments derived from tumor cells or other types of cancer cells that may be released into an individual's bloodstream as a result of biological processes such as apoptosis or necrosis of dying cells, or that may be actively released by viable tumor cells.

[0035] As used herein, the terms "comprising," "comprise," or "comprised," and conjugations thereof, are intended to be inclusive or open-ended, allowing for additional elements with respect to a defined or described element of an item, composition, apparatus, method, process, system, etc., thereby indicating that the defined or described item, composition, apparatus, method, process, system, etc. includes the specified element, or equivalents thereof, as appropriate, and includes other elements, but is still within the scope / definition of the defined item, composition, apparatus, method, process, system, etc.

[0036] "Diagnostic" or "diagnosed" means to determine the presence or nature of a pathological condition. Diagnostic methods differ in their sensitivity and specificity. The "sensitivity" of a diagnostic assay is the percentage of diseased individuals who test positive (percent "true positives"). Diseased individuals not detected by the assay are "false negatives." Subjects who are not diseased and test negative in the assay are referred to as "true negatives." The "specificity" of a diagnostic assay is 1 - the false positive rate, where the "false positive" rate is defined as the proportion of individuals without the disease who test positive. While a particular diagnostic method may not definitively diagnose a condition, it suffices if the method provides a positive indication that aids in diagnosis.

[0037] As used herein, "effective amount" means an amount that provides a therapeutic or prophylactic benefit.

[0038] As used herein, the terms "fragmentation profile", "position-dependent difference in fragmentation pattern", and "position-dependent difference in fragment size and coverage across the genome" are synonymous and can be used interchangeably. In some embodiments, determining a cfDNA fragmentation profile in a mammal can be used to identify a mammal as having cancer. For example, cfDNA fragments obtained from a mammal (e.g., from a sample obtained from the mammal) can be subjected to low-coverage whole genome sequencing, and the sequenced fragments can be mapped to the genome (e.g., in non-overlapping windows) and evaluated to determine a cfDNA fragmentation profile. As described herein, the cfDNA fragmentation profile of a mammal with cancer is more heterogeneous (e.g., in terms of fragment length) than the cfDNA fragmentation profile of a healthy mammal (e.g., a mammal without cancer). Thus, the present disclosure also provides methods and materials for evaluating, monitoring, and / or treating a mammal (e.g., a human) with or suspected of having cancer. In some embodiments, this document provides methods and materials for identifying a mammal as having cancer. For example, a sample (e.g., a blood sample) obtained from a mammal can be evaluated to determine the presence of cancer in the mammal, and optionally the tissue of origin of the cancer, based at least in part on the cfDNA fragmentation profile of the mammal. In some embodiments, methods and materials are provided for monitoring that a mammal has cancer. For example, a sample (e.g., a blood sample) obtained from a mammal can be evaluated to determine the presence of cancer in the mammal, based at least in part on the cfDNA fragmentation profile of the mammal. In some embodiments, methods and materials are provided for determining that a mammal has cancer and for administering one or more cancer treatments to the mammal to treat the mammal. For example, a sample (e.g., a blood sample) obtained from a mammal can be evaluated to determine whether the mammal has cancer, based at least in part on the cfDNA fragmentation profile of the mammal, and one or more cancer treatments can be administered to the mammal.

[0039] As used herein, the "frequency" of a mutation is defined as the number of variants per million positions evaluated across all sequenced DNA molecules.

[0040] The term "genomic nucleic acid" or "genomic DNA" refers to nucleic acid, including chromosomal DNA, derived from one or more healthy (e.g., non-tumor) cells. In various embodiments, genomic DNA can be extracted from cells derived from blood cell lineages, such as white blood cells (WBCs).

[0041] The term "mutation profile" as used herein refers to the type and frequency of mutation observed in bins across the genome.Mutation profile comparison between the genomic regions that are more frequently altered in cancer and the mutation profiles derived from the regions that are more frequently mutated in normal cfDNA can be used to determine multi-regional differences.

[0042] "Optional" or "optionally" means that the subsequently described event or circumstance may or may not occur, and that the description includes instances where the event or circumstance occurs and instances where it does not occur.

[0043] As used in this specification and the appended claims, the term "or" is generally used in its sense including "and / or" unless the content clearly dictates otherwise.

[0044] "Parenteral" administration of the immunogenic compositions includes, for example, subcutaneous (sc), intravenous (iv), intramuscular (im), or intrasternal injection or infusion techniques.

[0045] The terms "patient" or "individual" or "subject" are used interchangeably herein and refer to a mammalian subject to be treated, with human patients being preferred. In some embodiments, the methods of the present invention are used in the development of animal models of disease in laboratory animals, in veterinary applications, including but not limited to rodents, including mice, rats, and hamsters, and primates.

[0046] The term "reference genome" as used herein may refer to a digital or previously identified nucleic acid sequence database assembled as representative of a species or subject. A reference genome may be assembled from nucleic acid sequences from multiple subjects, samples, or organisms and does not necessarily represent the nucleic acid makeup of one person. A reference genome may be used to map sequencing reads from a sample to chromosomal locations. For example, reference genomes for human subjects, as well as many other organisms, can be found at the National Center for Biotechnology Information at ncbi.nlm.nih.gov.

[0047] The term "read segment" or "read" refers to any nucleotide sequence comprising a sequence read obtained from an individual and / or a nucleotide sequence derived from an initial sequence read from a sample obtained from an individual.

[0048] The terms "sample", "patient sample", "biological sample", and the like encompass a variety of sample types obtained from a patient, individual, or subject and can be used in diagnostic, prognostic, and / or monitoring assays. A patient sample may be obtained from a healthy subject, a diseased patient, or a lung cancer patient. In certain embodiments, a "provided" sample may be obtained by the person (or machine) performing the assay, or may be obtained by another person (or machine) and transferred to the person (or machine) performing the assay. Additionally, a sample obtained from a patient may be divided and only a portion may be used for diagnosis. Additionally, the sample, or a portion thereof, may be stored under conditions that preserve the sample for later analysis. This definition specifically encompasses blood as well as other liquid samples of biological origin (including, but not limited to, peripheral blood, serum, plasma, umbilical cord blood, amniotic fluid, cerebrospinal fluid, urine, saliva, feces, and synovial fluid), solid tissue samples, such as biopsy specimens or tissue cultures or cells derived therefrom and their progeny. In certain embodiments, the sample comprises cerebrospinal fluid. In certain embodiments, the sample comprises a blood sample. In other embodiments, the sample comprises a plasma sample. In yet another embodiment, a serum sample is used. The definition of "sample" also includes samples that are obtained and then manipulated in any manner, for example, by centrifugation, filtration, precipitation, dialysis, chromatography, treatment with a reagent, washed, or enriched for a certain cell population. These terms further encompass clinical samples, including cells in culture, cell supernatants, tissue samples, organs, etc. Samples may also include fresh frozen and / or formalin-fixed, paraffin-embedded tissue blocks, such as blocks prepared from clinical or pathological biopsies, blocks prepared for pathological analysis or immunohistochemical studies.

[0049] The term "sequence read" refers to a nucleotide sequence read from a sample obtained from an individual. Sequence reads can be obtained by various methods known in the art.

[0050] As defined herein, a "therapeutically effective" amount (i.e., an effective dosage) of a compound or agent means an amount sufficient to produce a therapeutically (e.g., clinically) desired result. The composition can be administered from one or more times per day, including once every other day, to one or more times per week. Those skilled in the art will appreciate that certain factors, including, but not limited to, the severity of the disease or disorder, previous treatments, the overall health and / or age of the subject, and other diseases present, may affect the dosage and timing required to effectively treat a subject. Furthermore, treatment of a subject with a therapeutically effective amount of a compound of the present invention may include a single treatment or a series of treatments.

[0051] As used herein, the terms "treat," "treating," "treatment," and the like refer to reducing or ameliorating a disorder and / or symptoms associated with the disorder. Although not excluded, it is understood that treating a disorder or condition does not require that the disorder, condition, or symptoms associated therewith be completely ablated.

[0052] Gene: All genes, gene names, and gene products disclosed herein are intended to correspond to homologs from any species to which the compositions and methods disclosed herein are applicable. It is understood that when a gene or gene product from a particular species is disclosed, this disclosure is intended to be illustrative only and should not be construed as limiting unless clearly indicated by the relevant context. Thus, for example, a gene or gene product disclosed herein is intended to encompass homologous and / or orthologous genes and gene products from other species.

[0053] Ranges: Throughout this application, various aspects of the invention may be presented in a range format. It should be understood that the description in range format is merely for convenience and brevity and should not be construed as an inflexible limitation on the scope of the invention. Thus, the description of a range should be considered to have specifically disclosed all possible subranges as well as individual numerical values ​​within that range. For example, the description of a range such as 1-6 should be considered to have specifically disclosed subranges such as 1-3, 1-4, 1-5, 2-4, 2-6, 3-6, etc., as well as individual numerical values ​​within that range, for example, 1, 2, 2.7, 3, 4, 5, 5.3, and 6. This applies regardless of the breadth of the range.

[0054] Any composition or method provided herein can be combined with one or more of any of the other compositions and methods provided herein. [Brief description of the drawings]

