Cell-free DNA for assessing and / or treating cancer

JP2024069295A5Pending Publication Date: 2024-11-13JOHNS HOPKINS UNIVERSITY
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
JP2024032720
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Priority Date
2019-01-23
Filing Date
2024-03-05
Publication Date
2024-11-13

AI Technical Summary

Technical Problem

Current methods for diagnosing and treating cancer lack effective, widely available biomarkers for early detection and treatment, leading to delayed diagnosis and reduced treatment efficacy.

Method used

The use of cell-free DNA (cfDNA) fragmentation profiles, analyzed through low-coverage whole-genome sequencing and machine learning, to identify cancer presence and monitor treatment response, utilizing methods like DELFI for assessing and treating cancer.

Benefits of technology

DELFI provides non-invasive early detection and localization of cancer with high sensitivity (57% to >99%) and specificity (98%), enabling effective treatment and monitoring, even with limited DNA samples.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000035_0000
    Figure 00000035_0000
  • Figure 00000036_0000
    Figure 00000036_0000
  • Figure 00000037_0000
    Figure 00000037_0000
Patent Text Reader

Abstract

To provide methods and materials for identifying a mammal as having cancer (e.g., localized cancer), and also to provide methods and materials for assessing, monitoring, and / or treating a mammal having cancer.SOLUTION: Provided is a method of determining a cell free DNA (cfDNA) fragmentation profile of a mammal, the method comprising: processing cfDNA fragments obtained from a sample obtained from the mammal into sequencing libraries; subjecting the sequencing libraries to low-coverage whole genome sequencing to obtain sequenced fragments; mapping the sequenced fragments to a genome to obtain windows of mapped sequences; and analyzing the windows of mapped sequences to determine cfDNA fragment lengths.SELECTED DRAWING: None
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims the benefit of U.S. Application No. 62 / 673,516, filed May 18, 2018, and claims the benefit of U.S. Application No. 62 / 795,900, filed January 23, 2019. The disclosures of the prior applications are considered part of (and incorporated by reference into) the disclosure of this application.

[0002] Federal Funding Statement This invention was made with United States Government support under Grant No. CA121113 from the National Institutes of Health. The United States Government has certain rights in this invention. [Background technology]

[0003] 1.Technical Field This document relates to methods and materials for assessing and / or treating a mammal (e.g., a human) with cancer. For example, this document provides methods and materials for identifying a mammal as having cancer (e.g., localized cancer). For example, this document provides methods and materials for monitoring and / or treating a mammal with cancer.

[0004] 2.Background circumstances Much of the morbidity and mortality from human cancer worldwide is the result of delayed diagnosis of these diseases, where treatment is less effective (Torre et al., 2015 CA Cancer J Clin 65:87; and World Health Organization, 2017 Guide to Cancer Early Diagnosis). Unfortunately, clinically proven biomarkers that can be used to universally diagnose and treat patients are not widely available (Mazzucchelli, 2000 Advances in clinical pathology 4:111; Ruibal Morell, 1992 The International journal of biological markers 7:160; Galli et al., 2013 Clinical chemistry and laboratory medicine 51:1369; Sikaris, 2011 Heart, lung & circulation 20:634; Lin et al., 2016 in Screening for Colorectal Cancer: A Systematic Review for the US Preventive Services Task Force. (Rockville, MD); Wanebo et al., 1978 N Engl J Med 299:448; and Zauber, 2015 Dig Dis Sci 60:681). [Prior art documents] [Non-patent literature]

[0005] [Non-Patent Document 1] Torre et al. 2015 CA Cancer J Clin 65:87 [Non-Patent Document 2] World Health Organization,2017 Guide to Cancer Early Diagnosis [Non-Patent Document 3] Mazzucchelli,2000 Advances in clinical pathology 4:111 [Non-Patent Document 4] Ruibal Morell,1992 The International journal of biological markers 7:160 [Non-Patent Document 5] Galli et al. 2013 Clinical chemistry and laboratory medicine 51:1369 [Non-Patent Document 6] Sikaris,2011 Heart, lung & circulation 20:634 [Non-Patent Document 7] Lin et al., 2016 in Screening for Colorectal Cancer: A Systematic Review for the US Preventive Services Task Force. (Rockville, MD) [Non-Patent Document 8] Wanebo et al., 1978 N Engl J Med 299:448 [Non-Patent Document 9] Zauber,2015 Dig Dis Sci 60:681 Summary of the Invention

[0006] Recent analyses of cell-free DNA suggest that such approaches may provide new avenues for early diagnosis (Phallen et al., 2017 Sci Transl Med 9; Cohen et al., 2018 Science 359:926; Alix-Panabieres et al., 2016 Cancer discovery 6:479; Siravegna et al., 2017 Nature reviews. Clinical oncology 14:531; Haber et al., 2014 Cancer discovery 4:650; Husain et al., 2017 JAMA 318:1272; and Wan et al., 2017 Nat Rev Cancer 17:223).

[0007] This document provides methods and materials for determining a cell-free DNA (cfDNA) fragmentation profile in a mammal (e.g., in a sample obtained from a mammal). In some cases, 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 a 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 assessed to determine a cfDNA fragmentation profile. This document also provides methods and materials for assessing and / or treating a mammal (e.g., a human) having or suspected of having cancer. In some cases, 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 assessed to determine whether the mammal has cancer based at least in part on a cfDNA fragmentation profile. In some cases, this document provides methods and materials for monitoring and / or treating a mammal having cancer. For example, one or more cancer treatments can be administered to a mammal identified as having cancer (e.g., based at least in part on a cfDNA fragmentation profile) to treat the mammal.

[0008] A non-invasive method for early detection and localization of cancer is described herein. Blood cfDNA can provide a non-invasive diagnostic route for patients with cancer. As demonstrated herein, DNA Evaluation of Fragments for Early Interception (DELFI) was developed and used to evaluate genome-wide fragmentation patterns of cfDNA in 236 patients with breast, colorectal, lung, ovarian, pancreatic, gastric or bile duct cancer, as well as 245 healthy individuals. These analyses revealed that cfDNA profiles of healthy individuals reflect the nucleosomal fragmentation patterns of white blood cells, whereas patients with cancer have altered fragmentation profiles. DELFI had a sensitivity of detection ranging from 57% to over 99% among seven cancer types, with a specificity of 98%, and in 75% of cases, localized the primary tissue of the cancer to a limited number of sites. Assessing cfDNA (e.g., using DELFI) can provide a screening approach for early detection of cancer, which can increase the chances of successful treatment of patients with cancer. Assessing cfDNA (e.g., using DELFI) can also provide an approach for monitoring cancer, which can increase the chances of successful treatment and improved outcomes of patients with cancer. Furthermore, cfDNA fragmentation profiles can be obtained from limited amounts of cfDNA using inexpensive reagents and / or equipment.

[0009] In general, one aspect of this document features a method for determining a cfDNA fragmentation profile of a mammal. The method can include, or consist essentially of, processing cfDNA fragments obtained from a sample obtained from the mammal into a sequencing library, subjecting the sequencing library to whole genome sequencing (e.g., low-coverage whole genome sequencing) to obtain sequenced fragments, mapping the sequenced fragments to a genome to obtain windows of mapped sequences, and analyzing the windows of mapped sequences to determine cfDNA fragment lengths. The mapped sequences can include tens to thousands of windows. The windows of mapped sequences can be non-overlapping windows. Each of the windows of mapped sequences can include about 5 million base pairs. A cfDNA fragmentation profile can be determined within each window. The cfDNA fragmentation profile can include a median fragment size. The cfDNA fragmentation profile can include a fragment size distribution. The cfDNA fragmentation profile can include a ratio of small to large cfDNA fragments in a window of mapped sequences. The cfDNA fragmentation profile can span the entire genome. The cfDNA fragmentation profile can span a subgenomic interval (e.g., an interval in a portion of a chromosome).