[0055] [Figure 1] Figure 1 is a schematic of the overall approach to detect cancer using single molecule cfDNA sequencing. Blood is collected from a population of individuals. A portion of the population has cancer. cfDNA is extracted from the plasma and subjects and subjected to single molecule sequencing using a massively parallel sequencing approach. Sequence alterations are used to obtain genome-wide mutation profiles and machine learning is used to identify regional differences in cancer and non-cancer mutation frequencies to distinguish individuals with and without cancer. [Figure 2A]Figures 2A-2J are a series of graphs and plots showing single molecule mutation analysis of lung cancer and normal samples from the PCAWG consortium. Figure 2A: Number of mutations detected in PCAWG lung cancer samples from smokers when downsampled across a range of sequencing coverage amounts and tumor fractions. Figure 2B: Percentage of PCAWG lung cancer mutations observed in single DNA molecules at different sequence coverages and tumor fractions as indicated. Figure 2C: Frequency of single molecule somatic and background C>A alterations in lung cancer samples and blood-derived matched normal samples without quality or germline filters. Figure 2D: Frequency of single molecule somatic and background C>A alterations in lung cancer samples and blood-derived matched normal samples with quality or germline filters including filtering for 8-oxo-dG-related sequence alterations. Figure 2E: Frequency of single molecule somatic and background C>A alterations across a 50 Mb region of chromosome 1 in patient DO25320. C>A frequencies were calculated in sliding 2.5 Mb windows with a step size of 100 kb. The red and black dashed lines represent the mutation frequencies of the top decile bins most enriched in C>A alterations in lung cancer samples and blood-derived matched normal samples. Figure 2F: Background C>A frequencies of the top decile bins most enriched in C>A alterations in lung cancer samples and blood-derived matched normal samples obtained after removing known PCAWG somatic mutations. For example, as can be seen from the identity solid lines, the background C>A frequencies between these regions are similar for each sample. Figure 2G: Number of molecules with background C>A alterations in lung cancer samples and blood-derived normal samples, respectively. Most background alterations were observed in only one molecule even at >30x coverage. FIG. 2H: Difference in region C>A frequency in normal or tumor samples after subtracting the C>A frequency in the top decile bin enriched in normal samples from the top decile bin enriched in mutations in tumor samples using the GEMINI approach.Regional C>A frequency differences preferentially remove background alterations, thereby enriching the frequency of observed somatic mutations. Figure 2I: Association between regional differences in single molecule C>A frequency and the frequency of high-confidence somatic C>A alterations reported in these samples by the PCAWG consortium. Figure 2J: Receiver operating characteristic (ROC) curves for discriminating between lung cancer and normal samples using the GEMINI approach and a test set downsampled to 1x coverage compared to using the total single molecule C>A frequency after quality and germline filtering. An AUC of 0.47 was obtained from the ROC of the GEMINI approach without 8-oxo-dG-associated alteration filtering. This highlights the importance of removing these artifacts to identify tumor-specific alterations. [Figure 2B] See legend to Figure 2A. [Figure 2C-1] See legend to Figure 2A. [Figure 2C-2] See legend to Figure 2A. [Figure 2D-1] See legend to Figure 2A. [Figure 2D-2] See legend to Figure 2A. [Figure 2E] See legend to Figure 2A. [Figure 2F] See legend to Figure 2A. [Figure 2G] See legend to Figure 2A. [Figure 2H-1] See legend to Figure 2A. [Figure 2H-2] See legend to Figure 2A. [Figure 2I] See legend to Figure 2A. [Figure 2J] See legend to Figure 2A. [Figure 3A]Figures 3A-3B are a series of plots and graphs demonstrating that genome-wide mutation profiles of tissue and plasma samples were associated with replication timing. Figure 3A: Somatic mutation frequency PCAWG lung cancer derived from smoking individuals (n=65) was calculated in a genome-wide sliding 2.5 Mb window with a step size of 100 kb and expressed as an average across individuals. Figure 3B: Association of mutation frequency across tissue-specific replication timing strata in tissues and cfDNA derived from patients with NSCLC, melanoma, BNHL, or no cancer. Replication timing was obtained as wavelet-smoothed transform of six fraction profiles representing different time points during replication in 1 kb bins derived from IMR90 cell line for NSCLC analysis, NHEK cell line for melanoma analysis, and GM12878 cell line for BNHL analysis47,48. The weighted average of the replication timing values ​​was calculated in 2.5Mb bins, and then the bins were grouped into 5 equal bin sets, including the bins with the earliest to the latest replication timing.In each bin set, we calculated the mutation frequency in tissues at different replication layers using the number of somatic mutations reported by the PCAWG consortium per 1Mb of genome, and compared it with the single molecule mutation frequency in plasma using Pearson correlation.To take into account the potential systematic variation in the measured genome-wide mutation frequency, we subtracted the single molecule mutation frequency in each bin set in a separate panel of 20 non-cancer cfDNA samples from cancer and non-cancer cfDNA samples.Then, we scaled the mutation frequency within each sample and mutation type to have a minimum value of 0. [Figure 3B-1] See legend to Figure 3A. [Figure 3B-2] See legend to Figure 3A. [Figure 3B-3] See legend to Figure 3A. [Figure 4-1]Figures 4A-4I are a series of plots and ROC curves demonstrating lung cancer detection using GEMINI and the combined GEMINI / DELFI approach. Figure 4A: GEMINI scores in high-risk individuals aged 50-80 years with a smoking history of ≥20 packs per year with or without lung cancer. The number of individuals at each stage or histology is shown. Importantly, the GEMINI scores of non-cancer individuals with and without benign nodules were similar, while those with cancer had higher GEMINI scores. Figure 4B: GEMINI scores of high-risk individuals without lung cancer, as well as individuals who did not have lung cancer when confirmed by imaging at baseline, but later developed lung cancer. Figure 4C: GEMINI scores in current or ex-smokers aged 50-80 years with and without cancer in the validation cohort. The validation cohort was enriched for early stage disease (stage I=25, stage II=2, stage III=2, stage IV=2, 1 individual with unknown stage). Figure 4D: ROC curves for lung cancer detection in high-risk individuals in the LUCAS cohort (n=89 with lung cancer, n=74 without cancer) show high performance using GEMINI or GEMINI and DELFI approaches. Figure 4E: ROC curves for lung cancer detection in a subset of high-risk individuals in the LUCAS cohort with at least 40 pack-years (n=63 with lung cancer, n=46 without cancer) show high GEMINI performance with long-term smoking history. Figure 4F: ROC curves for detection among high-risk individuals diagnosed with stage I lung cancer (n=13 with lung cancer, n=74 without cancer) from the LUCAS cohort. Figure 4G: ROC curves for stage I lung cancer detection among individuals in the validation cohort (n=25 with lung cancer, n=14 without cancer). Figure 4H: ROC curves for detection of high-risk individuals from the LUCAS cohort and ≧40 pack-years of smoking history diagnosed with stage I lung cancer (n=9 with lung cancer, n=46 without cancer). Figure 41: ROC curve for detection of stage I lung cancer and ≧40 pack-year smoking history among individuals in the validation cohort (n=13 with lung cancer, n=5 without cancer). [Figure 4-2] See description of Figure 4-1. [Figure 4-3] See description of Figure 4-1. [Figure 5-1] Figures 5A-5F are a series of graphs and ROC curves proving the GEMINI approach for non-invasive detection across multiple cancer types. Figure 5A: The GEMINI score performs well for cancer detection in SCLC patients and cancer-free high-risk individuals in the LUCAS and validation cohorts (Supplementary Table 4). Figure 5B: Regional differences in single molecule C>A frequency in the LUCAS and validation cohorts prove that the GEMINI approach can be used to identify the most altered bins between SCLC and NSCLC. Figure 5C: ROC curves (orange) for detecting SCLC (n=13) compared to non-cancer controls (n=88), as well as ROC curves (purple) for distinguishing SCLC (n=13) from NSCLC (n=99) in the combined LUCAS and validation cohorts. Figure 5D: Cross-validated regional differences in single molecule mutation frequencies in cfDNA in the liver cancer cohort, median-centered within each mutation type, show high levels of T>C mutations in HCC patients. P values ​​were generated using Wilcoxon rank sum test and corrected for multiple comparisons using the Benjamini-Hochberg method. Horizontal dashed lines indicate p-values ​​of 0.05. Figure 5E: GEMINI scores in the liver cancer cohort and the number of individuals represented at each stage demonstrate high sensitivity of liver cancer detection across all stages. Figure 5F: Principal coordinate analysis of Euclidean distance matrices reflecting cross-validated pairwise differences in regional mutation frequencies between NSCLC, SCLC, and HCC. The first two principal coordinates are shown by contour lines showing kernel density estimates of 0.7 and 0.95 probability for each cancer type. The composition of cancer types in clusters derived from K-means clustering with k=3 is shown on the right. [Figure 5-2] See description of Figure 5-1. [Figure 6]Figure 6 is a schematic diagram outlining the analysis cohort. Each box represents an analysis cohort and indicates whether the GEMINI approach was evaluated by cross-validation or validated using a fixed model. The dashed lines indicate the analysis of cohort subsets for the evaluation of individual tumor types or comparison of cancer subtypes. [Figure 7-1] Figure 7 is a series of plots showing genomic mutation profiles in common cancers. Mean somatic mutation frequencies calculated in sliding 2.5 Mb windows across available chromosome 1 with a step size of 100 kb were obtained from the analysis of 2,511 PCAWG samples across 25 common cancer types. [Figure 7-2] See description of Figure 7-1. [Figure 7-3] See description of Figure 7-1. [Figure 8] Figure 8 is a schematic of a dilution and downsampling experiment. In this example, we consider a tumor sample containing N somatic mutations at genomic positions 1, 2, ..., N. We start with 30 non-tumor-derived observations and 10 tumor-derived mutations (25% tumor purity). During the dilution step, we spike in non-tumor observations until we reach the desired tumor fraction. After dilution, we randomly draw fragments from the set of all fragments to achieve the desired average coverage across the genomic positions. We count the resulting number of observed mutations and finally calculate the proportion of observed mutations that were only observed in a single fragment. In this example, there are three observed mutations, one of which was only observed in one molecule. [Figure 9A]Figures 9A, 9B are plots and graphs demonstrating the identification of background changes in single molecule sequencing associated with 8-oxo-dG lesions. Figure 9A: For both background changes and known germline variants, the ratio of the frequency of each type of single base change when the purine guanine or adenine (pu) is on read 1 (R1) or the pyrimidine cytosine or thymine (py) is on read 2 (R2) before mutation to when the pyrimidine is on read 1 and the purine is on read 2 in 62 tissue samples (31 lung cancer samples and 31 blood-derived matched normal samples) from the PCAWG. Background changes reflect sequence changes identified by single molecule analysis that were not reported as somatic variants by the PCAWG. Here, germline variants reported by the PCAWG were also removed from the background variants to enrich for likely artifactual changes. Figure 9B: Ratio of known somatic mutations:background mutations identified by single molecule analysis before removing sequence changes likely related to 8-oxo-dG (R1pu or py, R2pu or py) and after filtering these changes. Here, only bases with a cytosine on R1 and a guanine on read 2 are considered (R1py, R2pu). [Figure 9B] See legend to Figure 9A. [Figure 10] Figures 10A, 10B are plots demonstrating the analysis of single molecule sequence alterations in PCAWG lung cancer and normal samples. Figure 10A: Single molecule mutation frequency in PCAWG lung cancer (n=31) and blood-derived matched normal samples (n=31). p-values ​​were corrected for multiple comparisons using the Benjamini-Hochberg method. The horizontal dashed line indicates a p-value of 0.05. Figure 10B: Cross-validated regional differences in single molecule mutation frequency in PCAWG lung cancer (n=31) and blood-derived matched normal samples (n=31), centered around the median within each mutation type. p-values ​​were generated using the Wilcoxon rank sum test and corrected for multiple comparisons using the Benjamini-Hochberg method. The horizontal dashed line indicates a p-value of 0.05. [Figure 11] FIG. 11 is a graph demonstrating the analysis of somatic and background alterations across mutation types in PCAWG lung cancer. For each mutation type analyzed, the ratio of somatic:background alterations identified by single molecule analysis after removal of potential 8-oxo-dG-related artifacts. Somatic alterations reflect sequence alterations identified by single molecule analysis that were also reported as somatic by PCAWG, whereas background alterations were identified by single molecule analysis but not reported as somatic by PCAWG. Overall, C:G>A:T alterations represented the largest proportion of somatic alterations. [Figure 12-1] Figure 12 is a series of plots demonstrating the analysis of single molecule sequence alterations across sequencing lanes in PCAWG lung cancer and normal samples. Single molecule mutation frequency in PCAWG lung cancer samples and blood-derived normal samples across sequencing lanes. For each sample, sequencing reads were separated into separate binary alignment map (BAM) files based on these related groups of reads, representing sequencing reads from one lane of the NGS experiment. The resulting BAM files contained a median of 464 million reads (range: 6-738 million). Approximately 1 million reads were randomized five times with replacement from each sequencing lane (up to 6 lanes per sample were shown). Single molecule mutation frequencies varied widely within individual samples depending on the lane and type of sequence alteration analyzed. [Figure 12-2] See description of Figure 12-1. [Figure 12-3] See description of Figure 12-1. [Figure 13] Figure 13 is a series of plots demonstrating genome-wide somatic single-molecule C>A mutation profiles in lung cancer. Single-molecule C>A somatic mutation frequencies calculated in sliding 2.5 Mb windows across autosomes with a step size of 100 kb from an aggregated analysis of 31 PCAWG lung cancer samples showed extensive mutation frequency differences depending on genomic location. [Figure 14] Figure 14 is a series of plots demonstrating somatic single C>A mutation profiles across chromosome 4 in PCAWG lung cancers. Single C>A somatic mutation frequencies calculated in sliding 2.5 Mb windows across chromosome 4 from PCAWG lung cancer samples with a step size of 100 kb revealed that mutation profiles were similar across different lung cancers. [Figure 15] Figure 15 is a schematic diagram of GEMINI region mutation frequency analysis. The genome is divided into 1,144 non-overlapping 2.5 Mb bins (20 bins are shown here), and the single molecule mutation frequency in each bin is calculated as the number of sequence changes per million evaluable bases, defined as the number of positions in the fragment where each sequence change could be detected after quality filtering and germline filtering. The training set samples are used to identify the most differentially mutated bins between cancer and non-cancer samples. In the training set, sequence data from all cancer and non-cancer samples are combined, and the single molecule mutation frequency of cancer and non-cancer is calculated in each bin. The difference in single molecule mutation frequency between cancer and non-cancer samples in each bin is then calculated, and the 10% of the most mutated bins in cancer samples relative to non-cancer samples and the 10% of the most mutated bins in non-cancer samples relative to cancer samples are identified (shown as triangles and circles, respectively). In the test set, the difference in single molecule mutation frequency is calculated between these two sets of bins in new samples not included in the training set to create regional differences in mutation frequency that can be used to classify samples into samples from healthy individuals or samples from individuals with cancer. By choosing the difference in single molecule mutation frequency between two sets of regions in the genome within an individual sample, this approach takes into account the total number of sequence changes in that sample that may be due to technical variations in the sequencing run. [Figure 16]Figure 16 is a graph demonstrating the effect of matched WBC filtering in PCAWG lung cancer on enrichment of somatic alterations by single molecule sequencing. Single molecule C>A frequency in PCAWG lung cancer (n=31) after removing sequence alterations identified in matched blood-derived normal samples at >30x coverage. Analysis revealed that subtraction of mutations observed in matched normal samples was not effective in removing background alterations, as such alterations were typically observed once and were absent in both tumor and matched non-cancer samples. [Figure 17A]Figures 17A-17C are a series of plots demonstrating the association of single molecule genome-wide mutation profiles of tissue and plasma samples with genomic features. These figures show genome-wide mutation frequencies in tissue and cfDNA from patients with NSCLC, melanoma, BNHL, or cancer-free patients across strata of tissue-specific gene expression, A / B compartmentalization, and H3K9me3 abundance, respectively. Weighted averages of each feature value were calculated in 2.5 Mb bins, and the bins were then grouped into five equal bin sets ordered by the feature values. In each bin set, we calculated the mutation frequency in tissues in different strata using the number of somatic mutations reported by the PCAWG consortium per Mb of genome, which was compared to the single molecule mutation frequency in plasma using Pearson correlation. To account for differences in the overall frequency of each mutation type in each bin in cfDNA, the single molecule mutation frequency in each bin set in the non-cancer sample panel (n=20) was subtracted from the single molecule mutation frequency in each bin set in the cancer and non-cancer cfDNA samples, and the resulting values ​​were scaled to have a minimum value of 0 for each mutation type and sample type. Figure 17A: Gene expression was calculated as the sum of transcripts per million (TPM) overlapping each 2.5 Mb bin, weighted by transcript length averaged across TCGA NSCLC, melanoma, and BNHL samples. Figure 17B: A / B compartmentalization represents primarily open and closed regions of the genome, respectively, measured as the first eigenvector of the correlation matrix of average methylation beta values ​​in 100 kb bins across TCGA NSCLC samples for NSCLC analysis and averaged across 12 TCGA cancer types for melanoma analysis. The first eigenvector for the genomic contact matrix from Hi-C analysis of lymphoblastoid cells (GM12878 cell line) was used for BNHL analysis 33 .Figure 17C: H3K9me3 abundance, a known marker of heterochromatin, was obtained as fold change in coverage in enriched samples compared to control samples48 from ChIP-seq of A549 cells (3 pooled replicates), GM23248, and Karpas 422 cells (2 pooled replicates) for NSCLC, melanoma, and BNHL analysis, respectively. [Figure 17B] See legend to Figure 17A. [Figure 17C] See legend to Figure 17A. [Figure 18] Figure 18 is a plot demonstrating regional differences in single molecule mutation frequencies in the high-risk LUCAS cohort. Cross-validated regional differences in single molecule mutation frequencies in cfDNA of lung cancer (n=89) and cancer-free (n=74) individuals, centered on the median within each mutation type. Region C>A mutation frequency changed preferentially between lung cancer and non-cancer samples, but not when class labels were randomly permuted (p=0.36, Wilcoxon rank sum test). p-values ​​were generated using Wilcoxon rank sum test and corrected for multiple comparisons using the Benjamini-Hochberg method. Horizontal dashed line indicates a p-value of 0.05. [Figure 19] Figure 19 is a plot showing the analysis of C>A sequence variation by flow cell and sequencing lane in non-cancer individuals. Single molecule C>A frequency and regional difference in single molecule C>A frequency across flow cells and sequencing lanes for all non-cancer individuals (n=158) from the LUCAS cohort. Sequencing background mutation rates vary by lane, and as a result, single molecule C>A frequencies of multiple samples within a sequencing lane are nearly the same, explaining 99% of their variance (p<0.0001, F-test), but this association was eliminated using the regional difference in single molecule C>A frequency obtained by the GEMINI approach (p=0.17, F-test). [Figure 20-1]Figures 20A-20K are a series of plots and schematics showing genome-wide fixed bins utilized for analysis of single molecule mutation frequencies and lung cancer detection in cfDNA. Figure 20A: Percentage similarity of bins found to be enriched for mutations in lung cancer and non-cancer samples for each training fold compared to the set of bins utilized in the fixed model identified from the analysis of all samples. High similarity between training folds indicates that bin selection was not driven by individual samples. Figure 20B: Chromosomal locations of bins enriched for mutations in cfDNA of lung cancer patients and cfDNA of individuals without cancer. Figure 20C: Compared to samples from individuals without cancer, samples from individuals with lung cancer have more C>A changes per genomic bin between samples in bins enriched in lung cancer and fewer of these changes in bins enriched in non-cancer. Figure 20D-20E: The mean number of evaluable bases per genomic bin and copy number were similar in bins enriched in lung cancer and bins enriched in non-cancer from non-cancer and lung cancer individuals. Copy number was estimated using ichorCNA. Figure 20F-20K: Bins in the fixed model were associated with replication timing, gene expression, A / B compartmentalization, and H3K9me3 abundance, GC content, but not sequence mappability. Figure 20F: Replication timing was obtained as a wavelet-smooth transform of six-compartment profiles representing different time points during replication in 1 kb bins derived from IMR90 cells47,49. A weighted average was then calculated in each 2.5 Mb bin. Large values ​​indicate earlier replication timing. Figure 20G: Gene expression was calculated as the sum of transcripts per million (TPM) overlapping each 2.5 Mb bin, weighted by transcript length averaged across TCGA NSCLC samples, and log-transformed as log10 (TPM).Figure 20H: A / B compartmentalization, representing primarily open and closed regions of the genome, respectively, was measured as the first eigenvector of the correlation matrix of mean methylation beta values ​​in 100 kb bins among TCGA lung cancer samples33. Figure 20I: H3K9me3 abundance, a known marker of heterochromatin, was obtained from ChIP-seq in A549 cells48 and shown as fold change in coverage in enriched samples compared to control samples derived from three pooled replicates. Figure 20J: GC content in each genomic bin was obtained from the hg19 reference genome. Bins enriched in lung cancer tend to be AT-rich (GC-poor) compared to bins enriched in non-cancer. This may be explained by our previous results that late replicated regions enriched for mutations in lung cancer have a lower GC content (Spearman's rho=0.83, p<0.0001). FIG. 20K: Mappability, which reflects how uniquely a 100-mer sequence aligns to a genomic region, was calculated as a weighted average in 2.5 Mb bins. [Figure 20-2] See description of Figure 20-1. [Figure 21-1]Figures 21A-21F are a series of plots showing the analysis of double base substitutions in tissue and plasma samples from lung cancer patients. Figure 21A: The number of somatic double base substitutions identified by the PCAWG consortium in lung cancer tissue samples from smoking individuals (n=65) revealed a higher number of CC>AA changes compared to other double mutations. The solid horizontal lines indicate the median number of each mutation type across individuals. Figure 21B: The ratio of single molecule CC>AA frequency when CC or CC>AA in read 1 and GG or GG>TT in read 2 (R1CC,R2GG) to when GG or GG>TT in read 1 and CC or CC>AA in read 2 (R1GG,R2CC) aggregated across samples in the high-risk LUCAS cohort. Background CC>AA changes represent alterations observed only in a single cfDNA fragment in cancer-free individuals, whereas likely somatic alterations represent alterations observed only in individual samples from lung cancer patients and in multiple cfDNA fragments. Within the high-risk LUCAS cohort, 67 private CC>AA changes were observed in more than one fragment among 89 lung cancer individuals, and only one such change was observed among 74 cancer-free individuals, indicating that the origin of most of these alterations is likely somatic. Bars represent 95% bootstrap confidence intervals of the ratios. Background CC>AA changes were often detected as R1CC,R2GG, but no imbalance in likely somatic CC>AA changes was observed, indicating enrichment of likely artifactual background CC>AA changes detected as R1CC,R2GG. Figure 21C: Sequence context surrounding CC>AA changes (+ / - 5bp) in the high-risk LUCAS cohort. Here, the number of mutations is shown for each group. The total height of the letter at each position indicates the information content of the position measured in bits. FIG. 21D: Single molecule CC>AA frequency was elevated in lung cancer individuals compared to non-cancer individuals, with greater separation observed after filtering of CC>AA changes detected as R1GG, R2CC.Figures 21E-21F: Single molecule CC>AA frequency was positively correlated with regional differences in single molecule C>A frequency in cfDNA (Figure 21E) and lung tumors (Figure 21F) after filtering for CC>AA alterations detected as R1CC, R2GG. [Figure 21-2] See description of Figure 21-1. [Figure 22] Figures 22A-22F are a series of plots demonstrating the effect of clinical features on GEMINI scores in non-cancer individuals in the LUCAS cohort. GEMINI scores are shown for men (n=87) and women (n=71) in Figure 22A, for individuals with (n=43) or without (n=115) autoimmune disease in Figure 22B, for individuals with (n=28) or without (n=130) COPD in Figure 22C, and for individuals of different ages in Figure 22D, and compared with CRP (mg / L) and IL-6 levels (pg / mL) in Figures 22E-22F. [Diagram 23] Figures 23A, 23B are a series of plots demonstrating that GEMINI scores reflect tumor DNA content in cfDNA. Figure 23A: GEMINI scores in cancer-free and lung cancer individuals with different ctDNA levels in the high-risk LUCAS cohort. A score of >0.55 reflects a positive test for detecting lung cancer with a specificity of 80%. b, GEMINI scores in cirrhotic and liver cancer individuals in the liver cancer cohort with <3% or ≥3% ctDNA. A score of >0.86 reflects a positive test for detecting liver cancer with a specificity of 80%. ichorCNA was used to estimate the ctDNA percentage in each sample. [Figure 24A]Figures 24A, 24B are a series of ROC curves demonstrating the performance of GEMINI or the combined GEMINI / DELFI approach for lung cancer detection. Figure 24A: ROC curves for lung cancer detection in the high-risk LUCAS cohort using GEMINI or the combined GEMINI / DELFI approach in the subset of patients with stage II-IV disease and who smoked >40 pack-years. Figure 24B: ROC curves for lung cancer detection in the high-risk LUCAS cohort using GEMINI or the combined GEMINI / DELFI approach in the subset of patients with adenocarcinoma, squamous cell carcinoma, or small cell lung cancer and who smoked >40 pack-years. Performance for stage I disease is shown in Figures 4F, 4H. [Figure 24B] See legend to Figure 24A. [Diagram 25] Figure 25 is a graph that demonstrates the GEMINI and DELFI scores and their combined performance to detect cancer in the LUCAS cohort. The GEMINI and DELFI scores of each patient in the high-risk LUCAS cohort (n=163) are shown. The vertical and horizontal dashed lines indicate the thresholds for positive GEMINI and DELFI tests at 80% specificity, respectively, and the black circles indicate positive tests by the combined approach at the same specificity. Some individuals with cancer were detected by one approach but not the other, and more lung cancer individuals were detected separately using the combined score compared to either approach alone. [Figure 26] Figure 26 is a graph demonstrating GEMINI / DELFI score and clinical outcome in lung cancer patients. Lung cancer patients (n=89) in the high-risk LUCAS cohort were stratified into two groups based on the median GEMINI / DELFI score among lung cancer patients of 0.84. Overall survival of patients with a GEMINI / DELFI score of ≥0.84 (yellow) was significantly worse than that of patients with a GEMINI / DELFI score of <0.84 (blue) (p=0.004, Log-rank test). [Figure 27A]Figures 27A-27D are a series of graphs and plots showing a comparison of cfDNA features between non-cancer patients in the LUCAS, DECAMP, and AHN cohorts. Figure 27A: Mean genome-wide coverage in non-cancer samples across cohorts. Horizontal dashed lines represent the median coverage of samples in each cohort. Figure 27B: Regional differences in single molecule C>A frequency are similar between cohorts (p=0.17, Kruskal-Wallis test). Horizontal solid lines represent the median for each group. Figure 27C: For each non-cancer sample, the ratio of short (100-150bp) to long (151-220bp) fragments was calculated in 473 non-overlapping 5Mb bins and mean-centered. Median fragmentation profiles represent the median of these values ​​among samples in each bin and are highly correlated between cohorts (Pearson correlation coefficients >0.97 for each pairwise comparison). Figure 27D: Chromosome arm-level Z-scores in non-cancer samples are similar between cohorts (p>0.05 for each chromosome arm, Kruskal-Wallis test with Bonferroni correction). [Figure 27B] See legend to Figure 27A. [Figure 27C] See legend to Figure 27A. [Figure 27D-1] See legend to Figure 27A. [Figure 27D-2] See legend to Figure 27A. [Figure 28]Figures 28A-28C are a series of graphs demonstrating GEMINI scores and smoking exposure in lung cancer patients. Figure 28A: Single molecule C>A frequency was similar in non-smokers with lung cancer (n=3) or without lung cancer (n=34) in the LUCAS cohort. In current or former smokers in the high-risk group with a smoking history of ≧20 pack-years and age 50-80 years, the single molecule C>A frequency was slightly higher in individuals with lung cancer (n=89) compared to individuals without lung cancer (n=74). Figure 28B: GEMINI scores were similar in non-smokers with lung cancer (n=3) or without lung cancer (n=34). The GEMINI scores of individuals with lung cancer (n=89) in the high-risk group were higher compared to individuals without lung cancer (n=74). Similarly, for individuals with a 40 pack-year smoking history and age 50-80, individuals with lung cancer (n=63) had higher GEMINI scores compared to individuals without lung cancer (n=46). Figure 28C: Current / former smokers aged 50-80 with lung cancer (n=32) and without lung cancer (n=14) in the validation cohort, as well as the subsets with a 40 pack-year smoking history and with lung cancer (n=18) and without lung cancer (n=5), had higher GEMINI scores. [Figure 29] Figure 29 shows a graph showing principal coordinate analysis in cancer patients after excluding the most frequent mutation types. Regional differences in single molecule mutation frequencies between NSCLC, SCLC, and HCC were calculated using a leave-one-out procedure for C>G, C>T, T>A, and T>G mutations, resulting in 12 feature values. From these 12 feature values, a Euclidean distance matrix was created that reflects pairwise differences between samples. Principal coordinate analysis of the Euclidean distance matrix revealed a reduced separation of samples by cancer type compared to when C>A and T>C mutations were also analyzed (Figure 5F). [Diagram 30]Figure 30 is a series of graphs showing GEMINI scores and MAF levels during therapy. Individuals with a smoking history and availability of targeted deep sequencing11 and low-coverage whole genome sequencing data13 were analyzed before and during treatment with tyrosine kinase inhibitors (arrows indicate treatment initiation). GEMINI scores were associated with the highest mutant allele fraction at each time point (Spearman's rho=0.50, p=0.03). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0056] Detailed Description Somatic mutations are a hallmark of tumorigenesis and may be useful for noninvasive diagnosis of cancer. However, detecting somatic alterations in circulation has been a challenge due to the limited number of tumor-derived molecules in cell-free DNA (cfDNA). An ultrasensitive analysis of single cfDNA molecules to detect genome-wide somatic mutation frequencies is developed herein and found that cancer patients have altered chromatin organization-associated mutation profiles compared to healthy individuals. Combining genome-wide cfDNA mutation profiles with fragmentation features followed by CT imaging detected 95% of cancer patients across stages and subtypes, including 95% of stage I and II patients, with a combined specificity of 90%. The model was independently validated in a separate screening cohort of high-risk individuals with early stage lung cancer. Genome-wide mutation profiles were able to distinguish between individuals with small cell lung cancer and those with non-small cell lung cancer and identify lung cancer earlier than standard approaches. This approach paves the way for noninvasive cancer detection using a combination of genome-wide mutation and fragmentation features of cfDNA, which may facilitate cancer screening.

[0057] GEMINI Although sequence alterations are abundant in cancer genomes, the proportion of fragments with tumor-specific (somatic) mutations in cell-free DNA (cfDNA) is often low 7,8 Therefore, it is difficult to detect genuine variants in the background noise due to sequence changes introduced during library construction, gene selection, PCR amplification, and sequencing. Significant efforts have been made to detect mutations present at low frequencies in cfDNA. However, these methods have typically relied on deep sequencing and have been limited to interrogating specific genes that constitute a small subset of the genome. 9-11 Due to the low number of genomic equivalents from cancer cells in cfDNA, such approaches have limited efficacy in detecting the presence of cancer, especially in early disease. 12-14 Furthermore, sequence alterations in cfDNA can originate from white blood cells (WBCs), thus undermining the use of sequence variants to detect cancer patients. 7,15,16 .

[0058] Disclosed herein is a method for detecting genome-wide mutation rates ( ge nome-wide m utational i Incidence for n on i A method called Gemini (gemini non-vasive detection of cancer) has identified a very large number of tumor-derived alterations in cfDNA for cancer detection (Figure 1). The method is based on sequencing individual cfDNA molecules to estimate the mutation frequency and type of alteration across the genome using non-overlapping bins ranging in size from thousands to millions of bases. For each individual, the mutation profile in genomic regions more frequently altered in cancer is compared to a profile derived from regions more frequently mutated in normal cfDNA to determine multiregional differences in the mutation profile. In this way, the GEMINI approach enriches for mutations that are likely to be somatic while taking into account individual variability in the total number of mutations.

[0059] Thus, in certain embodiments, a method of determining the frequency of somatic mutations in a subject includes extracting cell-free DNA (cfDNA) from a biological sample of the subject; generating a genomic library from the extracted cfDNA; sequencing individual cfDNA molecules to obtain a mutation profile; determining multiregional differences in the mutation profile; and determining the frequency of somatic mutations in the subject.

[0060] The generation of genome-wide mutation profiles included identifying mutations in the sequences of individual cfDNA molecules. The subject's genome-wide mutation profile is determined using non-overlapping bins that span a size range of at least about 100 bases to at least about 20 million bases. In certain embodiments, the subject's genome-wide mutation profile is determined using non-overlapping bins that span a size range of at least about 500 bases to at least about 15 million bases. In certain embodiments, the subject's genome-wide mutation profile is determined using non-overlapping bins that span a size range of at least about 750 bases to at least about 10 million bases. In certain embodiments, the subject's genome-wide mutation profile is determined using non-overlapping bins that span a size range of at least about 900 bases to at least about 10 million bases. In certain embodiments, the subject's genome-wide mutation profile is determined using non-overlapping bins that span a size range of at least about 1000 bases to at least about 5 million bases.

[0061] In certain embodiments, the frequency and mutation type of single molecule somatic mutations across the genome of a subject is diagnostic of cancer when compared to the frequency and mutation type of single molecule somatic mutations across the genome of a normal subject.

[0062] In certain embodiments, the frequency of somatic mutations at various loci is indicative of cancer, hi certain embodiments, the type of mutation is indicative of cancer.

[0063] cfDNA Fragmentation Profile: The cfDNA fragmentation profile may include one or more cfDNA fragmentation patterns. The cfDNA fragmentation pattern may include any suitable cfDNA fragmentation pattern. Examples of cfDNA fragmentation patterns include, but are not limited to, median fragment size, fragment size distribution, ratio of small cfDNA fragments to large cfDNA fragments, and cfDNA fragment coverage. In some embodiments, the cfDNA fragmentation pattern includes two or more (e.g., two, three, or four) of median fragment size, fragment size distribution, ratio of small cfDNA fragments to large cfDNA fragments, and cfDNA fragment coverage. In some embodiments, the cfDNA fragmentation profile may be a genome-wide cfDNA profile (e.g., genome-wide cfDNA profile in a genome-wide window). In some embodiments, the cfDNA fragmentation profile may be a target region profile. The target region may be any suitable part of the genome (e.g., a chromosomal region). Examples of chromosomal regions that can determine cfDNA fragmentation profile as described herein include, but are not limited to, parts of chromosomes (e.g., parts of 2q, 4p, 5p, 6q, 7p, 8q, 9q, 10q, 11q, 12q, and / or 14q) and chromosomal arms (e.g., chromosome mm of 8q, 13q, 11q, and / or 3p). In some embodiments, cfDNA fragmentation profile can include two or more target region profiles.

[0064] In some embodiments, cfDNA fragmentation profile can be used to identify the change (e.g., alteration) of cfDNA fragment length.The alteration can be genome-wide alteration or one or more target regions / locus alteration.The target region can be any region that contains one or more cancer-specific alterations. In some embodiments, the cfDNA fragmentation profile can be used to identify (e.g., simultaneously identify) about 10 alterations to about 500 alterations (e.g., about 25 to about 500, about 50 to about 500, about 100 to about 500, about 200 to about 500, about 300 to about 500, about 10 to about 400, about 10 to about 300, about 10 to about 200, about 10 to about 100, about 10 to about 50, about 20 to about 400, about 30 to about 300, about 40 to about 200, about 50 to about 100, about 20 to about 100, about 25 to about 75, about 50 to about 250, or about 100 to about 200 alterations).

[0065] The cfDNA fragmentation profile can be obtained using any suitable method. In some embodiments, cfDNA from a mammal (e.g., a mammal having or suspected of having cancer) can be processed into a sequencing library, which can be subjected to whole genome sequencing (e.g., low-coverage whole genome sequencing), mapped to the genome, and analyzed to determine cfDNA fragment lengths. The mapped sequences can be analyzed in non-overlapping windows covering the genome. The windows can be of any suitable size. For example, the windows can be thousands to millions of bases in length. As a non-limiting example, the windows can be about 5 megabases (Mb) in length. Any suitable number of windows can be mapped. For example, tens to thousands of windows can be mapped in the genome. For example, hundreds to thousands of windows can be mapped in the genome. The cfDNA fragmentation profile can be determined within each window.

[0066] In some embodiments, the methods and materials described herein can also include machine learning.For example, machine learning can be used to identify mutation frequency, fragmentation profile change (for example, by using cfDNA fragment coverage, cfDNA fragment size, chromosome coverage, and mtDNA).

[0067] Treatment The method embodied herein comprises the steps of determining that a mammal has cancer.The method comprises the steps of extracting cell-free DNA (cfDNA) from the biological sample of the subject; preparing a genomic library from the extracted cfDNA; sequencing each cfDNA molecule to obtain a mutation profile; determining the multi-region difference in the mutation profile and the frequency of somatic mutation in the subject; and administering a cancer treatment to the subject.

[0068] In certain embodiments, the subject is diagnosed with cancer, for example, early cancer.In certain embodiments, the type of cancer is identified, and the cancer is treated with various therapeutic agents, including cancer type-specific therapeutic agents.In certain embodiments, the cancer comprises colorectal cancer, lung cancer, breast cancer, gastric cancer, pancreatic cancer, bile duct cancer, brain cancer, or ovarian cancer.In certain embodiments, the lung cancer is small cell lung cancer (SCLC).In certain embodiments, the lung cancer is non-small cell lung cancer (NSCLC).

[0069] The cancer treatment may be surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormone therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy, targeted therapy, or any combination thereof. The method may also include administering a cancer treatment to the mammal (e.g., surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormone therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy, targeted therapy, or any combination thereof). The mammal can be monitored for the presence of cancer after administration of the cancer treatment.

[0070] Cancer therapy also commonly involves a variety of combination therapies with chemical- and radiation-based treatments. Combination chemotherapy includes, for example, cisplatin (CDDP), carboplatin, procarbazine, mechlorethamine, cyclophosphamide, camptothecin, ifosfamide, melphalan, chlorambucil, busulfan, nitrosourea, dactinomycin, daunorubicin, doxorubicin, bleomycin, plicomycin, mitomycin, etoposide (VP16), tamoxifen, raloxifene, estrogen receptor binding agents, taxol, gemcitabien, navelbine, famesyl-protein transferase inhibitors, transplatinum, 5-fluorouracil, vincristine, vinblastine and methotrexate, temazolomide (water-based form of DTIC), or any analog or derivative variant of the above. The combination of chemotherapy and biological therapy is known as biochemotherapy. Chemotherapy may also be given in successive low doses, known as metronomic chemotherapy.

[0071] Still further combination chemotherapy includes, for example, alkylating agents, such as thiotepa and cyclosphosphamide; alkylsulfonates, such as busulfan, improsulfan, and piposulfan; aziridines, such as benzodopa, carboquone, meturedopa, and uredopa; altretamine, triethylenemelamine, triethylenephosphoramide, triethylenethiophosphoramide, and trimethylol melamine. ethylenimines and methylamelamines, including trimethylolomelamimes; acetogenins (especially bullatacin and bullatacinone); camptothecins (including the synthetic analog topotecan); bryostatin; callystatin; CC-1065 (including its synthetic analogs adozelesin, carzelesin, and biceresin); cryptophycins (especially cryptophycin 1 and cryptophycin 2); dolastatins; duocarmycins (including synthetic analogs KW-2189 and CB1-TM1); erytherobin; pancratistatin; sarcodictyin; spongistatins; nitrogen mustards, e.g., chlorambucil, chlornaphazine, chlorophosphamide, estramustine, ifosfamide, mechlorethamine, mechlorethamine oxide hydrochloride, melphalan, novem novembichin, phenesterine, prednimustine, trofosfamide, uracil mustard; nitrosoureas, such as carmustine, chlorozotocin, fotemustine, lomustine, nimustine, and ranimustine; antibiotics, such as enediyne antibiotics (e.g., calicheamicin, especially calicheamicin gamma ll and calicheamicin omega ll; dynemicins, including dynemicin A; bisphosphonates, such as clodronate; esperamicin;and neocarzinostatin chromophore and related chromoprotein enediyne antibiotic chromophores, aclacinomycin, actinomycin, authrarnycin, azaserine, bleomycin, cactinomycin, carabicin, carminomycin, carzinophilin, chromomycinis, dactinomycin, daunorubicin, detorubicin, 6-diazo-5-oxo-L-norleucine, doxorubicin (morpholino-doxin), sorbicin, cyanomorpholino-doxorubicin, 2-pyrrolino-doxorubicin, and deoxydoxorubicin), epirubicin, esorubicin, idarubicin, marcelomycin, mitomycins, such as mitomycin C, mycophenolic acid, nogalamycin, olivomycin, peplomycin, potfiromycin, puromycin, quelamycin, rodornbicin, streptonigrin, streptozocin, tubercidin, ubenimex, zinostatin, and zorubicin; antimetabolites such as methotrexate and 5-fluorouracil (5-FU); folic acid analogues such as denopterin, pteropterin, trimetrexate; purine analogues such as fludarabine, 6-mercaptopurine, thiamiprine, thioguanine; pyrimidine analogues such as ancitabine, azacitidine, 6-azauridine, carmofur, cytarabine, dideoxyuridine, doxifluridine, enocitabine, floxuridine; androgens such as calsterone, Dromostanolone propionate, epithiostanol, mepitiostane, testolactone; antiadrenal drugs such as mitotane, trilostane; folic acid supplements such as folinic acid; aceglatone; aldophosphamide glycosides; aminolevulinic acid; eniluracil; amsacrine; bestravcil; bisantrene; edatraxate; defofarnine; demecolcine; diaziquone; elformithine; elliptinium acetate; epothilone; etoglucide; gallium nitrate;Hydroxyurea;Lentinan;Lonidynin;Maytansinoids, such as maytansine and ansamitocin;Mitoguazone;Mitoxantrone;Mopidammol;Nitraerine;Pentostatin;Phenamet;Pirarubicin;Losoxantrone;Podophyllinic acid;2-Ethylhydrazide;Procarbazine;PSK polysaccharide complex;Razoxane;Rhizoxin xin; sizofiran; spirogermanium; tenuazonic acid; triaziquone; 2,2',2''-trichlorotriethylamine; trichothecenes (especially T-2 toxin, verracurin A, roridin A, and anguidine); urethane; vindesine; dacarbazine; mannomustine; mitobronitol; mitolactol; pipobroman; gacytosine; arabinoside ("Ar a-C"); cyclophosphamide; taxoids, e.g., paclitaxel and docetaxel gemcitabine; 6-thioguanine; mercaptopurine; platinum coordination complexes, e.g., cisplatin, oxaliplatin, and carboplatin; vinblastine; platinum; etoposide (VP-16); ifosfamide; mitoxantrone; vincristine; vinorelbine; novantrone; teniposide; edatrexate; daunomycin; aminopterin; xeloda; ibandronate; irinotecan (e.g. For example, CPT-11); the topoisomerase inhibitor RPS2000; difluoromethylornithine (DMFO); retinoids, such as retinoic acid; capecitabine; carboplatin, procarbazine, plicomycin, gemcitabien, navelbine, farnesyl-protein transferase inhibitors, transplatinum, as well as pharmaceutically acceptable salts, acids, or derivatives of any of the foregoing;

[0072] Immunotherapeutics generally rely on the use of immune effector cells and molecules that target and destroy cancer cells. The immune effector may be, for example, an antibody specific for some marker on the surface of the tumor cell. The antibody alone may serve as the therapeutic effector or may recruit other cells that actually effect cell death. The antibody may also be conjugated with a drug or toxin (chemotherapeutic agent, radionuclide, ricin A chain, cholera toxin, pertussis toxin, etc.) and simply serve as a targeting agent. Alternatively, the effector may be a lymphocyte carrying a surface molecule that interacts directly or indirectly with the tumor cell target. Various effector cells include cytotoxic T cells and NK cells as well as genetically engineered variants of these cell types that have been modified to express chimeric antigen receptors.

[0073] Immunotherapy may include suppression of T regulatory cells (Tregs), myeloid-derived suppressor cells (MDSCs), and cancer-associated fibroblasts (CAFs). In some embodiments, immunotherapy is tumor vaccines (e.g., whole tumor cell vaccines, peptides, and recombinant tumor-associated antigen vaccines), or adoptive cell therapy (ACT) (e.g., T cells, natural killer cells, TILs, and LAK cells). T cells may be engineered with chimeric antigen receptors (CARs) or T cell receptors (TCRs) for specific tumor antigens. Chimeric antigen receptors (or CARs), as used herein, can refer to any engineered receptor specific to an antigen of interest that, when expressed in T cells, confers the specificity of the CAR to the T cells. Once created using standard molecular techniques, T cells expressing chimeric antigen receptors can be introduced into patients, such as techniques such as adoptive cell transfer. In some aspects, the T cells are CD4 and / or CD8 T cells that produce gamma-IFN and / or activated CD4 and / or CD8 T cells in the individual characterized by enhanced cytolytic activity compared to before administration of the combination. The CD4 and / or CD8 T cells may exhibit increased release of cytokines selected from the group consisting of IFN-gamma, TNF-alpha, and interleukins. The CD4 and / or CD8 T cells may be effector memory T cells. In certain embodiments, the CD4 and / or CDS effector memory T cells are CD44 high CD62L low The present invention is characterized in that it has expression of

[0074] The immunotherapy may be a cancer vaccine comprising one or more cancer antigens, in particular proteins or immunogenic fragments thereof, DNA or RNA encoding said cancer antigens, in particular proteins or immunogenic fragments thereof, cancer cell lysates, and / or protein preparations from tumor cells. A cancer antigen as used herein is an antigenic substance present in cancer cells. In principle, any protein produced in cancer cells that has an abnormal structure due to mutation can act as a cancer antigen. In principle, cancer antigens can be the products of mutated oncogenes and tumor suppressor genes, the products of other mutated genes, overexpressed or aberrantly expressed cellular proteins, cancer antigens produced by tumor viruses, oncofetal antigens, altered cell surface glycolipids and glycoproteins, or cell type specific differentiation antigens. Examples of cancer antigens include the abnormal products of ras and p53 genes. Other examples include tissue differentiation antigens, mutated protein antigens, tumor virus antigens, cancer testis antigens, and vascular or stroma specific antigens. Tissue differentiation antigens are tissue differentiation antigens that are specific to a certain type of tissue. These proteins are likely to be quite specific to cancer cells, since normal cells should not contain mutant protein antigens. Normal cells display normal protein antigens on their MHC molecules, whereas cancer cells display mutant versions. Some viral proteins are linked to the formation of cancer, and some viral antigens are cancer antigens. Cancer testis antigens are antigens that are expressed primarily in testicular germ cells, but also in fetal ovaries and trophoblasts. Some cancer cells aberrantly express these proteins, thus presenting these antigens and thus allowing attack by T cells specific for these antigens. An exemplary antigen of this type is Rindopepimut, a 14-mer intradermally injected peptide vaccine targeted against CTAG1 B and MAGEA1, as well as epidermal growth factor receptor vIII (EGFRvIII; exon 2-7 deletion) variants. Rindopepimut is particularly suitable for treating glioblastoma when combined with CD95 / CD95L signaling inhibitors as described herein.Similarly, proteins that are normally produced in very low amounts, but are dramatically increased in cancer cells, can also elicit an immune response. One example of such a protein is the enzyme tyrosinase, which is required for melanin production. Normally, tyrosinase is produced in minute amounts, but in melanoma cells, its levels are significantly elevated. Carcinoembryonic antigens are another important class of cancer antigens. Examples are alpha-fetoprotein (AFP) and carcinoembryonic antigen (CEA). These proteins are normally produced early in embryonic development and disappear by the time the immune system is fully developed. Thus, self-tolerance to these antigens does not develop. Abnormal proteins are also produced in cells infected with oncoviruses, such as EBV and HPV. Cells infected with these viruses contain latent viral DNA, which is transcribed and the resulting proteins generate an immune response. Cancer vaccines can include peptide cancer vaccines, which in some embodiments are personalized peptide vaccines. In some aspects, the peptide cancer vaccine is a multivalent long peptide vaccine, a multi-peptide vaccine, a peptide cocktail vaccine, a hybrid peptide vaccine, or a peptide-pulsed dendritic cell vaccine.

[0075] The immunotherapy may be an antibody, for example, part of a polyclonal antibody preparation, or a monoclonal antibody. The antibody may be a humanized antibody, a chimeric antibody, an antibody fragment, a bispecific antibody, or a single chain antibody. The antibodies disclosed herein include antibody fragments, including, but not limited to, Fab, Fab' and F(ab')2, Fd, single chain Fvs (scFv), single chain antibodies, disulfide-linked Fv (sdfv), and fragments containing either the VL or VH domain. In some aspects, the antibody or fragment thereof specifically binds to epidermal growth factor receptor (EGFR1, Erb-B1), HER2 / neu (Erb-B2), CD20, vascular endothelial growth factor (VEGF), insulin-like growth factor receptor (IGF-1R), TRAIL-receptor, epithelial cell adhesion molecule, carcinoembryonic antigen, prostate-specific membrane antigen, mucin-1, CD30, CD33, or CD40.