[0010] In another embodiment, this document features a method for identifying a mammal as having cancer. The method can include, or consist essentially of, determining a cfDNA fragmentation profile in a sample obtained from the mammal, comparing the cfDNA fragmentation profile to a reference cfDNA fragmentation profile, and identifying the mammal as having cancer if the cfDNA fragmentation profile in the sample obtained from the mammal differs from the reference cfDNA fragmentation profile. The reference cfDNA fragmentation profile can be a cfDNA fragmentation profile of a healthy mammal. The reference cfDNA fragmentation profile can be generated by determining a cfDNA fragmentation profile in a sample obtained from a healthy mammal. The reference DNA fragmentation pattern can be a reference nucleosomal cfDNA fragmentation profile. The cfDNA fragmentation profile can include a fragment size median, and the fragment size median of the cfDNA fragmentation profile can be shorter than the fragment size median of the reference cfDNA fragmentation profile. The cfDNA fragmentation profile can include a fragment size distribution, and the fragment size distribution of the cfDNA fragmentation profile can differ by at least 10 nucleotides compared to the fragment size distribution of the reference cfDNA fragmentation profile. The cfDNA fragmentation profile can include position-dependent differences in fragmentation patterns, including ratios of small to large cfDNA fragments, where small cfDNA fragments can be 100 base pairs (bp) to 150 bp in length and large cfDNA fragments can be 151 bp to 220 bp in length, and the correlation of fragment ratios in the cfDNA fragmentation profile can be lower than the correlation of fragment ratios in a reference cfDNA fragmentation profile. The cfDNA fragmentation profile can include sequence coverage of small cfDNA fragments, large cfDNA fragments, or both small and large cfDNA fragments across the genome. The cancer can be colorectal cancer, lung cancer, breast cancer, bile duct cancer, pancreatic cancer, gastric cancer, or ovarian cancer. The comparing step can include comparing the cfDNA fragmentation profile to a reference cfDNA fragmentation profile in a window across the entire genome. The comparing step can include comparing the cfDNA fragmentation profile to a reference cfDNA fragmentation profile across a subgenomic interval (e.g., an interval within a portion of a chromosome). The mammal can have previously undergone a cancer treatment to treat the cancer. The cancer treatment can be surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T-cell therapy, targeted therapy, or any combination thereof. The method can also include administering a cancer treatment (surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T-cell therapy, targeted therapy, or any combination thereof) to the mammal. The mammal can be monitored for the presence of cancer after administration of the cancer treatment.

[0011] In another embodiment, this document features a method for treating a mammal with cancer. The method can include or consist essentially of identifying a mammal as having cancer, wherein the identification includes determining a cfDNA fragmentation profile in a sample obtained from the mammal, comparing the cfDNA fragmentation profile with a reference cfDNA fragmentation profile, and identifying the mammal as having cancer if the cfDNA fragmentation profile obtained from the mammal is different from the reference cfDNA fragmentation profile; and administering a cancer treatment to the mammal. The mammal can be a human. The cancer can be colorectal cancer, lung cancer, breast cancer, gastric cancer, pancreatic cancer, bile duct cancer, or ovarian cancer. The cancer treatment can be surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy, targeted therapy, or a combination thereof. The reference cfDNA fragmentation profile can be a cfDNA fragmentation profile of a healthy mammal. The reference cfDNA fragmentation profile can be generated by determining a cfDNA fragmentation profile in a sample obtained from a healthy mammal. The reference DNA fragmentation pattern can be a reference nucleosomal cfDNA fragmentation profile. The cfDNA fragmentation profile can include a median fragment size, where the median fragment size of the cfDNA fragmentation profile is shorter than the median fragment size of the reference cfDNA fragmentation profile. The cfDNA fragmentation profile can include a fragment size distribution, where the fragment size distribution of the cfDNA fragmentation profile differs by at least 10 nucleotides compared to the fragment size distribution of the reference cfDNA fragmentation profile. The cfDNA fragmentation profile can include a ratio of small to large cfDNA fragments in a window of mapped sequences, where the small cfDNA fragments are 100bp to 150bp in length and the large cfDNA fragments are 151bp to 220bp in length, and the correlation of the fragment ratios in the cfDNA fragmentation profile is lower than the correlation of the fragment ratios of the reference cfDNA fragmentation profile.The cfDNA fragmentation profile can include sequence coverage of small cfDNA fragments in a window spanning the genome. The cfDNA fragmentation profile can include sequence coverage of large cfDNA fragments in a window spanning the genome. The cfDNA fragmentation profile can include sequence coverage of small and large cfDNA fragments in a window spanning the genome. The comparing step can include comparing the cfDNA fragmentation profile with a reference cfDNA fragmentation profile across the entire genome. The comparing step can include comparing the cfDNA fragmentation profile with a reference cfDNA fragmentation profile across a subgenomic interval. The mammal can have previously undergone a cancer treatment to treat the cancer. The cancer treatment can be surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy, targeted therapy, or a combination thereof. The method can also include monitoring the mammal for the presence of cancer after the administration of the cancer treatment.