[0076] Examples of monoclonal antibodies include trastuzumab (anti-HER2 / neu antibody); pertuzumab (anti-HER2 mAb); cetuximab (chimeric monoclonal antibody against epidermal growth factor receptor EGFR); panitumumab (anti-EGFR antibody); nimotuzumab (anti-EGFR antibody); zalutumumab (anti-EGFR mAb); necitumumab (anti-EGFR mAb); MDX-210 (humanized anti-HER-2 bispecific antibody); MDX-210 (humanized anti-HER-2 bispecific antibody); MDX-447 (humanized anti-EGF receptor bispecific antibody); rituximab (chimeric mouse / human anti-CD20 mAb); obinutuzumab (anti-CD20 mAb); ofatumumab (anti-CD20 mAb); tositumomab-I131 (anti-CD20 mAb); ibritumomab tiuxetan (anti-CD20 mAb); bevacizumab (anti-VEGF mAb); ramucirumab (anti-VEGFR2 mAb); ranibizumab (anti-VEGF mAb); aflibercept (extracellular domains of VEGFR1 and VEGFR2 fused to IgGl Fc); AMG386 (angiopoietin-1 and -2 binding peptide fused to IgGl Fc); dalotuzumab (anti-IGF-1R mAb); gemtuzumab ozogamicin (anti-CD33 mAb); alemtuzumab (anti-Campath-1 / CD52 mAb); brentuximab vedotin (anti-CD30 mAb); catumaxomab (dual specificity mAb targeting epithelial cell adhesion molecule and CD3); naptumomab (anti-5T4 mAb); Girentuximab (anticarbonic anhydrase ix); or Farletuzumab (antifolate receptor).Other examples include Panorex™ (17-1A) (mouse monoclonal antibody); Panorex (MAb17-1A) (chimeric mouse monoclonal antibody); BEC2 (combined with an anti-idiotypic mAb mimicking the GD epitope) (with BCG); Oncolym (Lym-1 monoclonal antibody); SMART Ml95Ab, humanized 13'1 LYM-1 (Oncolym), Ovarex (B43.13, an anti-idiotypic mouse mAb); 3622W94 mAb that binds to the EGP40 (17-1A) pancarcinoma antigen located on adenocarcinoma; Zenapax (SMART anti-Tac (IL-2 receptor); SMART M195 Ab, humanized Ab, humanized); NovoMAb-G2 (pancarcinoma-specific Ab); TNT (chimeric mAb against histone antigen); TNT (chimeric mAb against histone antigen); Gliomab-H (monoclonal humanized Ab); GNI-250 Mab; EMD-72000 (chimeric-EGF antagonist); LymphoCide (humanized IL.L.2 antibody); and antibodies such as MDX-260 bispecific targeting GD-2, ANA Ab, SMART IDIO Ab, SMART ABL 364 Ab, or ImmuRAIT-CEA.Further examples of antibodies include zanulimumab (anti-CD4 mAb), keliximab (anti-CD4 mAb); ipilimumab (MDX-101; anti-CTLA-4 mAb); tremilimumab (anti-CTLA-4 mAb); (daclizumab (anti-CD25 / IL-2R mAb); basiliximab (anti-CD25 / IL-2R mAb); MDX-1106 (anti-PDl mAb); antibodies against GITR; GC1008 (anti-TGF-β antibody); metelimumab / CAT-192 (anti-TGF-β antibody); lerdelimumab / CAT-152 (anti-TGF-β antibody); ID11 (anti-TGF-β antibody); denosumab (anti-RANKL mAb);BMS-663513 (humanized anti-4-lBB mAb);SGN-40 (humanized anti-CD40 mAb);CP870,893 (human anti-CD40 mAb);Infliximab (chimeric anti-TNF mAb);Adalimumab (human anti-TNF mAb);Certolizumab (humanized Fab anti-TNF);Golimumab (anti-TNF);Etanercept (extracellular domain of TNFR fused to IgG1 Fc);Belatacept (extracellular domain of CTLA-4 fused to Fe);Abatacept (extracellular domain of CTLA-4 fused to Fe);Belimumab (anti-B lymphocyte stimulator);Muromonab-CD3 (anti-CD3 mAb);Otelixizumab (anti-CD3 mAb);Teplizumab (anti-CD3 mAb);Tocilizumab (anti-IL6R mAb); REGN88 (anti-IL6R mAb); ustekinumab (anti-IL-12 / 23 mAb); briakinumab (anti-IL-12 / 23 mAb); natalizumab (anti-α4 integrin); vedolizumab (anti-α4β7 integrin mAb); T1h (anti-CD6 mAb); epratuzumab (anti-CD22 mAb); efalizumab (anti-CD11a mAb); and atacicept (extracellular domain of transmembrane activator-calcium-modulating ligand interactor fused to Fc).

[0077] system In some examples, the present disclosure provides a system, method, or kit that may include software code that executes on data analysis, computational hardware implemented in a measurement device (e.g., laboratory equipment, e.g., sequencing device). The software can be stored in memory and executed on one or more hardware processors. The software can be organized into routines or packages that can communicate with each other. A module may comprise one or more software routines / packages that execute on one or more devices / computers and potentially on one or more devices / computers. For example, an analysis application or system may comprise at least a data reception module, a data pre-processing module, a data analysis module (that can operate on one or more types of genomic data), a data interpretation module, or a data visualization module.

[0078] The data receiving module can connect laboratory hardware or instrumentation along with a computer system that processes the laboratory data. The data pre-processing module can perform operations on the data ready for analysis. Examples of operations that can be applied to the data in the pre-processing module include affine transformations, denoising operations, data cleaning, reformatting, or sub-sampling. The data analysis module can be specialized in analyzing genomic data from one or more genomic materials, for example, taking assembled genomic sequences and performing probabilistic and statistical analysis to identify abnormal patterns associated with a disease, pathology, state, risk, condition, or phenotype. The data interpretation module can use analytical methods, for example, statistical, mathematical, or biology-derived analytical methods, to support an understanding of the relationship between the identified abnormal patterns and health status, functional status, prognosis, or risk. The data analysis module and / or data interpretation module can comprise one or more machine learning models that can be executed in hardware, for example, in hardware executing software that embodies the machine learning models. The data visualization module may use methods of mathematical modeling, computer graphics, or rendering to generate data visual displays that may facilitate understanding or interpretation of the results. The present disclosure provides computer systems programmed to perform the methods of the present disclosure.

[0079] In some embodiments, the methods disclosed herein may include computational analysis of nucleic acid sequencing data of samples from one individual or multiple individuals. The analysis may identify variants inferred from sequence data, such as identifying sequence variants based on probability modeling, statistical modeling, mechanistic modeling, network modeling, or statistical inference. Non-limiting examples of analysis methods include principal component analysis, autoencoders, singular value decomposition, Fourier bases, wavelets, discriminant analysis, regression, support vector machines, tree-based methods, networks, matrix factorization, and clustering. Non-limiting examples of variants include germline variations or somatic mutations. In some cases, the variant may refer to a known variant. The known variant may be scientifically confirmed or reported in the literature. In some cases, the variant may refer to a putative variant associated with a biological change. The biological change may be known or unknown. In some cases, the putative variant may be reported in the literature but has not yet been biologically confirmed. Alternatively, a putative variant may have never been reported in the literature, but may be inferred based on the computational analyses disclosed herein. In some instances, a germline variant may refer to a nucleic acid that induces natural or normal variation.

[0080] In certain embodiments, the computer system comprises a central processing unit (CPU, also referred to herein as "processor" and "computer processor"), which may be a single-core processor or a multi-core processor, or multiple processors for parallel processing; memory (e.g., cache, random access memory, read-only memory, flash memory, or other memory); electronic storage (e.g., hard disk), a communication interface (e.g., network adapter) for communicating with one or more other systems; and peripherals, e.g., adapters for cache, other memory, data storage, and / or electronic displays. The memory, storage, interfaces, and peripherals may be in communication with the CPU through a communication bus (solid line), e.g., a motherboard. The storage may be a data storage (or data repository) for storing data. One or more analytical feature inputs may be input from one or more measurement devices. Examples of analytes and measurement devices are described herein.

[0081] The computer system may be operably connected to a computer network ("network") with the aid of a communication interface. The network may be the Internet, an Internet and / or an extranet, or an intranet and / or an extranet in communication with the Internet. The network may optionally be a telecommunications and / or data network. The network may comprise one or more computer servers that may enable distributed computing, e.g., cloud computing on a network ("cloud"), for performing various aspects of the analysis, calculation, and production of the present disclosure, such as actuating valves or pumps to transfer reagents or samples from one chamber to another, or heating the sample (e.g., during an amplification reaction), processing the sample and / or other aspects of the assay, performing sequencing analysis, measuring a set of values ​​representative of molecular classes, identifying a set of features and feature vectors from the assay data, processing the feature vectors with a machine learning model to obtain an output classification, and training the machine learning model (e.g., iteratively searching for optimal values ​​of machine learning model parameters). Such cloud computing may be provided by cloud computing platforms such as, for example, Amazon Web Services (AWS), Microsoft Azure, Google Cloud Platform, and IBM Cloud. The network may in some cases, with the help of a computer system, implement a peer-to-peer network that may enable devices connected to the computer system to act as clients or servers.

[0082] The CPU may execute a sequence of machine-readable instructions, which may be embodied in the form of a program or software. The instructions may be stored in a memory location, such as a memory. The instructions may be directed to the CPU, which may then be programmed or otherwise configured to execute the methods of the present disclosure. The CPU may be part of a circuit, such as an integrated circuit. One or more other components of the system may be included in the circuit. In some cases, the circuit is an application specific integrated circuit (ASIC).

[0083] The storage device can store files, such as drivers, libraries, and saved programs. The storage device can store user data, such as user preferences and user programs. The computer system may optionally include one or more additional data storage devices external to the computer system, for example, one or more additional data storage devices located on a remote server in communication with the computer system through an intranet or the Internet.

[0084] The computer system can communicate with one or more remote computer systems through a network. For example, the computer system can communicate with a user's remote computer system. Examples of remote computer systems include a personal computer (e.g., a portable PC), a slate or tablet PC (e.g., an Apple® iPad, a Samsung® Galaxy Tab), a phone, a smartphone (e.g., an Apple® iPhone, an Android-enabled device, a Blackberry®), or a personal digital assistant. A user can access the computer system through a network.

[0085] The methods described herein can be implemented as machine-executable code (e.g., a computer processor) stored in an electronic storage location, e.g., memory or electronic storage device, of a computer system. Machine-executable or machine-readable code can be provided in the form of software. In use, the code can be executed by a CPU. In some cases, the code can be retrieved from storage device and stored in memory for rapid access by the CPU. In some situations, electronic storage device can be omitted and machine-executable instructions are stored in memory.

[0086] The code may be pre-compiled and configured for use with a machine having a processor adapted to execute the code, or may be compiled during run-time. The code may be provided in a programming language that can be selected to enable the code to be executed in a pre-compiled or as-compiled manner.

[0087] Aspects of the systems and methods provided herein, e.g., computer systems, can be embodied in programming. Various aspects of the technology can be considered as "products" or "articles of manufacture" that typically take the form of machine (or processor) executable code and / or associated data carried or embodied on some type of machine-readable medium. The machine executable code can be stored in electronic storage, e.g., memory (e.g., read-only memory, random access memory, flash memory), or hard disk. A "storage" type medium can include any or all of the computer's tangible memory, processor, etc., or its associated modules, e.g., various semiconductor memories, tape drives, disk drives, etc., that can provide non-transitory storage at any time for software programming. All or a portion of the software may sometimes be communicated over the Internet or various other telecommunications networks. Such communication may, for example, enable loading of the software from one computer or processor to another, e.g., from a management server or host computer to an application server computer platform. Thus, other types of media that may have software elements include light waves, radio waves, and electromagnetic waves, e.g., those used across physical interfaces between local devices, through wired and optical land-line networks, and across various air-links. The physical elements that carry such waves, e.g., wired or wireless links, optical links, etc., may also be considered media that have software. Unless limited to non-transitory tangible "storage" media, the term computer or machine "readable medium" as used herein refers to any medium that participates in providing instructions to a processor for execution.

[0088] Thus, a machine-readable medium, such as a computer executable code, may take many forms, including but not limited to a tangible storage medium, a carrier wave medium, or a physical transmission medium. Non-volatile storage media include, for example, optical or magnetic disks, such as any storage device present in any computer, and may be used to execute the databases shown in the figures. Volatile storage media include dynamic memory, such as the main memory of such a computer platform. Tangible transmission media include coaxial cables; copper wire and fiber optics, including the wires that comprise a bus within a computer system.

[0089] Carrier wave transmission media may take the form of electric or electromagnetic signals, or acoustic or light waves, such as those generated during radio frequency (RF) and infrared (IR) data communications. Common forms of computer readable media thus include, for example, floppy disks, flexible disks, hard disks, magnetic tapes, any other magnetic media, CD-ROMs, DVDs or DVD-ROMs, any other optical media, punch cards paper tapes, any other physical storage media with patterns of holes, RAMs, ROMs, PROMs and EPROMs, FLASH-EPROMs, any other memory chips or cartridges, carrier waves carrying data or instructions, cables or links carrying such carrier waves, or any other medium from which a computer can read programming code and / or data. Many of these types of computer readable media can be involved in carrying one or more sequences of one or more instructions to a processor for execution.

[0090] The computer system may include or be in communication with an electronic display that includes a user interface (UI), for example, to indicate the current stage of sample processing or assay (e.g., a particular step, e.g., lysis step, or sequencing step being performed). Input is received by the computer system from one or more measurements. Examples of UIs include, but are not limited to, graphical user interfaces (GUIs) and web-based user interfaces. The algorithm may, for example, process and / or assay the sample, perform a sequencing analysis, measure a set of values ​​representative of molecular classes, identify a set of features and feature vectors from the assay data, process the feature vectors with a machine learning model to obtain an output classification, and train the machine learning model (e.g., iteratively search for optimal values ​​of machine learning model parameters).

[0091] In some embodiments, a system, e.g., a laptop, desktop, iPad, mobile device, etc., capable of executing one or more algorithms to determine changes in cfDNA mutational profile, mutation frequency, and / or fragmentation profile, classifies a subject as a cancer patient based on the subject's cfDNA mutational profile, mutation frequency, and / or fragmentation. Additionally, these systems may implement machine learning algorithms (e.g., penalized logistic regression) that can be used to create models of high-risk and low-risk general populations, etc., as described in Mathios et al. (Mathios D, Johansen JS, Cristiano S, Medina JE, Phallen J, Larsen KR, et al. Detection and characterization of lung cancer using cell-free DNA fragmentomes. Nat Commun. 2021;12(1):5060) with features and coverage from transcription factor binding sites. These models can be trained on the subject cohort for 10 iterations using 5-fold cross-validation, and scores for each sample are calculated by averaging across the iterations and evaluated using AUC-ROC. For example, the first model used high-risk non-cancer and HCC patients, while the second model used non-cancer individuals without liver pathology. The locked high-risk model trained on the cohort was applied to a second, different cohort to make cancer predictions on an external validation set. A "class label" can be applied to each sample, indicating sample classification for any number of input features. For example, the class label for a cohort set may indicate identity information for cfDNA mutation profile, mutation frequency, and / or fragmentation profile based on genomic location, etc. The resulting training set is provided to a machine learning unit, such as a neural network or support vector machine. Using the training set, the machine learning unit may create a model to classify samples according to cfDNA mutation profile, mutation frequency, and / or fragmentation profile.

[0092] In some embodiments, methods are provided for generating a trained classifier comprising: (a) preparing a plurality of distinct classes, each class representing a set of subjects (e.g., from one or more cohorts) having a common characteristic; (b) preparing a multi-metric model representative of cell-free DNA molecules from each of a plurality of samples belonging to each class, thereby providing a training dataset; and (c) training a learning algorithm on the training dataset to generate one or more trained classifiers, each trained classifier classifying test samples into one or more of the plurality of classes.

[0093] As an example, the trained classifier may use a learning algorithm selected from the group consisting of random forests, neural networks, support vector machines, and linear classifiers. Each of the multiple distinct classes may be selected from the group consisting of healthy, breast cancer, colon cancer, lung cancer, pancreatic cancer, prostate cancer, ovarian cancer, melanoma, and liver cancer.

[0094] The trained classifier can be applied to a method for classifying a sample from a subject.The classification method can include: (a) preparing a multimetric model representing cell-free DNA molecules derived from a test sample from a subject; and (b) classifying the test sample using the trained classifier.After the test sample is classified into one or more classes, a therapeutic intervention can be performed on the subject based on the classification of the sample.

[0095] In some embodiments, the training set is provided to a machine learning unit, such as a neural network or a support vector machine. Using the training set, the machine learning unit can create a model to classify samples according to treatment response to one or more therapeutic inventions. This is also called "calling." The developed model can use information from any part of the test vector.

[0096] In general, machine learning can be used to reduce a data set created from all (primary sample / analyte / test) combinations to an optimal predictive set of features, for example, that meets a specified criterion. In various examples, statistical learning and / or regression analysis can be applied. Simple to complex and small to large models making various modeling assumptions can be applied to the data in a cross-validation paradigm. Simple to complex includes considerations of linear to nonlinear and non-hierarchical to hierarchical representation of features. Small to large models include considerations of the size of the basis vector space to project the data to the number of interactions between features included in the modeling process.

[0097] Machine learning methods can be used to evaluate commercial test modalities that are optimal for cost / performance / commercial reach as defined in the initial questions. Threshold checks can be performed, i.e., if the method applied to a holdout data set not used in cross-validation is better than the initialized constraints, the assay is locked and production begins. For example, assay performance thresholds may include a desired minimum accuracy, positive predictive value (PPV), negative predictive value (NPV), clinical sensitivity, clinical specificity, area under the curve (AUC), or combinations thereof. For example, the desired minimum accuracy, PPV, NPV, clinical sensitivity, clinical specificity, or a combination thereof may be at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 81%, at least about 82%, at least about 83%, at least about 84%, at least about 85%, at least about 86%, at least about 87%, at least about 88%, at least about 89%, at least about 90%, at least about 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, or at least about 99%. As another example, a desirable minimum AUC may be at least about 0.50, at least about 0.55, at least about 0.60, at least about 0.65, at least about 0.70, at least about 0.75, at least about 0.80, at least about 0.81, at least about 0.82, at least about 0.83, at least about 0.84, at least about 0.85, at least about 0.86, at least about 0.87, at least about 0.88, at least about 0.89, at least about 0.90, at least about 0.91, at least about 0.92, at least about 0.93, at least about 0.94, at least about 0.95, at least about 0.96, at least about 0.97, at least about 0.98, or at least about 0.99.A subset of assays may be selected from the set of assays to be performed on a particular sample based on the total cost of performing the subset of assays, subject to assay performance thresholds, such as desired minimum accuracy, positive predictive value (PPV), negative predictive value (NPV), clinical sensitivity, clinical specificity, area under the curve (AUC), and combinations thereof. If the thresholds are not met, the assay engineering procedure can revert to the constraint setting for possible relaxation or to the wet lab to change the parameters under which the data were obtained. Given a clinical question, biological constraints, budget, lab machines, etc. may constrain this problem.

[0098] In certain embodiments, the computational processing of the machine learning method may include methods of statistics, mathematics, biology, or any combination thereof. In various examples, any one of the computational processing methods may include dimensionality reduction methods, logistic regression, dimensionality reduction, principal component analysis, autoencoders, singular value decomposition, Fourier basis, singular value decomposition, wavelets, discriminant analysis, support vector machines, tree-based methods, random forests, gradient boost trees, logistic regression, matrix factorization, network clustering, statistical testing, and neural networks.

[0099] In certain embodiments, the computerized machine learning methods may include logistic regression, multiple linear regression (MLR), dimensionality reduction, partial least squares (PLS) regression, principal component regression, autoencoder, variational autoencoder, singular value decomposition, Fourier basis, wavelets, discriminant analysis, support vector machines, decision trees, classification and regression trees (CART), tree-based methods, random forests, gradient boosted trees, logistic regression, matrix factorization, multidimensional scaling (MDS), dimensionality reduction methods, t-distributed stochastic neighbor embedding (t-SNE), multilayer perceptron (MLP), network clustering, neuro-fuzzy, neural networks (shallow and deep), artificial neural networks, Pearson's product moment correlation coefficient, Spearman's rank correlation coefficient, Kendall tau rank correlation coefficient, or any combination thereof. In some examples, the computational methods are supervised machine learning methods, including, for example, regression, support vector machines, tree-based methods, and neural networks. In some examples, the computational methods are unsupervised machine learning methods, including, for example, clustering, networks, principal component analysis, and matrix factorization.

[0100] In the case of supervised learning, the training samples (e.g., in the thousands) may include measurement data (e.g., of various analytes) and known labels that may be determined by other time-consuming processes such as imaging of the subject and analysis by a trained practitioner. Examples of labels may include classification of the subject, e.g., a discrete classification of whether the subject has cancer or not, or a continuous classification indicating a probability of a discrete value (e.g., risk or score). The learning module can optimize the parameters of the model such that a quality metric (e.g., the accuracy of prediction for the known label) is achieved by one or more specified criteria. The quality metric determination can be performed for any arbitrary function, including the set of all risk, loss, utility, and decision functions. A gradient can be used with the learning step (e.g., a measure of how much the parameters of the model should be updated for a given time step of the optimization process).

[0101] As mentioned above, examples can be used for a variety of purposes. For example, plasma (or other samples) can be collected from subjects that exhibit symptoms of a condition (e.g., known to have a condition) and healthy subjects. Genetic data (e.g., cfDNA) can be obtained and analyzed to obtain a variety of different features. The features can include features based on genome-wide analysis. These features can form a feature space that is searched, stretched, rotated, translated, and linearly or non-linearly transformed to create an accurate machine learning model that can distinguish healthy subjects from subjects with a condition (e.g., identify a diseased or non-disease state of a subject). This data and the output derived from the model (which can include probabilities of a condition, a stage (level) of a condition, or other values) can be used to create another model that can be used to recommend further procedures, such as recommending a biopsy, or to continue to monitor the subject's condition.

[0102] In some embodiments, DNA from a population of several individuals can be analyzed by a set of multiplex arrays. The data of each multiplex array can be self-normalized using the information contained in that particular array. This normalization algorithm can be adjusted for the nominal intensity variations observed in the two color channels, background differences between channels, and possible crosstalk between dyes. The behavior of each base position can then be modeled using a clustering algorithm that incorporates some biological heuristics for mutation profile, mutation frequency, and / or fragmentation profile. In cases where few cfDNA fragments are observed (e.g., due to low minor allele frequency), the location and shape of missing sequences can be estimated using neural networks. Depending on the profile and percent sequence identity information, a statistical score can be devised (training score). Scores such as the GenCall score are designed to mimic the assessment made by the visual and cognitive systems of human experts. Moreover, it has been evolved using genotyping data from top and bottom strands. This score can be combined with some penalty terms (e.g., low intensity, mismatch between existing and predicted cfDNA fragments) to form a training score. The training score is saved for use by the calling algorithm.

[0103] To call the treatment response, the calling algorithm can obtain the genetic information and treatment response of multiple individuals with a disease or condition. First, the data can be normalized (using the same procedure as the clustering algorithm). The calling operation (classification) can be performed, for example, using a Bayesian model. The score of each call, the Call Score, can be the product of the training score and the data-to-model fit score. After scoring all the treatment responses, the application can calculate a composite score.

[0104] In some embodiments, the training dataset comprises clinical data selected from the group consisting of cancer stage, type of surgical procedure, age, tumor grading, depth of tumor invasion, occurrence of postoperative complications, and presence of venous invasion. In some embodiments, the training dataset is pre-processed, including converting the provided data into class-conditional probabilities.

[0105] In another embodiment, machine learning methods are used to train statistical classifiers, specifically support vector machines, for each cancer stage category based on word occurrences in the corpus of each patient's histology reports. New reports can then be classified according to their most likely stage, facilitating the collection and analysis of population staging data.

[0106] In some embodiments, the machine learning algorithm is selected from the group consisting of supervised or unsupervised learning algorithms selected from support vector machines, random forests, nearest neighbor analysis, linear regression, binary decision trees, discriminant analysis, logistic classifiers, and cluster analysis.

[0107] In general, the system may include a report generator for reporting on cancer test results and treatment options. The report generator system may be a central data processing system configured to establish communication with a remote data site or laboratory, a medical clinic / healthcare provider (treatment specialist) and / or a patient / subject directly through a communication link. The laboratory may be a medical laboratory, a diagnostic laboratory, a medical facility, a medical clinic, a point-of-care testing device, or any other remote data site capable of generating clinical information of a subject. The clinical information of a subject includes, but is not limited to, laboratory test data, x-ray data, examinations, and diagnoses. The medical provider or clinic 26 includes medical service providers, such as doctors, nurses, home health aides, technicians, and physician assistants, and the clinic is any medical facility where a medical provider is located. In certain cases, the medical provider / clinic is also a remote data site. In cancer treatment aspects, the subject may specifically be suffering from cancer.

[0108] Other clinical information for cancer subjects includes the results of laboratory tests, imaging, or medical treatments directed at the specific cancer, which can be easily identified by those skilled in the art.The list of suitable sources of clinical information for cancer includes, but is not limited to, CT scan, MRI scan, ultrasound scan, bone scan, PET scan, bone marrow examination, barium x-ray, endoscopy, lymphangiogram, IVU (intravenous urogram) or IVP (IV pyelogram), lumbar puncture, cystoscopy, immunological tests (antimalignin antibody screen), and cancer marker tests.

[0109] The subject's clinical information may be obtained manually or automatically from the laboratory. To simplify the system, the information is obtained automatically at a predetermined or fixed time interval. Fixed time interval refers to the time interval in which the collection of laboratory data is performed automatically by the method and system described herein based on a time measurement such as hours, days, weeks, months, years, etc. In one embodiment of the present invention, data collection and processing is performed at least once a day. In one embodiment, data transfer and collection is performed once a month, every two weeks, or once a week, or once every few days. Alternatively, information retrieval may be performed at a predetermined but not fixed time interval. For example, a first retrieval step may be performed after one week, and a second retrieval step may be performed after one month. Data transfer and collection can be customized according to the nature of the disorder being managed and the frequency of the subject's required tests and medical examinations.

[0110] In certain embodiments, genetic report is generated from the sample of the subject, for example, cfDNA.The polynucleotide in the sample can be sequenced, for example, whole genome sequencing, NGS sequencing, to generate multiple sequence reads.In some embodiments, genetic information comprises the variables that define the genomic composition of multiple cancer cells, or the genomic composition of one sporadic cancer cell.In some embodiments, genetic information comprises sequence or abundance data from one or more loci in cell-free DNA from an individual.

[0111] The cfDNA genetic information is processed (72). Genetic variants can also be identified. Genetic variants include sequence variants, copy number variants, and nucleotide modification variants. Sequence variants are those in which the gene nucleotide sequence is altered. Copy number variants are those in which the copy number of a portion of the genome deviates from the wild type. Genetic variants include, for example, single nucleotide variations (SNPs), insertions, deletions, inversions, transversions, translocations, gene fusions, chromosomal fusions, gene truncations, copy number variations (e.g., aneuploidy, partial aneuploidy, polyploidy, gene amplification), abnormal changes in nucleic acid chemical modifications, abnormal changes in epigenetic patterns, and abnormal changes in nucleic acid methylation. This process then determines the frequency of the genetic variant in the sample containing the genetic material. This process is noisy, so it separates the information from the noise (73). The sensitivity of detecting genetic variants can be increased by increasing the read depth of the polynucleotides (e.g., by sequencing to a greater read depth in samples derived from the subject at two or more time points).

[0112] Multiple measurements can be taken to increase diagnostic accuracy. Alternatively, measurements at multiple time points (e.g., 2, 3, 4, 5, 6, 7, 8, 9, 10, or more) can be used to determine whether the cancer is progressing, in remission, or stable. Diagnostic accuracy can be used to identify disease state. For example, cell-free polynucleotides taken from a subject can include polynucleotides derived from normal cells, as well as polynucleotides derived from diseased cells, such as cancer cells. Polynucleotides derived from cancer cells can have genetic variants, such as somatic mutations and copy number variants. When cell-free polynucleotides derived from a sample from a subject are sequenced, cfDNA mutation profiles, mutation frequencies, and / or fragmentation profiles can be generated as described in the Examples section below.

[0113] The methods and systems described herein can be used to detect a large number of cancers. Cancer cells, like most cells, can be characterized by a turnover rate, where old cells die and are replaced by new cells. Generally, dead cells are in contact with the vasculature in a given subject and may release DNA or DNA fragments into the bloodstream. This also applies to cancer cells during different stages of disease. Cancer cells may also be characterized by different genetic abnormalities, such as copy number variations and mutations, depending on the stage of disease. This phenomenon may be used to detect the presence or absence of cancer individuals using the methods and systems described herein.

[0114] In the early detection of cancer, any system or method described herein can be used to detect cancer, including mutation detection or copy number variation detection.These systems and methods can be used to detect any number of genetic abnormalities that can cause cancer or can be caused by cancer.These can include, but are not limited to, cfDNA mutation profile, mutation frequency, cfDNA fragmentation profile, mutation, mutation, indel, copy number variation, transversion, translocation, inversion, deletion, aneuploidy, partial aneuploidy, polyploidy, chromosomal instability, chromosomal structure change, gene fusion, chromosomal fusion, gene truncation, gene amplification, gene duplication, chromosomal damage, DNA damage, abnormal change in nucleic acid chemical modification, abnormal change in epigenetic pattern, abnormal change in nucleic acid methylation, infectious disease, and cancer.

[0115] Furthermore, the systems and methods described herein may also be used to help characterize a particular cancer. Genetic data generated from the systems and methods of the present disclosure may enable practitioners to better characterize a particular type of cancer. Often, cancers are heterogeneous in both composition and staging. Genetic profile data may enable characterization of a particular subtype of cancer, which may be important in the diagnosis or treatment of a particular subtype. This information may also provide clues to a subject or practitioner regarding the prognosis of a particular type of cancer.

[0116] The systems and methods provided herein may be used to monitor known cancer or other diseases in a particular subject. This may allow the subject or practitioner to adapt treatment options according to the progression of the disease. In this example, the systems and methods described herein may be used to build a genetic cfDNA mutation profile, mutation frequency, and / or fragmentation profile of a particular subject's disease progression. In some cases, cancer may progress, become aggressive, and become genetically unstable. In other examples, cancer may remain benign, inactive, or dormant. The systems and methods of the present disclosure may be useful in determining disease progression.

[0117] Furthermore, the systems and methods described herein may be useful in determining the efficacy of a particular treatment option.In one example, a particular treatment option may be correlated with the genetic cfDNA mutation profile, mutation frequency, and / or fragmentation profile of cancer over time.This correlation may be useful in selecting a therapy.Furthermore, if cancer is observed to be in remission after treatment, the systems and methods described herein may be useful in monitoring residual disease or disease recurrence.

[0118] Furthermore, the method of the present disclosure may be used to characterize the heterogeneity of an abnormal condition in a subject, the method includes creating a cfDNA mutation profile, mutation frequency, and / or fragmentation profile of extracellular polynucleotides in the subject, the cfDNA mutation profile includes a plurality of data resulting from profile variation and mutation analysis. In some cases, including but not limited to cancer, a disease may be heterogeneous. Disease cells may not be identical. In the example of cancer, it is known that some tumors contain different types of tumor cells, and some cells are at different stages of cancer. In other cases, the heterogeneity may constitute multiple disease foci. Again in the example of cancer, there may be multiple tumor foci, and perhaps one or more foci are the result of metastasis that has spread from the primary site (also known as distant metastasis).

[0119] The methods of the present disclosure may be used to generate a profile, fingerprint, or set of data that is the sum of genetic information from different cells within a heterogeneous disease. This data set may include copy number variation and mutation analysis, either alone or in combination.

[0120] Additionally, these reports are submitted and accessed electronically via the internet. Analysis of the data is performed at a location other than the subject's site. Reports are generated and transmitted to the subject's site. The subject accesses the reports reflecting the subject's tumor burden via an internet enabled computer.

[0121] The annotated information can be used by health care providers to select other drug treatment options and / or to provide information about drug treatment options to insurance companies. The method includes annotating drug treatment options for a condition, for example, in the NCCN Clinical Practice Guidelines in Oncology™ or the American Society of Clinical Oncology (ASCO) clinical practice guidelines.

[0122] Reports are generated that map the genomic location and cfDNA mutation profile variations of subjects with cancer. These reports may indicate that a particular cancer is aggressive and resistant to treatment when compared to other profiles of subjects with known outcomes. Subjects are monitored and retested over a period of time. If at the end of the period the cfDNA mutation profile, mutation frequency, and / or fragmentation variation profile have not changed, this may indicate that the current treatment is not working. Comparisons are made with the cfDNA mutation profiles of other subjects. For example, if it is determined that changes in cfDNA mutation variations indicate that the cancer is progressing, the initial treatment regimen prescribed is no longer treating the cancer and a new treatment is prescribed.

[0123] In certain embodiments, the system receives genetic information from a DNA sequencer. This process then determines the specific cfDNA alterations and their frequency. These reports are submitted and accessed electronically via the Internet. Data analysis is performed at a location other than the subject's location. Reports are generated and transmitted to the subject's location. The subject accesses the report reflecting the subject's tumor burden via an Internet-enabled computer.

[0124] Although temporal information can be used to enhance the information of cfDNA mutation profile and mutation frequency, other consensus methods can be applied. In other embodiments, historical comparison can be used with other consensus cfDNA mutation profiles, mutation frequencies, and / or fragmentation profiles. Consensus cfDNA mutation profile and mutation frequency can be normalized against control samples. Molecular mapping measurements against reference sequences can also be compared across the genome to identify regions in the genome where cfDNA mutation profile and mutation frequency have changed or remain the same. Consensus methods include linear or nonlinear methods (e.g., voting, averaging, statistical, maximum a posteriori, or maximum likelihood detection, dynamic programming, Bayesian, hidden Markov, or support vector machine methods, etc.) that generate consensus cfDNA mutation profile and mutation frequency, for example, derived from digital communication theory, information theory, or bioinformatics. After the sequence read coverage is determined, a probabilistic modeling algorithm is applied to convert the normalized nucleic acid sequence read coverage of each window region into a discrete copy number state. In some cases, the algorithm may include one or more of the following: hidden Markov model, dynamic programming, support vector machine, Bayesian network, trellis decoding, Viterbi decoding, expectation maximization, Kalman filtering, and neural network.

[0125] Artificial neural networks (NNets) mimic networks of "neurons" based on the neural structure of the brain. Artificial neural networks process records one at a time, or in batches, and "learn" the classification of the (initially almost arbitrary) record by comparing it to the known actual classification of the record. In MLP-NNets, the error from the initial classification of the first record is fed back into the network and used to improve the network's algorithm over the second, and so on, over and over again. Neural networks use an iterative learning process in which data examples (rows) are presented to the network one at a time, and the weights associated with the input values ​​are adjusted each time.

[0126] After all the examples have been presented, the process often begins again. During this learning phase, the network learns by adjusting the weights so that it can predict the correct class label of the input sample. Neural network learning is also called "connectionist learning" due to the connections between the units. The advantages of neural networks include a high tolerance to noisy data, as well as the ability to classify patterns that have never been trained on. One of the neural network algorithms is the back-propagation algorithm, e.g., Levenberg-Marquadt. Once a network has been constructed for a particular application, it is ready to be trained. To begin this process, the initial weights are randomly selected. Then, training, or learning, begins.

[0127] The network uses the weights and functions in the hidden layer to process the records in the training data one at a time, and then compares the resulting output against the desired output. The error is then back-propagated through the system, which causes the system to adjust the weights to apply to the next record to be processed. This process is carried out many times, with the weights being successively fine-tuned. During training of the network, the same data set is processed many times, with the connection weights being continually refined.

[0128] In one embodiment, the training step of the machine learning unit on the training dataset may generate one or more classification models for application to test samples. These classification models may be applied to the test samples to predict a subject's response to a therapeutic intervention.

[0129] Comparison of sequence coverage with control samples or reference sequences can aid in normalization across windows. In this embodiment, cell-free DNA is extracted and isolated from easily accessible body fluids such as blood. For example, cell-free DNA can be extracted using a variety of methods known in the art, including but not limited to isopropanol precipitation and / or silica-based purification. Cell-free DNA can be extracted from any number of subjects, for example, subjects without cancer, subjects at risk of cancer, or subjects known to have cancer (e.g., by other means).

[0130] After the isolation / extraction step, any of a number of different sequencing operations can be performed on the cell-free polynucleotide sample. The sample may be treated with one or more reagents (e.g., enzymes, unique identifiers (e.g., barcodes), probes, etc.) before sequencing. In some cases, once the sample has been treated with a unique identifier such as a barcode, the sample or fragments of the sample may be tagged with a unique identifier, either individually or in subgroups. The tagged sample may then be used in downstream applications such as sequencing reactions where individual molecules can be traced back to the parent molecule.

[0131] Cell-free polynucleotides can be tagged or tracked to allow for later identification and origin of specific polynucleotides. Assigning identifiers (e.g., barcodes) to individual polynucleotides or subgroups of polynucleotides may allow for unique identification information to be assigned to individual sequences or fragments of sequences. This may allow for data acquisition from individual samples, not limited to sample averages. In some cases, nucleic acids or other molecules derived from a single strand may share a common tag or identifier, and thus may later be traced back to originating from that strand. Similarly, all of the fragments derived from a single strand of nucleic acid may be tagged with the same identifier or tag, thereby later identifying the fragments derived from the parent strand. In other cases, gene expression products (e.g., mRNA) may be tagged to quantify expression, allowing barcodes, or barcodes in combination with sequences attached to barcodes, to be counted. In still other cases, the systems and methods can be used as PCR amplification controls. In such cases, multiple amplification products from a PCR reaction can be tagged with the same tag or identifier. If the products are later sequenced and show sequence differences, differences between products with the same identifier can be attributed to PCR errors. Additionally, individual sequences may be identified based on the sequence data characteristics of the read itself. For example, detection of unique sequence data at the beginning (start) and end (stop) of each sequencing read may be used alone or in combination with the length, i.e., number of base pairs, of each sequence read unique sequence to assign unique identity to each molecule. Fragments from a single strand of nucleic acid that have been assigned unique identity may then be used to identify fragments from the parent strand. This may be used in conjunction with bottlenecking of the initial starting genetic material to limit diversity.

[0132] Generally, the methods and systems provided herein are useful for preparing cell-free polynucleotide sequences for downstream sequencing reactions.In many cases, the sequencing method is next-generation sequencing (NGS), classical Sanger sequencing, whole genome bisulfite sequencing (WGSB), small RNA sequencing, low-coverage whole genome sequencing (IcWGS) and so on.

[0133] As used herein, the term "sequencing" refers to any of a number of techniques used to determine the sequence of a biomolecule, e.g., a nucleic acid such as DNA or RNA. Exemplary sequencing methods include targeted sequencing, single molecule real-time sequencing, exon sequencing, electron microscope-based sequencing, panel sequencing, transistor-mediated sequencing, direct sequencing, random shotgun sequencing, Sanger dideoxytermination sequencing, whole genome sequencing, sequencing by hybridization, pyrosequencing, capillary electrophoresis, gel electrophoresis, duplex sequencing, cycle sequencing, single-base extension sequencing, solid-phase sequencing, high-throughput sequencing, massively parallel signature sequencing, emulsion PCR, co-amplification at lower denaturation temperature-PCR (COLD-PCR), multiplex PCR, and the like. PCR, sequencing by reversible dye terminator, paired-end sequencing, near-term sequencing, exonuclease sequencing, sequencing by ligation, short-read sequencing, single molecule sequencing, sequencing-by-synthesis, real-time sequencing, reverse-terminator sequencingThe sequencing methods include, but are not limited to, 454 sequencing, nanopore sequencing, 454 sequencing, Solexa Genome Analyzer sequencing, SOLiD™ sequencing, MS-PET sequencing, and combinations thereof. In some embodiments, the sequencing can be performed by a genetic analyzer, such as a genetic analyzer commercially available from Illumina or Applied Biosystems. In some embodiments, the sequencing method can be massively parallel sequencing, which simultaneously (or in rapid succession) sequences at least any of 100, 1000, 10,000, 100,000, 1 million, 10 million, 100 million, or 1 billion polynucleotide molecules.

[0134] After sequencing, the reads are assigned a quality score. The quality score can be a representation of the reads that indicates whether the reads can be useful in later analysis based on a threshold. In some cases, some reads are not of sufficient quality or length to perform a later mapping step. Sequencing reads with a quality score of at least 90%, 95%, 99%, 99.9%, 99.99%, or 99.999% can be filtered from the data set. In other cases, sequencing reads that are assigned a quality score of at least 90%, 95%, 99%, 99.9%, 99.99%, or 99.999% can be filtered from the data set. Genome fragment reads that meet a specified quality score threshold are mapped to a reference genome or a reference sequence that is known not to contain mutations. After mapping alignment, the sequence reads are assigned a mapping score. The mapping score can be a representation or a read that is mapped back to a reference sequence that indicates whether each position is uniquely mappable. In some cases, the reads can be sequences that are not relevant for mutation analysis. For example, some sequence reads may be generated from contaminating polynucleotides.Sequencing reads with mapping scores of at least 90%, 95%, 99%, 99.9%, 99.99%, or 99.999% can be filtered from the data set.In other cases, sequencing reads assigned with mapping scores less than 90%, 95%, 99%, 99.9%, 99.99%, or 99.999% can be filtered from the data set.For each mappable base, bases that do not meet the minimum threshold of mappability, i.e., low-quality bases, can be replaced by the corresponding bases found in the reference sequence.

[0135] The methods and systems described herein can be used to detect a large number of cancers. Cancer cells, like most cells, can be characterized by a turnover rate, where old cells die and are replaced by new cells. Generally, dead cells are in contact with the vasculature in a given subject and may release DNA or DNA fragments into the bloodstream. This also applies to cancer cells in different stages of disease. Cancer cells may also be characterized by different genetic abnormalities, such as copy number variations and mutations, depending on the stage of disease. This phenomenon may be used to detect the presence or absence of cancer individuals using the methods and systems described herein.

[0136] The types and number of cancers that may be detected may include, but are not limited to, blood cancer, brain cancer, lung cancer, skin cancer, nasal cancer, pharyngeal cancer, liver cancer, bone cancer, lymphoma, pancreatic cancer, skin cancer, intestinal cancer, rectal cancer, thyroid cancer, bladder cancer, kidney cancer, oral cancer, stomach cancer, solid tumors, heterogeneous tumors, homogeneous tumors, and the like.