[0012] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art. Similar or equivalent methods and materials to those described herein can be used to carry out the present invention, but the following describes suitable methods and materials. All publications, patent applications, patents and other references cited herein are incorporated by reference in their entirety. In case of conflict, the present specification, including definitions, will prevail. In addition, the materials, methods and examples are illustrative only and are not intended to be limiting.

[0013] The details of one or more embodiments of the invention are set forth in the accompanying drawings and the description below. Other features, objects, and advantages of the invention will become apparent from the description and drawings, and from the claims. [Brief description of the drawings]

[0014] [Figure 1] Schematic of an exemplary DELFI approach. Blood is collected from a cohort of healthy individuals and patients with cancer. Nucleosome-protected cfDNA is extracted from the plasma fraction, processed into a sequencing library, interrogated through whole-genome sequencing, mapped to the genome, and analyzed to determine cfDNA fragment profiles in different windows across the genome. A machine learning approach is used to classify individuals as healthy or with cancer, and to identify tumor tissue of origin using whole-genome cfDNA fragmentation patterns. [Diagram 2] Simulation of non-invasive cancer detection based on the number of variations analyzed and tumor-derived cfDNA fragment distribution. Monte Carlo simulations were performed using different numbers of tumor-specific variations to evaluate the probability of detecting cancer variations in cfDNA at the indicated proportion of tumor-derived molecules. Simulations were performed assuming an average of 2000 genome equivalents of cfDNA and the need for 5 or more observations of any variation. These analyses show that increasing the number of tumor-specific variations improves the sensitivity of detection of circulating tumor DNA. [Diagram 3] Distribution of tumor-derived cfDNA fragments. Cumulative density functions of cfDNA fragment lengths for 42 loci containing tumor-specific variations from 30 patients with breast, colorectal, lung, or ovarian cancer are shown with 95% confidence bands (blue). Mutant cfDNA fragment lengths were significantly different in size compared to wild-type cfDNA fragments (red) at these loci. [Figure 4] Figure 4A and B: GC content and fragment length of tumor-derived cfDNA. A, GC content was similar for mutated and nonmutated fragments. B, GC content did not correlate with fragment length. [Diagram 5] Germline cfDNA fragment distribution. Cumulative density functions of fragment lengths for 44 loci containing germline variations (non-tumor derived) from 38 patients with breast, colorectal, lung, or ovarian cancer are shown with 95% confidence bands. Fragment lengths with germline mutations (blue) were comparable to wild-type cfDNA fragment lengths (red). [Figure 6] Distribution of hematopoietic cfDNA fragments. Cumulative density functions of fragment lengths for 41 loci containing hematopoietic variations (non-tumor derived) from 28 patients with breast, colorectal, lung, or ovarian cancer are shown with 95% confidence bands. After correction for multiple testing, there were no significant differences in size distributions of mutated hematopoietic cfDNA fragments (blue) and wild-type cfDNA fragments (red) (α=0.05). [Figure 7] Figure 7A-F: cfDNA fragmentation profiles in healthy individuals and patients with cancer. A, Whole genome cfDNA fragmentation profiles (defined as the ratio of short to long fragments) obtained from approximately 9x whole genome sequencing are shown in 5 Mb bins for 30 healthy individuals (top) and 8 lung cancer patients (bottom). B, Analysis of fragmentation profiles of healthy cfDNA (top), lung cancer cfDNA (middle) and healthy lymphocytes (bottom) as well as lymphocyte profiles obtained from chromosome 1 at 1 Mb resolution. Healthy lymphocyte profiles were scaled with a standard deviation equal to the standard deviation of the median healthy cfDNA profile. The healthy cfDNA patterns closely mirrored the cfDNA patterns in healthy lymphocytes, whereas the lung cancer cfDNA profiles were more diverse and distinct from both the healthy and lymphocyte profiles. C, Smoothed median distances between adjacent nucleosomes using 100k b-bins and centered at zero from healthy cfDNA (top) and nuclease-digested healthy lymphocytes (middle) are illustrated along with the first eigenvector for the genome contact matrix obtained through previously reported Hi-C analysis of lymphoblastoid cells (bottom). Healthy cfDNA nucleosome distances closely mirrored those in nuclease-digested