[0137] Furthermore, the systems and methods described herein may also be used to help characterize a particular cancer. Genetic data generated from the systems and methods of the present disclosure may enable practitioners to better characterize a particular type of cancer. Often, cancers are heterogeneous in both composition and staging. Genetic profile data may enable characterization of a particular subtype of cancer, which may be important in the diagnosis or treatment of a particular subtype. This information may also provide clues to a subject or practitioner regarding the prognosis of a particular type of cancer.

[0138] The systems and methods provided herein may be used to monitor known cancers or other diseases in a particular subject. This may allow the subject or practitioner to adapt treatment options as the disease progresses. In this example, the systems and methods described herein may be used to build a genetic profile of a particular subject's disease progression. In some cases, the cancer may progress, become aggressive, and become genetically unstable. In other cases, the cancer may remain benign, inactive, or dormant. The systems and methods of the present disclosure may be useful in determining disease progression.

[0139] Furthermore, the systems and methods described herein may be useful in determining the efficacy of a particular treatment option. In one example, successful treatment may actually increase the amount of copy number variation or mutation detected in the subject's blood, since more cancer may die and shed DNA if treatment is successful. In other examples, this may not happen. In another example, perhaps a particular treatment option may correlate with the genetic profile of the cancer over time. This correlation may be useful in selecting a therapy. Furthermore, if the cancer is observed to be in remission after treatment, the systems and methods described herein may be useful in monitoring residual disease or recurrence of the disease.

[0140] The data is transmitted to a computer for processing by a direct connection or over the Internet. The data processing aspects of the system may be implemented in digital electronic circuitry, or in computer hardware, firmware, software, or combinations thereof. The data processing apparatus of the invention may be implemented in a computer program product tangibly embodied in a machine-readable storage device for execution by a programmable processor, and the data processing method steps of the invention may be performed by the programmable processor executing a program of instructions to perform the functions of the invention by operating on input data and producing output. The data processing aspects of the invention may conveniently be implemented in one or more computer programs executable on a programmable system comprising at least one programmable processor coupled to receive data and instructions from a data storage system, at least one input device, and at least one output device, and to transmit data and instructions to the data storage system, at least one input device, and at least one output device. Each computer program may be implemented in a high-level procedural or object-oriented programming language, or in assembly or machine language if desired. In any case, the language may be a compiled or interpreted language. Suitable processors include, by way of example, general-purpose and special-purpose microprocessors. Typically, a processor receives instructions and data from a read-only memory and / or a random access memory. Suitable storage devices for tangibly embodying computer program instructions and data include, by way of example, all kinds of non-volatile memories, including semiconductor memory devices, such as EPROM, EEPROM, and flash memory devices; magnetic disks, such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM disks. Any of the foregoing may be supplemented by or incorporated in ASICs (application-specific integrated circuits).

[0141] To interact with a user, the method can be carried out using a computer system having a display device, such as a monitor or LCD (liquid crystal display) screen, for displaying information to the user, and an input device, such as a keyboard, a two-dimensional pointing device, such as a mouse or trackball, or a three-dimensional pointing device, such as a data glove or a gyroscopic mouse, that allows the user to input information into the computer system.The computer system can be programmed to provide a graphical user interface through which a computer program interacts with the user.The computer system can be programmed to provide a virtual reality three-dimensional display interface. EXAMPLES

[0142] Example 1: Single-molecule genome-wide mutation and fragmentation profiling of cell-free DNA for non-invasive detection of lung cancer We investigated whether identifying somatic sequence alterations across the genome could detect an increased number of tumor-derived cfDNA alterations and improve the ability to detect early stage disease. Tumor genomes contain thousands of somatic alterations. 19,20 Knowledge of such alterations, derived from tumor tissue, has been used to guide targeted analysis in the circulation during therapy. 21,22 In principle, such genome-wide alterations could be identified directly in cfDNA without knowledge of the alterations in the tumor, which could be useful for early cancer detection. However, such an approach would require the ability to efficiently detect somatic alterations and to distinguish these from the many other non-tumor-derived alterations.

[0143] To address these challenges, we developed herein an approach called Genome-wide Mutational Incidence for NonInvasive detection of cancer (GEMINI), which can identify a much larger number of tumor-derived alterations in cfDNA for cancer detection (Figure 1). We applied this method to analyze tissue and cfDNA samples from multiple patient cohorts (Figure 6). The method is based on sequencing individual cfDNA molecules to estimate genome-wide mutation frequencies and types of alterations using non-overlapping bins ranging in size from thousands to millions of bases. For each individual, the mutation types and frequencies in genomic regions more frequently altered in cancer are compared to a profile derived from regions more frequently mutated in normal cfDNA to determine multiregional differences in the mutation profile. Thus, the GEMINI approach enriches for mutations that are likely to be somatic while taking into account individual variability in total background alterations.

[0144] result To develop this method, we used the Pan-Cancer Analysis of Whole Genomes (PCAWG) study. 25,26We sequenced whole genomes of cancers from 2,511 individuals across 25 different cancers from the US, Canada, and Canada. This identified distinct genome-wide somatic mutation frequencies in different tumor types (Figure 7; Supplementary Table 1). As an example, analysis of lung tumor and matched normal tissue genomes from 65 smoking-exposed individuals revealed that cancers harbor an average of 52,209 (range 6,031–193,539) bona fide somatic mutations per genome. To quantify the number of tumor-specific alterations expected to be found in plasma in these individuals, we performed in silico dilution and downsampling experiments (Figure 8). In these simulations, we found that all patients theoretically harbored a subset of detectable somatic mutations when analyzed at 1x coverage using whole genome sequencing, even in tumors with a low proportion of somatic mutations (Figure 2A).

[0145] Because the majority of mutations detected at low genome-wide coverage are expected to be observed in single DNA molecules (Figure 2B), we developed a rigorous method to examine the frequency of single-molecule somatic mutations in a mixture of germline variants, WBC alterations, and experimental and sequencing artifacts (all while accounting for background variation). We scanned each sequenced molecule for single-nucleotide changes and calculated the frequency of putative mutations in high-quality reads after removing common germline variants and non-evaluable regions. The frequency of putative mutations was defined as the number of variants per million evaluated positions across all sequenced DNA molecules (Methods). A specific transversion likely associated with the accumulation of 7,8-dihydro-8-oxoguanine (8-oxo-dG) was identified. 27Because were more overrepresented among PCAWG single molecule alterations than other changes and more abundant than expected from analysis of similar transversions at sites of known polymorphisms, these alterations were filtered from further consideration when they occurred in certain read combinations (Figure 9A, 9B, Methods). These alterations were examined in a set of PCAWG lung tumor samples and matched normal blood cells (n=31) because blood cells are the largest source of cfDNA in non-cancer individuals. 28 Given its high abundance in tumors of current and former smokers, 29 We then focused our analysis on the remaining C:G>A:T mutations (hereafter C>A). As expected, given the high and variable frequency of all background alterations, we found that the C>A mutation frequency in tumors was similar compared to normal samples (Figure 2C) and only slightly higher even after the filtering steps described above and removal of germline variants. In this case, a small proportion of tumor alterations was somatic in origin (mean = 7.5%, range 0.8%-22%) (Figures 2D, 10A, 11).

[0146] We examined the high amount and variability of total background alterations between samples. These were found to be primarily sequencing lane- and run-specific artifacts (Figure 12). We hypothesized that taking into account the overall background rate in a sample-specific manner may improve the detection of tumor-derived alterations. Previous analyses have shown that mutation rates vary between individual cancer genomes, with lower mutation rates in regions associated with euchromatin, including expressed genes and early replicating regions, compared to heterochromatic regions corresponding to unexpressed genes and late replicating regions. 30,31 To examine variation in mutation frequency across the genome, 31 PCAWG paired samples were analyzed by binning sequence data containing 3,076,901 mutations into 1,144 non-overlapping 2.5 megabase (Mb) bins, and regions of increased mutation frequency across the genome were found to be shared by many tumors (Figures 13, 14).

[0147] To evaluate the GEMINI approach for tumor-derived DNA detection, we identified genomic regions with the largest C>A alterations in the cancer and control training sets and calculated the average C>A difference in these regions for patients not represented in the training set (Fig. 15, Methods). Regions enriched for C>A alterations were identified in 31 PCAWG cancers but not in normal samples (Fig. 2E), and background alterations were found to be highly correlated in cancer and control regions for each patient sample (Pearson correlation coefficient = 0.99, p < 0.0001) (Fig. 2F). This suggests that subtracting the alteration frequency between cancer and control regions within a given patient sample is useful for removing background mutations. In contrast, subtracting specific mutations observed in matched normal samples observed in single molecule sequencing data was not effective for removing background alterations (Fig. 16). This is because such alterations were typically de novo and observed once (Fig. 2G). After background subtraction, the remaining region mutation frequency was clearly higher in tumors compared to normal samples (median 13.4 compared to 1.3, p<0.0001, Wilcoxon rank-sum test), with a large proportion of alterations attributable to somatic mutations (mean=80%, range=31%-100%) (Figure 2H), which correlated highly with the frequency of high-confidence somatic C>A alterations reported in these samples by the PCAWG consortium (Pearson correlation coefficient=0.96, p<0.0001) (Figure 2I). The GEMINI approach using C>A region frequency was able to distinguish PCAWG cancer from non-cancer samples with high accuracy (AUC=0.91, 95% CI=0.84-0.99) compared to mutation frequency alone using low-coverage whole genome sequencing (AUC=0.64, 95% CI=0.50-0.79) (Figures 2J, 10A, 10B). The overall approach for filtering background variation using variant quality filters, removing germline variants, and subtracting regional mutation frequencies from single-molecule sequencing resulted in a 1,903-fold enrichment of somatic mutations in these samples.

[0148] To determine whether the GEMINI method could be used to non-invasively detect alterations in cfDNA derived from cancer patients, we applied this approach to a prospective lung cancer diagnostic cohort (LUCAS). 18 We assessed the ability to detect sequence alterations in individuals from the 1990s and 2000s. We analyzed low-coverage plasma whole-genome sequence data (approximately 2x coverage) from 365 individuals examined in this study. The majority of them were at high risk for lung cancer (age 50–80 years, smoking history ≥20 pack-years; Supplementary Table 2) and their blood samples were collected before clinical diagnosis. Given the short length of the cfDNA fragments, 13 In addition to the above filtering steps and removal of 8-oxo-dG-associated alterations, we used an additional filter that restricted the analysis to regions with identical sequence calls in overlapping reads in the paired-end libraries. Requiring a Phred quality score of ε30 in both reads would, in theory, reduce the benefit from sequencing errors and a high degree of overlap for shorter tumor-derived cfDNA sequences. 32 The error rate of mutations due to genomic DNA mutations will be smaller, thereby potentially enriching the detection of circulating tumor DNA (ctDNA) alterations. Supplementary Table 5 summarizes the cfDNA samples and genomic analysis of lung cancer patients receiving targeted therapy.

[0149] Comparison of single molecule sequencing of cfDNA from a subset of the LUCAS cohort and PCAWG tumor tissue samples revealed that genomic regions with increased frequency of cancer type-specific mutations were largely similar between tumor tissue and cfDNA from lung cancer patients, and between tissue and cfDNA from patients with melanoma and B-cell non-Hodgkin lymphoma (BNHL) (Pearson correlation >0.80, p<0.001 in all cases) and were located in genomic regions associated with tissue-specific slow replication timing (Figures 3A, 3B). Different mutation types among the tumors analyzed contributed to increased mutation frequency, including C>A changes in lung cancer, C>T in melanoma, and T>G in lymphoma. Tumor type- and mutation type-specific regional mutation frequencies were correlated with gene expression. 30 , genome compartmentalization measured by eigenvector analysis of methylation33 , as well as histone 3 lysine 9 trimethylation (H3K9me3), a known marker of heterochromatin. 34 We also found concordance between tumor and cfDNA analysis (Pearson correlation >0.80, p<0.001 in all cases) (Figures 17A-17C). Mutation types or regions not enriched in cancer-free individuals or cancers did not have these features or were weakly correlated with these features (Figures 3B, 17A-17C). Overall, these results suggest that genome-wide mutation rate variation in cfDNA is associated with chromatin organization and can be exploited by the GEMINI approach to detect tumor-derived sequence alterations in circulation.

[0150] Using the GEMINI approach, cross-validated regional differences in single molecule mutation frequencies were identified for individuals in the LUCAS cohort. Similar to the analysis in PCAWG lung cancer, regional C>A mutation frequencies in lung cancer individuals were preferentially altered compared to non-cancer individuals (p<0.0001, Wilcoxon rank sum test) (Figure 18) and were stable across sequencing lanes in contrast to the overall C>A frequency (Figure 19). The identified regions were largely consistent across cross-validation folds and contained high-quality sequences with similar evaluable bases, copy number levels, and mappability, but were located in genomic locations associated with distinct mutation frequencies between individuals with and without lung cancer, reflecting the epigenomic features described above (Figures 20A-20K, Supplementary Table 6). We further compared the identified regional differences in C>A mutation frequency to CC>AA double mutations, because these double mutations were enriched in lung cancers from smoking individuals. 26, given the requirement for two identical alterations at adjacent positions, the likelihood of this occurring by chance is extremely low (Figures 21A-21F). We found that the frequency of high-quality CC>AA alterations was highly correlated with regional differences in single-molecule C>A frequency in both tissues (Spearman's rho=0.62, p=0.0002) and cfDNA samples (Spearman's rho=0.65, p<0.0001) (Figures 21E, 21F). These data strongly support the idea that GEMINI mutation frequencies reflect tumor-derived sequence alterations in circulation.

[0151] Regional differences in single molecule C>A frequency were calibrated to the GEMINI score, which reflects an individual's probability of cancer (Methods). To assess whether clinical characteristics may influence genome-wide cfDNA mutation profiles, we investigated whether non-malignant nodules, sex, age, or the presence of chronic obstructive pulmonary disease (COPD) or autoimmune disease were associated with the GEMINI score. No differences in GEMINI scores were observed when comparing non-cancer individuals with benign lesions to those without benign lesions (median GEMINI score 0.30 vs. 0.33, p=0.94, Wilcoxon rank sum test) (Figure 4A), nor between men and women (p=0.14) (Figures 22A-22F). No correlation of GEMINI score with age (rho=-0.15, p=0.053), no correlation with levels of the inflammatory marker CRP (rho=0.01, p=0.89), and no correlation with levels of IL-6 (rho=-0.07, p=0.40) was observed (Figures 22A-22F). Similarly, no change in GEMINI score was observed between individuals with or without COPD (p=0.73) or with or without autoimmune disease (p=0.31) (Figures 22A-22F). Taken together, these analyses show that single molecule mutation frequency in cfDNA is not significantly influenced by demographic characteristics or the presence of acute or chronic inflammatory conditions.

[0152] We next assessed the association between GEMINI scores and cancer stage and histology. While non-cancer individuals had lower GEMINI scores (median scores of 0.30 for individuals with benign lesions and 0.33 for those without benign lesions), cancer patients had significantly higher median scores across stages (stage I=0.74, stage II=0.67, stage III=0.76, and stage IV=0.74) (p<0.001 for stages I, II, III, or IV, Wilcoxon rank-sum test) (Figure 4a) and histological subtypes (adenocarcinoma=0.71, squamous cell carcinoma=0.72, small cell lung cancer (SCLC)=0.98) (p<0.0001 for all subtypes, Wilcoxon rank-sum test) (Figure 4a). As expected, GEMINI scores were largely associated with ctDNA levels, with ichorCNA 35 The high GEMINI scores in SCLC patients likely reflect the known large ctDNA fraction in this tumor type. 36 , Patients with low GEMINI scores (e.g., <0.5) were more likely to have NSCLC histology, Supplementary Table 7 (p=0.03, Fisher's exact test) and small ichorCNA tumor fraction estimates (p=0.003, Wilcoxon rank sum test). Receiver operating characteristic (ROC) curves, representing the sensitivity and specificity of the GEMINI approach to identify cancer patients, revealed an overall area under the curve (AUC) of 0.85 (95% CI=0.79-0.91) (Figure 4D) and a high level of detection across stages and subtypes (Figures 4F, 24A, 24B).

[0153] Using the fixed GEMINI model, we evaluated samples from seven patients who were cancer-free at the time of blood draw and later diagnosed with lung cancer. The median GEMINI score for these individuals was 0.78, significantly higher than the median GEMINI score for non-cancer individuals (p=0.0005, Wilcoxon rank-sum test) (Figure 4B). Six of these seven individuals had scores above the threshold at 80% specificity, with time to lung cancer diagnosis ranging from 231 to 1868 days. This provides evidence that abnormalities in cfDNA mutational profiles are useful for detecting cancer several years earlier than standard diagnosis. Five of these patients were diagnosed with NSCLC (two patients had stage I disease, one patient had stage III disease, and stage information was not available for the other two patients), one was diagnosed with SCLC (stage unknown), and the other patient, for whom we had no stage or histology information, died within a few months of diagnosis. The patient not detected by GEMINI had the longest time from blood draw to diagnosis (1,954 days). Interestingly, at the time of initial blood draw, cancer was not suspected based on CT imaging in four of these patients, and no biopsies were performed. Cancer was suspected based on CT imaging in the remaining three patients, and the patients underwent biopsies; however, their pathology reports showed benign pulmonary nodules, highlighting the limitations of current diagnostic approaches.

[0154] We next investigated whether the genome-wide fragmentation features used by the DELFI approach could be combined with the GEMINI mutation profile, as we hypothesized that these methods measure complementary cfDNA features and may be used to improve the ability to detect individuals with early-stage lung cancer. We combined the GEMINI and DELFI scores into a combined score to assess the predictive accuracy of these features used in isolation (Methods). Although the GEMINI and DELFI scores were positively correlated (Spearman's rho=0.50, p<0.0001), several samples missed by either approach in isolation were detected using the combined approach, resulting in a reduction in false negatives, e.g., 56% at a specificity of 80% (Figure 25). The combined approach increased overall performance with an overall AUC of 0.93 (95% CI=0.89-0.97) (p<0.05 when compared to GEMINI alone or DELFI alone) (Figure 4D). For stage I patients (n=13), DELFI fragmentation analysis alone or GEMINI analysis alone yielded an AUC of 0.73 (95%CI=0.59-0.88) and 0.80 (95%CI=0.67-0.93), respectively, whereas the combined approach yielded an AUC of 0.87 (95%CI=0.76-0.98) (p<0.05 when compared with DELFI alone or GEMINI alone) (Fig. 4f). The performance of the combined GEMINI and fragmentation approach yielded an overall sensitivity of 91% with a specificity of 80% (GEMINI / DELFI score>0.38) (Table 1). If this approach is considered as a prescreen for LDCT, the sensitivity of the combined approach with LDCT would be >95% with a combined specificity of 85% (Table 1). Importantly, individuals with low GEMINI / DELFI scores had a better prognosis compared to those with high scores (p=0.004, log-rank test) (FIG. 26). Therefore, this approach reduces the concern of false negatives, as individuals with low scores have a better prognosis and may be detectable in subsequent screens.

[0155] To externally validate the individual GEMINI methods as well as the combined GEMINI / DELFI approach, an additional cohort of individuals from a lung cancer screening program was evaluated (n=57, Supplementary Table 3). This cohort included asymptomatic high-risk individuals with primarily early stage cancer (stage I=32, II=4, III=3, IV=2, and unknown=1) whose samples were collected before clinical diagnosis, as well as individuals who were ultimately confirmed to be cancer-free (n=15). Twenty-one of the 42 lung cancer individuals (50%) were diagnosed with stage IA disease, a proportion similar to that detected by LDCT in the National Lung Screening Trial. 5 cfDNA was isolated from the plasma of these individuals and subjected to low-coverage whole genome sequencing using similar coverage and feature metrics as the LUCAS cohort (Figures 27A-27D). These samples were analyzed using the fixed GEMINI and fragmented machine learning models from the LUCAS cohort analysis. Consistent with the initial study, we observed that high-risk individuals with cancer (age 50-80 years and smoking history) had higher GEMINI scores compared to individuals without cancer (p=0.001, Wilcoxon rank-sum test) (Figure 4C). Across the validation and LUCAS cohorts, the GEMINI scores of patients with late-stage lung cancer (stage III / IV, median GEMINI score=0.74) were significantly higher than those with early-stage cancer (stage I / II, median GEMINI score=0.64) (p=0.03, Wilcoxon rank-sum test). The performance of the GEMINI approach to detect stage I disease among individuals in this cohort was high, with an overall AUC of 0.81 (95% CI = 0.67-0.94) and 0.86 (95% CI = 0.74-0.97) when combined with fragmentation features (Figure 4G). Overall, these analyses suggest that genome-wide mutational profiling is broadly generalizable for detecting early-stage lung cancer in high-risk populations.

[0156] Because somatic alterations in lung cancer are associated with smoking, we hypothesized that there would be an association between cfDNA mutational profiles and smoking history. Although total cfDNA C>A mutation frequencies were similar between nonsmokers with and without lung cancer in the LUCAS cohort (p=0.65, Wilcoxon rank-sum test), total mutation frequencies were higher in smokers with lung cancer than smokers without cancer (p=0.01, Wilcoxon rank-sum test) and GEMINI scores were dramatically increased (p<0.0001, Wilcoxon rank-sum test) (Figures 23A, 23B). GEMINI scores were positively associated with years of smoking among cancer patients (rho=0.24, p=0.01). Interestingly, in cancer-free individuals, GEMINI scores were negatively correlated with smoking exposure (rho=-0.25, p=0.002), potentially reflecting smoking-related DNA damage in noncancerous tissues that may contribute to cfDNA alterations. 37 Analysis of patients in the LUCAS cohort and validation cohort suggested that the GEMINI approach may have high performance in detecting individuals with a long smoking history, including an increase in GEMINI performance in the LUCAS cohort with an AUC of 0.90 and 0.95 for the combined GEMINI / DELFI approach (p<0.05 compared with GEMINI alone with AUC 0.90 or DELFI alone with AUC 0.88, DeLong's test) (Figures 4E, 4H, 4I, 28A-28C). A positive GEMINI test at 80% specificity was associated with a 13.5-fold increase in cancer odds among ≥20 pack-year smokers (95% CI for odds ratio: 6.7-30.7, p<0.0001) and a 20.1-fold increase among ≥40 pack-year smokers (95% CI for odds ratio: 7.7-54.6, p<0.0001). These observations are consistent with the idea that smoking exposure causes sequence alterations in ctDNA and nontumor cfDNA affecting distinct genomic regions that may help improve cancer detection using the GEMINI approach.

[0157] Given the important differences between biological features and clinical management of SCLC and non-small cell lung cancer (NSCLC), we investigated whether genome-wide mutation profiling could be used to detect SCLC and non-invasively distinguish this cancer from other cancer types. GEMINI scores in SCLC patients (n=13) were significantly higher than in non-cancer individuals (n=88) (p<0.0001, Wilcoxon rank sum test) (Figure 5A, Supplementary Tables 2, 3), and we were able to distinguish between them with an AUC>0.99 (95% CI=0.99-1.00) (Figure 5C). The GEMINI approach was used to evaluate regional mutation differences in cfDNA of SCLC patients compared to NSCLC patients (n=99) and found that the mutation frequency obtained with this method was higher in SCLC (p<0.0001, Wilcoxon rank sum test) (Figure 5B, Supplementary Table 4) and could be used to distinguish this cancer type from NSCLC (AUC=0.86, 95 CI=0.75-0.96) (Figure 5C). These findings suggest that genome-wide mutation profiles may be useful to provide a non-invasive approach for detecting SCLC and for distinguishing between lung cancer and different histological subtypes.

[0158] To explore the generalizability of the GEMINI approach to detect other cancers, the method was applied to evaluate a prospective cohort (n=62) of individuals with or without liver cancer. Cross-validated regional differences in mutation frequency identified significant differences in genome-wide T>C mutation profiles in individuals with liver cancer (Figure 5D). The derived GEMINI scores were higher across all stages (0-A, B, and C) in individuals with liver cancer compared to individuals with cirrhosis (p<0.01 for each comparison) (Figure 5E). Similar to the analysis of lung cancer patients, GEMINI scores derived from patients with liver cancer were largely associated with ctDNA levels, and ichorCNA 35 The tumor fraction estimated by α-glucose administration increased with increasing tumor fraction (p=0.008, Wilcoxon rank sum test) (Figure 23B, Supplementary Table 8).

[0159] Because cfDNA mutational profiles appeared to be cancer type specific, we hypothesized that the GEMINI approach may be useful for distinguishing between different cancer types. Using GEMINI, we compared differences in mutational profiles in cfDNA between NSCLC, SCLC, and HCC (n=159). Profiles were found to cluster primarily into three groups, with each cancer type constituting the majority of observations in a cluster (Fig. 5F) (Methods). Exclusion of the most common tumor-specific alterations (Figs. 3A-3B, 7) prevented accurate grouping by cancer type (Fig. 29). Overall, these analyses suggested that mutational profiles may be a useful method to non-invasively determine the origin of cancer.

[0160] To explore whether the GEMINI approach can be used to monitor patients during therapy, we evaluated serial blood samples from lung cancer patients undergoing treatment with EGFR or ERBB2 inhibitors with mutant allele fractions (MAFs) as low as 0.1%. Using a fixed model trained on the high-risk LUCAS cohort, we found that all patients had a decrease in GEMINI score after starting therapy, consistent with an early response to therapy, and an increase in GEMINI score over time, consistent with the known progression of these individuals (Figure 30). Comparing GEMINI scores to mutant allele fractions from targeted sequencing of these patients revealed a significant correlation of ctDNA levels between the two methods (Spearman's correlation coefficient = 0.50, p = 0.03). This shows that GEMINI has high sensitivity to low MAF levels, reflecting the amount of ctDNA during therapy.

[0161] This study shows that single molecule mutation profiles derived from low-coverage whole genome sequencing of cfDNA can non-invasively detect individuals with cancer. Tumor-type-specific alterations in the mutational landscape are detectable in the plasma of cancer patients and appear to correlate with genome replication timing and other chromatin features that may impair DNA damage repair. 38 The method described herein includes deep sequencing of matched blood cells to filter for hematopoietic changes. 16 We also performed tumor sequencing to identify tumor-specific mutations that we aimed to evaluate in plasma. 22 This approach is therefore applicable to de novo detection and characterization of cancer. The combination of genome-wide sequencing and fragmentation analysis of cfDNA offers the opportunity for cost-effective and scalable cancer detection.

[0162] Although the majority of patients in the described cohort were individuals at risk for developing cancer, large-scale validation in lung, liver, and other cancer screening populations will be required before clinical use. However, extending the read length from 100 bp to 150 bp would approximately quadruple the evaluable bases sequenced by both reads. A variety of genome-wide tumor-specific mutation profiles were evaluated, including different lung cancer histologies, liver cancer, melanoma, and lymphoma, but further analysis of genome-wide mutation profiles using other sequence alterations in other settings will likely be useful. As mutation rates vary widely across the cancer genome, 31 , Detection of changes in regional mutation frequency in cfDNA offers a generalizable approach that may be useful in early cancer detection and monitoring.

[0163] Example 2: Methods and Materials Study population analyzed Tissue samples from the PCAWG consortium consisted of 2,778 tumors with somatic mutation calls 39Hypermutated tumors, including those with putative polymerase ε or mismatch repair defects, as well as tumors treated with temozolomide were excluded from the analysis (n=49), and cancer types with fewer than 20 samples (n=129 samples) and those with an average of <250 mutations per sample (pilocytic astrocytoma, n=89 samples) were also excluded, resulting in 2,511 tumors across 25 common cancer types. Single molecule mutation analysis consisted of lung cancers and corresponding solid tissues or blood cells from 86 donors that passed quality control metrics. 39 The cohort consisted of 30 women and 56 men, ages 41–83 years, diagnosed with lung cancer. Among these individuals, 38 had lung adenocarcinoma and 48 had lung squamous cell carcinoma, and of these, 65 had mutations attributable to smoking-associated signature 4. Thirty-one of these 65 patients had available tumor tissue and blood-derived normal sequencing data. Additional information regarding these samples is available at dcc.icgc.org / releases / PCAWG. See also Supplementary Table 1.

[0164] As previously stated 18The LUCAS cohort was a prospectively recruited group of 365 patients with positive imaging findings by chest X-ray or chest CT at the Department of Respiratory Medicine, Infiltrate Unite, Bispebjerg Hospital, Copenhagen. Patients diagnosed with cancer with known active disease or who were undergoing treatment at the time of enrollment were excluded. The study took place over a 7-month period from September 2012 to March 2013. All patients were clinically followed up until death or until April 2020. All patients provided written informed consent and the study was conducted in accordance with the Declaration of Helsinki. The LUCAS study was approved by the Danish Regional Ethics Committee and the Danish Data Protection Agency. Blood samples were collected from all patients at the first visit before a possible diagnosis of lung cancer. The cohort analyzed included 158 patients with no prior, baseline, or future cancer, 114 patients with baseline lung cancer, 15 patients with lung metastases, and 78 patients without lung cancer at blood draw but with early or later lung cancer or another cancer type. The high-risk LUCAS cohort was defined as individuals at high risk for lung cancer (age 50-80 years, ≥20 pack-year smoking history) and included individuals with primary lung cancer at baseline (n=89) and individuals with no prior, baseline, or future cancer (n=74) (Supplementary Table 2). There was a median of approximately 1,451 (range: 392-2,111) haploid genome equivalents per ml of plasma analyzed.

[0165] The validation cohort consisted of individuals (n = 57) from a lung cancer screening program, including asymptomatic high-risk individuals with nodules confirmed to be early stage cancer or benign (Supplementary Table 3). Individuals were recruited from the Detection of Early Lung Cancer Among Military Personnel (DECAMP) Consortium. 40 Individuals were enrolled through the Allegheny Health Network (AHN) or through a screening program at the Allegheny Health Network (AHN). The DECAMP-1 protocol included current or former smokers with ≥220 pack-years of exposure and radiological findings showing an indeterminate lung nodule measuring 0.7–3.0 cm in size identified within 12 months prior to enrollment and an additional CT scan within 3 months prior to enrollment. Individuals enrolled at AHN were identified based on eligibility for high-risk lung cancer screening using low-dose helical CT scanning or based on indication for lung cancer screening based on other high-risk features such as a family history of lung cancer. All patients provided written informed consent to participate in these collections, and the study was conducted in accordance with the Declaration of Helsinki. Liquid biopsies of all individuals were collected prior to a possible diagnosis of lung cancer.

[0166] The lung cancer monitoring cohort consisted of serial blood draws from a cohort of lung cancer patients undergoing treatment with EGFR or ERBB2 inhibitors 11 The study population included samples (n = 18) derived from serial blood draws from patients with a smoking history (n = 5) for whom both targeted and whole-genome sequencing was available. 13 Patients were aged 50-73 years and had stage II-IV lung adenocarcinoma (n=4) or mixed histology (n=1).

[0167] The liver cancer cohort consisted of 62 patients with liver cancer (n = 48) or cirrhosis (n = 14). Samples were collected prospectively as part of the HCC biomarker registry at the Johns Hopkins University School of Medicine under a protocol approved by the Johns Hopkins Institutional Review Board. Liver cancer was revealed by appropriate imaging features as defined by approved guidelines. Tumor staging was determined by the Barcelona Clinic Liver Cancer staging system (BCLC). Detailed clinical data were extracted from electronic medical records (Supplementary Table 8).

[0168] Blood sample collection and storage Sample collection for the LUCAS cohort took place at the screening visit and was performed as follows: Venous peripheral blood was collected into one K2-EDTA tube. The tubes were centrifuged at 2330g for 10 minutes at 4°C within 2 hours of blood collection. EDTA plasma was aliquoted after centrifugation and stored at -80°C.

[0169] For the validation cohort, venous peripheral blood from each individual was collected into one K2-EDTA tube (AHN) or one Streck tube (DECAMP). Tubes from the AHN and DECAMP collections were centrifuged at low speed (800–1600 g) for 10 min. The plasma portion from the first spin was spun a second time for 10 min. Plasma was aliquoted after centrifugation and stored at -80 °C for cfDNA analysis.

[0170] For the lung cancer monitoring cohort, whole blood was collected into EDTA tubes and processed immediately or within 1 day after storage at 4°C, or collected into Streck tubes and processed as previously described. 13and processed within 2 days of collection. Plasma and cellular components were separated by centrifugation at 800 g for 10 min at 4°C. Plasma was spun a second time at 18,000 g at room temperature to remove remaining cellular debris and stored at -80°C until time of DNA extraction.

[0171] For the liver cancer cohort, sample collection was performed as follows: venous peripheral blood was collected into one K2-EDTA tube. Within 2 hours of blood collection, the tube was centrifuged at 2330g for 10 minutes at 4°C, the plasma was transferred to a new tube, and the sample was spun at 14,000 rpm (18,000 rcf) for 10 minutes at room temperature to pellet remaining cellular debris. EDTA plasma was aliquoted after centrifugation and stored at -80°C for cfDNA analysis.

[0172] Plasma sequencing library preparation For all plasma samples, circulating cell-free DNA was isolated from 2–4 ml of plasma using the Qiagen QIAamp Circulating Nucleic Acids Kit (Qiagen GmbH), eluted in 52 μl of RNase-free water containing 0.04% sodium azide (Qiagen GmbH), and stored in LoBind tubes (Eppendorf AG) at −20 °C. cfDNA concentration and quality were assessed using a Bioanalyzer 2100 (Agilent Technologies).

[0173] Next-generation sequencing (NGS) cfDNA libraries from the LUCAS, validation, and liver cancer cohorts for whole genome sequencing were prepared using 15 ng of cfDNA when available, or the full purified volume when less than 15 ng was available. Briefly, genomic libraries were prepared using the NEBNext DNA Library Prep Kit from Illumina (New England Biolabs (NEB)) with four major modifications to the manufacturer's guidelines: (i) the library purification step used an on-bead AMPure XP (Beckman Coulter) approach to minimize sample loss during the elution and tube transfer steps; (ii) the volumes of NEBNext End Repair, A-tailing, and adapter ligation enzymes and buffers were adjusted accordingly to accommodate on-bead AMPure XP purification; (iii) the ligation reaction used an Illumina dual index adaptor; and (iv) the cfDNA libraries were amplified by Phusion Hot Start Polymerase. All of these samples underwent four-cycle PCR amplification after the DNA ligation step. For the lung cancer monitoring cohort, WGS and targeted sequencing were performed as previously described. 11,13 Next-generation sequencing (NGS) cfDNA libraries were prepared using 5–250 ng of cfDNA.

[0174] Whole genome sequencing data from PCAWG samples Somatic mutation calling, tumor purity, coverage statistics, and SigProfiler 26Mutation signature abundances generated by were downloaded from the International Cancer Genome Consortium (ICGC) Data Portal (https: / / dcc.icgc.org / releases / PCAWG). Bam files and germline variant calls were downloaded from the Bionimbus Protected Data Cloud (bionimbus.opensciencedatacloud.org). SAM tool 41 The Bam files were indexed using

[0175] Downsampling and dilution of somatic mutations derived from PCAWG lung cancer samples The downsampling and dilution experimental method is shown in Figure 8. Specifically, we obtained somatic mutation calls (n = 3,393,564 mutations) for individuals (n = 65) with lung cancer and PCAWG with the presence of signature 4. 26 Mutations with missing values ​​for the number of observed reference or mutant alleles were excluded (n=5,857), yielding 3,387,707 somatic mutations among 65 individuals. For a given individual, each observation of the reference or mutant allele was considered separately. The number of sequenced observations originating from the tumor was calculated as the total number of observations multiplied by the tumor purity of the sample. We then multiplied 10 of the observations by 100%. -1 , 10 -2 , 10 -3 , or 10 -4 We spiked in observations with the reference allele until the mutations were of tumor origin. We then calculated the average coverage of the mutation positions after dilution and randomized the observations to achieve the desired coverage of 8x, 4x, 2x, 1x, and 0.5x. For each known somatic mutation in an individual's cancer genome, we tallied the number of times we observed the mutation for each combination of dilution amount and genome coverage, and used this information to calculate the percentage of mutations observed in a single DNA molecule.

[0176] Whole genome sequencing of plasma samples Libraries prepared from whole genomes of cancer patients and cancer-free individuals were analyzed using Illumina HiSeq 2000 / 2500 (LUCAS 18 , validation, and lung cancer monitoring cohort 13 ) and NovaSeq 6000 (liver cancer cohort) using a 100 bp paired-end run (200 cycles) to approximately 2x coverage per sample. To assess concordance between tissue and cfDNA mutation profiles in cancer types with only a small number of samples available, LUCAS samples from melanoma (n=2) and lymphoma patients (n=1) as well as 40 non-cancer controls and 15 individuals with highly advanced lung cancer were resequenced to a median of 10x coverage by Illumina NovaSeq 6000. Prior to alignment, fastp 42 Adapter sequences were filtered from the reads using Bowtie2. 43 Sequence reads were aligned against the hg19 human reference genome using Sambamba 44 Duplicate reads were removed using . Sequencing data from each sample contained >75 million fragments, >15 million reads, >10 million reads mapped to the reference genome, >85% of bases had a Phred quality score of ε20 (Q20), and >80% of bases had a Phred quality score of ε30 (Q30).

[0177] Identifying single and double base changes in a single molecule Primary alignments of correctly paired read pairs mapped to autosomes in non-overlapping 100 kb bins were scanned to obtain base calls, Phred scores, and mapping quality (MAPQ) for each sequenced base using pysam. Only read pairs with a MAPQ of at least 40 and only one position in each read with a Phred score of at least 30 were considered. To avoid counting multi-nucleotide variants in the analysis of single base changes, positions where two adjacent positions both contain the reference allele and have a Phred score of at least 30 were filtered. In the analysis of double base changes, a similar filter was used to avoid counting even larger multi-base variants. Additionally, positions that overlapped with the Duke Excluded Regions track (hgdownload.cse.ucsc.edu / goldenpath / hg19 / encodeDCC / wgEncodeMapability) were removed. In each 100 kb bin, the number of sequenced bases that were C:G or A:T in the reference genome were counted. We also counted the number of times each type of single base change (C:G>A:T, C:G>G:C, C:G>T:A, T:A>A:T, T:A>C:G, and T:A>G:C) and CC:GG>AA:TT double base change were observed in 100 kb bins. Observations were counted separately based on whether the purine or pyrimidine of each base pair was in read 1 or read 2 of the paired-end sequencing data. To exclude potential germline variants, we used the gnomAD database (version 3.0), which contains gene variants from >70,000 whole genomes. 45We downloaded the gnomAD version 3.0 variant call format (VCF) files available in hg38 coordinates from the gnomAD browser. We first lifted the position of each sequence variation identified across hg19-hg38 using the R package rtracklayer. We removed sequence variations that did not lift over to hg38, that lifted over to hg38 but to multiple different positions, or that lifted over to hg38 but where the reference genome sequence differed between the hg19 and hg38 genome builds. We annotated the identified sequence variations along with the population allele frequency and whether the variant passed the gnomAD quality filter. We then removed any candidate variants if they were present in gnomAD but did not pass the gnomAD quality filter or if they were present in gnomAD with an allele frequency >1 / 100,000. For PCAWG samples, the remaining variants were annotated in each sample. This indicates that they were called as somatic or germline variants by the PCAWG consortium. For tissue sample analysis, if any position in a fragment was sequenced by both read pairs, the position was randomly moved away from read 1 or read 2. For plasma samples, positions in fragments sequenced by both read 1 and read 2 of a read pair with the same base call were analyzed. To filter 8-oxo-dG-related sequence changes from single base analysis, we excluded any base where guanine or G>T was in read 1 and cytosine or C>A was in read 2. To filter artifactual CC>AA changes, we excluded any base where CC or CC>AA was in read 1 and GG or GG>TT was in read 2.To account for potential differences in sequencing depth between samples, single molecule mutation frequencies were always calculated as the number of each sequence variation divided by the number of evaluable bases, defined as the number of positions in the fragment where each sequence variation could be detected after quality and germline filtering.

[0178] Estimation of 8-oxo-dG levels For each sample, 8-oxo-dG levels were estimated as the ratio of single molecule C>A frequencies when guanine or G>T was on read 1 and cytosine or C>A was on read 2 to when cytosine or C>A was on read 1 and guanine or G>T was on read 2.

[0179] Creating regional differences in single molecule mutation frequency The approach to calculating the regional difference in single molecule mutation frequency for a particular mutation type is shown in Figure 15. Specifically, 100 kb bins were first collected into 1144 non-overlapping 2.5 Mb bins. Let TIFF2025503433000002.tif5128 represent the number of sequence alterations (e.g., C>A) in bin i for non-cancer and cancer participants, respectively. We compute the corresponding number of evaluable positions (e.g., the number of C:G bases that pass the quality filter) as For a training set consisting of n-1 samples from J cancer participants and K non-cancer participants (J+K=n-1), the difference in the number of sequence changes in bin i relative to the number of evaluable bases, comparing cancer participants with non-cancer participants, is Represented by TIFF2025503433000004.tif11128.

[0180] δ (s) is δ (1) is the most depleted bin for sequence changes in cancer versus non-cancer, and δ (1144) is the order statistic s such that s is the bin most enriched for sequence changes in cancer versus non-cancerth The feature selection in the training set is done by selecting the bin (δ (1) ,...,δ (114) bins in the top decile (δ (1144) ,...,δ (1030) We proceed by identifying the bin with h th For the training set excluding the samples in A, the bin sets of the top decile and the bottom decile are denoted by {A -h} and {B -h The regional difference in single molecule mutation frequency of the test sample is expressed as: Represented by TIFF2025503433000005.tif10128.

[0181] This procedure was repeated using leave-one-out cross-validation such that every participant appeared once in the test set, obtaining regional differences in single molecule mutation frequencies for all n participants.

[0182] Downsampling of regional differences in single molecule C>A frequency to 1x coverage in PCAWG For simplicity, the area difference Use alternate symbols for TIFF2025503433000006.tif7128, where y Bh =Σ b∈{B-h} y bh It was downsampled ( * ) Area difference These quantities are first calculated using the hg19 reference genome, r A and r B We then define the set {1,...,x Ah} to r A Let the exponent be the set {1,...,x Bh ] to r B The index was randomly selected (without replacement). Ah and y BhThe number of exponents in two random samples that is less than or equal to y, respectively. * Ah and y * Bh The above procedure is repeated until all participants in the PCAWG have downsampled regional differences in single molecule C>A frequency.

[0183] Correlation between single molecule mutation frequency and tissue-specific genomic features Replication timing tracks created by the UW ENCODE group, calculated by averaging the wavelet-smooth transform of 6-fraction profiles representing different time points during replication in 1 kb bins, were downloaded from the UCSC Genome Browser for IMR90, NHEK, and GM12878 cell lines. Weighted averages were calculated in each 2.5 Mb bin; larger values ​​indicate earlier replication timing. Gene expression values ​​were calculated from lung adenocarcinoma (n=542), lung squamous cell carcinoma (n=504), melanoma (n=472), and N-cell non-Hodgkin lymphoma (n=48) generated by The Cancer Genome Atlas (TCGA) using recount3 46 Expression estimates were obtained as raw counts using qRT-PCR and converted to transcripts per million (TPM). For each cancer type, the TPM values ​​for each gene were first averaged across samples. In each cancer type, gene expression in each 2.5 Mb bin was calculated as the sum of the TPMs overlapping with each bin, weighted by the transcript length. These values ​​were then averaged between lung adenocarcinoma and lung squamous cell carcinoma to obtain single lung cancer gene expression estimates in each bin. A / B compartmentalization data generated by eigenvector analysis of 450K methylation array data at 100 kb resolution was obtained for 12 cancer types and by eigenvector analysis of Hi-C data for GM12878 cells. 33For each 2.5 Mb bin, a weighted average of the eigenvectors in 100 kb bins was calculated. The average of these values ​​from lung adenocarcinoma and lung squamous cell carcinoma was used for lung cancer analysis. GM12878 was used for BNHL analysis, and the average across all 12 cancer types was used for melanoma analysis in the absence of skin A / B compartmentalization data. H3K9me3 ChIP-seq data for A549 cells (3 pooled replicates), GM23248 cells, and Karpas422 cells (2 pooled replicates), expressed as fold change in coverage in enriched samples relative to control samples, were downloaded from the ENCODE portal (accessions: ENCFF425LVX, ENCFF098PML, and ENCFF574RYG). A weighted average of fold change was calculated in each 2.5 Mb bin for each cell type. GC content in each 2.5 Mb bin was obtained from the hg19 reference genome. Mappability scores, which reflect how uniquely 100-mer sequences align with genomic regions, were downloaded (hgdownload.cse.ucsc.edu / goldenpath / hg19 / encodeDCC / wgEncodeMapability / wgEncodeCrgMapabilityAlign100mer.bigWig) and assembled into 2.5 Mb bins as a weighted average of the mappability scores that overlap with each bin. Genome-wide copy number was estimated for each sample using ichorCNA. The average copy number per genomic bin was calculated as the weighted average of the copy numbers in the segments that overlap with each bin.

[0184] Creating a GEMINI score To provide a calibrated score capturing the association between regional differences in single molecule C>A frequency and the probability that an individual has lung cancer in the high-risk LUCAS cohort, we fitted a logistic regression model for cancer status with regional differences in single molecule C>A frequency as a covariate (Lung GEMINI model) and extracted a fitted probability of cancer for each individual (Lung GEMINI score). A Lung GEMINI score >0.55 reflects a positive test for lung cancer detection with a specificity of 80%. Furthermore, for the validation cohort, the cohort of patients who had a baseline negative test and later developed lung cancer, the cohort of lung cancer patients monitored during therapy, and the remaining samples in the LUCAS cohort, we created Lung GEMINI scores using fixed bins and the Lung GEMINI model. For the liver cancer cohort, we created a GEMINI score by fitting a logistic regression model for cancer status (Liver GEMINI model) with regional differences in unimolecular T>C frequency as a covariate and extracting the fitted probability of cancer (Liver GEMINI score) for each individual. A Liver GEMINI score of >0.86 reflects a positive test for liver cancer detection with a specificity of 80% (see Supplementary Tables 1-8).

[0185] Development of DELFI and combined GEMINI / DELFI scores To evaluate whether fragmentation features could further improve GEMINI cancer status prediction, we used a previously described approach on the same training set used to create the cross-validated GEMINI score. 18 Briefly, the hg19 reference genome was tiled into non-overlapping 5 Mb bins. Bins with an average GC content <0.3 and average mappability <0.9 were excluded, leaving 473 bins spanning approximately 2.4 GB of genome. Fragment size analysis was performed on fragments with a MAPQ of at least 30. The ratio of the number of short (100–150 bp) to long (151–220 bp) fragments across the 473 bins was calculated as previously described. 18, normalized for GC content and library size. For each training set, a principal component analysis was performed on the fragmentation profile, and the minimum number of principal components required to explain 90% of the variance between participants was retained. Chromosome arm copy numbers were summarized by calculating z-scores for each arm using expected coverage and standard deviations calculated from an external reference set of 54 non-cancer controls (github.com / cancer-genomics / PlasmaToolsHiseq.hg19). The 39 z-scores and principal components were incorporated as covariates in the logistic regression model and LASSO penalty. To generate the DELFI score in the validation cohort, we used a previously described model trained on 158 non-cancer and 129 cancer. 18 A combined GEMINI / DELFI score was calculated by averaging the individual GEMINI and DELFI scores for each patient.

[0186] Association between GEMINI score and tumor DNA percentage in plasma ichorCNA for samples from the LUCAS and liver cancer cohorts 35 was used to estimate the percentage of tumor DNA in plasma.

[0187] Generate regional differences in single molecule C>A frequency between SCLC and NSCLC Regional differences in single molecule C>A frequency were calculated as previously described. Here, individuals with SCLC were compared to individuals with NSCLC. To maximize the number of samples used to identify binsets A and B, we combined samples from the high-risk LUCAS cohort (n=10 SCLC, n=75 NSCLC) with individuals aged 50-80 who were smokers from the validation cohort (n=3 SCLC, n=24 NSCLC).

[0188] Analysis of different tumor types Regional differences in single molecule mutation frequencies were calculated by repeatedly holding out each individual (n=159) with either NSCLC, SCLC, or HCC and identifying bin sets A and B with all other individuals as described previously. For each mutation type (C>A, C>G, C>T, T>A, T>C, and T>G), individuals with NSCLC were compared to individuals with SCLC, individuals with NSCLC were compared to individuals with HCC, and individuals with SCLC were compared to individuals with HCC to obtain 18 regional differences in mutation frequency per individual. Principal coordinate analysis was performed on a similarity matrix created from Euclidean distances between pairwise samples using these 18 regional differences in mutation frequency. K-means clustering was performed on the matrix of 18 regional differences in mutation frequency with the number of clusters (k) set to 3. As a negative control, we also performed principal coordinate analysis on a similarity matrix constructed from Euclidean distances between pairwise samples after excluding the C>A and T>C mutations most frequently observed in lung and liver cancers, yielding 12 regional differences in mutation frequency per individual.

[0189] statistics Wilcoxon rank sum test was used to generate p values ​​for two-group comparisons. Correlations of continuous variables were performed using Pearson's product moment correlation coefficient or Spearman's rank correlation coefficient. All p values ​​were based on two-sided hypothesis testing. ROC curves were compared using DeLong's test. All confidence intervals for the areas under the ROC curves were presented at 95% confidence level and were based on DeLong's method. Confidence intervals for coefficients in logistic regression models were normalized and presented at 95% confidence level. Analysis of variance (ANOVA) was performed and F-test was used to evaluate whether the variation between sequencing lanes in C>A frequency or region C>A frequency was statistically significant. The analysis was performed using Rε 3.6.1 and Python 3.8.2. All box plots represent interquartile ranges with whiskers drawn to the maximum value within the upper and lower bounds (upper bound = 0.75 quantile + 1.5 x interquartile range; lower bound = 0.25 quantile - 1.5 x interquartile range). The solid middle line in the box plots corresponds to the median.

[0190] Data availability The computer code, software versions, and computational environment to reproduce the results from this study are provided as a GitHub repository (github.com / cancer-genomics / gemini_wflow). Sequence data and clinical variables from the LUCAS study are available from the European Genome-Phenome Archive (EGA) under the accession code EGAS00001005340.

[0191] (Table 1) Sensitivity of GEMINI and DELFI followed by LDCT for detecting lung cancer * TIFF2025503433000008.tif66166 *Sensitivity was determined at a specificity of 80% for GEMINI or GEMINI and DELFI, or at a combined specificity of 85% and GEMINI and DELFI prescreen specificity of 62% for GEMINI, DELFI, and LDCT. The overall sensitivity of DELFI at a specificity of 80% was 80%. Based on the high-risk subset of patients analyzed from the LUCAS cohort 13 The specificity of LDCT in this setting was 66%, with an estimated sensitivity of 95%. The overall blended sensitivity reflects the expected sensitivity in the screened population weighted by the proportion of lung cancers detected in the NLST population at each stage. 5 .

[0192] Other Aspects From the foregoing description, it is apparent that changes and modifications may be made to the disclosure described herein to adapt it to various uses and conditions, such embodiments also falling within the scope of the following claims.

[0193] All references to sequences, patents, and publications in this specification are hereby incorporated by reference to the same extent as if each individual patent and publication was specifically and individually indicated to be incorporated by reference.

[0194] References TIFF2025503433000009.tif52165TIFF2025503433000010.tif216165TIFF2025503433000011.tif216165TIFF2025503433000012.tif92165

[0195] Other Aspects From the foregoing description, it is apparent that changes and modifications may be made to the invention described herein to adapt it to various uses and conditions, such embodiments also falling within the scope of the following claims.

[0196] All references to sequences, patents, and publications in this specification are hereby incorporated by reference to the same extent as if each individual patent and publication was specifically and individually indicated to be incorporated by reference.

[0197] Supplementary Table 1. Summary of PCAWG cohort and genomic analyses * TIFF2025503433000013.tif230166

[0198] Supplementary Table 1. Summary of PCAWG cohort and genomic analyses * TIFF2025503433000014.tif235168

[0199] Supplementary Table 1. Summary of PCAWG cohort and genomic analyses * TIFF2025503433000015.tif238167

[0200] Supplementary Table 1. Summary of PCAWG cohort and genomic analyses * TIFF2025503433000016.tif22584

[0201] Supplementary Table 2. Effect of variant filters on tissue and cfDNA samples * TIFF2025503433000017.tif23281 * Sequence data from tissue samples were randomly downsampled to approximately 2x coverage similar to that of cfDNA samples, and values ​​reflect averages across samples. † All mutation frequencies in cfDNA samples were scaled to have a mean value of 0 for non-cancer samples.

[0202] Supplementary Table 3. Summary of LUCAS cohort and genomic analyses * TIFF2025503433000018.tif228159TIFF2025503433000019.tif223170TIFF2025503433000020.tif235168TIFF2025503433000021.t if233170TIFF2025503433000022.tif237170TIFF2025503433000023.tif237164TIFF2025503433000024.tif237169TIFF2025503433 000025.tif238166TIFF2025503433000026.tif238165TIFF2025503433000027.tif238170TIFF2025503433000028.tif236163TIFF20 25503433000029.tif237164TIFF2025503433000030.tif237168TIFF2025503433000031.tif236166TIFF2025503433000032.tif23049

[0203] Supplementary Table 3. Summary of LUCAS cohort and genomic analyses * (continuation) TIFF2025503433000033.tif237170TIFF2025503433000034.tif237170TIFF2025503433000035.tif238164TIFF20255034330 00036.tif234170TIFF2025503433000037.tif237169TIFF2025503433000038.tif233170TIFF20255503433000039.tif236169

[0204] Supplementary Table 4. Overview of high-coverage cfDNA samples and genomic analysis TIFF2025503433000040.tif236160

[0205] Supplementary Table 4. Overview of high-coverage cfDNA samples and genomic analysis TIFF2025503433000041.tif235118

[0206] Supplementary Table 5. Overview of cfDNA samples and genomic analysis of lung cancer patients receiving targeted therapy * TIFF2025503433000042.tif236117 * Clinical, coverage, and mutant allele fraction information for patients receiving targeted therapy was provided by Phallen et al. 11 and Cristiano et al. 13 Obtained from.

[0207] Supplementary Table 5. Overview of cfDNA samples and genomic analysis of lung cancer patients receiving targeted therapy * continuation TIFF2025503433000043.tif235135

[0208] Supplementary Table 6. Summary of validation cohorts and genomic analyses TIFF2025503433000044.tif237159

[0209] Supplementary Table 6. Summary of validation cohorts and genomic analyses TIFF2025503433000045.tif23785

[0210] Supplementary Table 7. Summary of SCLC vs. NSCLC analysis TIFF2025503433000046.tif117170

[0211] Supplementary Table 7. Summary of SCLC vs. NSCLC analysis TIFF2025503433000047.tif117170

[0212] Supplementary Table 7. Summary of SCLC vs. NSCLC analysis TIFF2025503433000048.tif117170

[0213] Supplementary Table 7. Summary of SCLC vs. NSCLC analysis TIFF2025503433000049.tif64170

[0214] Supplementary Table 8. Overview of liver cohort and genomic analyses TIFF2025503433000050.tif95170

[0215] Supplementary Table 8. Overview of liver cohort and genomic analyses TIFF2025503433000051.tif32170

Claims

1. 1. A method for determining the frequency of somatic mutations in a subject, comprising: Extracting cell-free DNA (cfDNA) from a biological sample of a subject; preparing a genomic library from the extracted cfDNA; sequencing each cfDNA molecule to obtain a mutation profile; Determining multi-region differences in the mutation profile; and determining the frequency of somatic mutations in said subject. The method comprising:

2. 2. The method of claim 1, wherein said determining a genome-wide mutation and fragmentation profile comprises identifying mutations and fragment length changes in the sequence of individual cfDNA molecules.

3. The method of claim 1 or 2, wherein the mutation profile comprises mutation frequency and mutation type across the subject's genome.

4. 4. The method of claim 3, wherein the genome-wide mutation profile of the subject is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 20 million bases.

5. 4. The method of claim 3, wherein the genome-wide mutation profile of the subject is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 10 million bases.

6. 4. The method of claim 3, wherein the genome-wide mutation profile of the subject is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 5 million bases.

7. The method of claim 3, wherein mutations in each sequenced molecule are determined after removal of common germline variants and non-evaluable regions.

8. The method of any one of claims 1 to 7, wherein the frequency and mutation type of single molecule somatic mutations across the genome of the subject is diagnostic of cancer compared to the frequency and mutation type of single molecule somatic mutations across the genome of a normal subject.

9. The method of any one of claims 1 to 8, wherein such analysis is performed in subjects for which tumor tissue is not available.

10. 1. A method of treating cancer in a subject, comprising: Extracting cell-free DNA (cfDNA) from a biological sample of a subject; preparing a genomic library from the extracted cfDNA; sequencing each cfDNA molecule to obtain a mutation profile; determining multi-region differences in the mutation profile and determining the frequency of somatic mutations in the subject; and administering a cancer treatment to the subject based on these. The method comprising:

11. 11. The method of claim 10, wherein the cancer treatment comprises surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy, targeted therapy, and combinations thereof.

12. 12. The method of claim 10 or 11, wherein the cancer comprises colorectal cancer, lung cancer, breast cancer, gastric cancer, pancreatic cancer, bile duct cancer, brain cancer, or ovarian cancer.

13. 13. The method of claim 12, wherein the lung cancer is small cell lung cancer (SCLC).

14. 13. The method of claim 12, wherein the lung cancer is non-small cell lung cancer (NSCLC).

15. The method of any one of claims 10 to 14, wherein the subject having cancer comprises an altered chromatin organization-associated mutation profile compared to a healthy individual.

16. The method of any one of claims 10-15, wherein the genome-wide mutation and fragmentation profile comprises identifying mutations and fragment length changes in the sequence of individual cfDNA molecules.

17. The method of claim 10, wherein the mutation profile comprises mutation frequency and mutation type across the subject's genome.

18. 18. The method of claim 17, wherein the genome-wide mutation profile of the subject is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 20 million bases.

19. 18. The method of claim 17, wherein the genome-wide mutation profile of the subject is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 10 million bases.

20. 18. The method of claim 17, wherein the genome-wide mutation profile of the subject is determined using non-overlapping bins spanning a size range of at least about 1000 bases to at least about 5 million bases.

21. The method of claim 16, wherein genome-wide mutations of each sequenced molecule are determined after removal of common germline variants and non-evaluable regions.

22. 1. A method for determining regional frequencies of mutations across a genome, comprising: Sequencing each cfDNA molecule isolated from the subject; estimating mutation frequencies and mutation types across the genome; determining the mutation types and frequencies in genomic regions that are altered in the cancer relative to the mutation profile and regions that are mutated in normal cfDNA to determine multi-regional differences in the mutation profile; Thereby determining the regional frequency of mutations across the genome. The method comprising:

23. 23. The method of claim 22, wherein said estimation of mutation frequencies and mutation types across the genome comprises using non-overlapping bins ranging in size from thousands to millions of bases.

24. The method of claim 22 or 23, wherein the tumor-specific changes are quantified by one or more assays.

25. 25. The method of claim 24, wherein the one or more assays comprise an in silico dilution assay and / or a downsampling assay.

26. The method of any one of claims 22 to 25, wherein after removal of common germline variants and / or non-evaluable regions, each sequenced molecule is scanned for single nucleotide changes.

27. The method of any one of claims 22 to 26, wherein the genomic region is characterized by slow replication timing, low gene expression, B compartmentalization, high H3K9me3 abundance, low GC content, or a combination thereof.

28. 27. The method of any one of claims 21 to 26, wherein the predicted mutation frequency is defined as the number of variants per million evaluated positions across all sequenced DNA molecules.

29. The method of any one of claims 21 to 28, further comprising combining the mutation profile and the genome-wide fragmentation profile.

30. running a machine learning model to determine changes in the genome-wide mutation profile that classifies or excludes the subject as having or at risk of having cancer based on the genome-wide mutation profile identified for the subject.

30. The method of any one of claims 21 to 29, further comprising:

31. A method for determining whether a subject is a responder to a treatment based on the results of performing the method according to any one of claims 1 to 30 or a combination thereof.

32. 32. The method of claim 31, wherein the treatment is selected from surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy, targeted therapy, and combinations thereof.

Images