lymphocytes and those obtained from lymphoblastoid Hi-C analysis. cfDNA fragmentation profiles from healthy individuals (n=30) were highly correlated to the median fragmentation profiles of lymphocytes (D), healthy cfDNA (E) and lymphocyte nucleosome (F) distances, whereas patients with lung cancer had lower correlations. [Figure 8]cfDNA fragment length density in healthy individuals and patients with lung cancer. cfDNA fragment lengths are shown for healthy individuals (n=30, grey) and patients with lung cancer (n=8, blue). [Figure 9] 9A and 9B. Subsampling of whole genome sequence data for analysis of cfDNA fragmentation profiles. A, High coverage (9×) whole genome sequencing data was subsampled to 2×, 1×, 0.5×, 0.2× and 0.1× fold coverage. For each subsampled fold coverage, mean-centered whole genome fragmentation profiles in 5 Mb bins for 30 healthy individuals and 8 patients with lung cancer are shown with the profile median shown in blue. B, Pearson correlation of subsampled profiles to the original profiles at 9× coverage for healthy individuals and patients with lung cancer. [Figure 10] cfDNA fragmentation profiles and sequence variations during treatment. Cancer detection and monitoring in serial blood draws from NSCLC patients (n=19) undergoing treatment with a targeted tyrosine kinase inhibitor (black arrow) was performed using targeted sequencing (top) and whole genome fragmentation profiles (bottom). For each case, the vertical axis of the lower panel indicates 1x the correlation of each sample to the median healthy cfDNA fragmentation profile. Error bars illustrate the confidence intervals obtained from binomial tests for variant allele frequencies and confidence intervals calculated using Fisher transformation for whole genome fragmentation profiles. Although these approaches analyze different aspects of cfDNA (whole genome compared to specific variations), targeted sequencing and fragmentation profiles were similar for patients responding to treatment and those with stable or progressing disease. Because fragmentation profiles reflect both genomic and epigenomic variations, whereas variant allele frequencies reflect only individual mutations, variant allele frequencies alone may not reflect the absolute level of correlation of fragmentation profiles to healthy individuals. [Figure 11]11A-11C. cfDNA fragmentation profiles in healthy individuals and patients with cancer. A, Fragmentation profiles (bottom) in relation to tumor copy number alterations (top) in colorectal cancer patients for whom parallel analysis of tumor tissue was performed. The distribution of segment means and integer copy numbers are shown at the top right in the indicated colors. Altered fragmentation profiles were present in regions of the genome that were copy neutral and were more affected in regions with copy number alterations. B, GC-prepared fragmentation profiles obtained from 1-2× whole genome sequencing for healthy individuals and patients with cancer are illustrated by cancer type using 5 Mb windows. Healthy profile medians are shown in black and 98% confidence bands in grey. For patients with cancer, each profile is colored based on their correlation to the healthy median. C, If more than 10% of the cancer samples had fragment ratios more than 3 standard deviations from the healthy fragment ratio median, the window is shown in orange. These analyses highlight the numerous position-dependent variations across the genome in the cfDNA of individuals with cancer. [Figure 12] Figure 12A and Figure 12B: cfDNA fragment length profiles in copy-neutral regions in healthy individuals and one patient with colorectal cancer. A, Fragmentation profiles in 211 copy-neutral windows in chromosomes 1-6 for 25 randomly selected healthy individuals (gray). For a patient with colorectal cancer (CGCRC291) with an estimated mutant allele frequency of 20%, the cancer fragment length profile was diluted to approximately 10% tumor contribution (blue). A and B, Marginal densities of fragment profiles for healthy samples and cancer patients show significant overlap (A, right), but the fragmentation profiles are distinct, as evident by visualization of fragmentation profiles (A, left) and separation of colorectal cancer patients from healthy samples in principal component analysis (B). [Figure 13]Figure 13A and Figure 13B: Genome-wide GC correction of cfDNA fragments. To estimate and adjust for the effect of GC content on sequencing coverage, coverage in non-overlapping 100 kb genomic windows was calculated across autosomes. For each window, the average GC of aligned fragments was calculated. A, loess smoothing of raw coverage for two randomly selected healthy subjects (CGPLH189 and CGPLH380) and two cancer patients (CGPLLU161 and CGPLBR24) with undetectable aneuploidy (PA score < 2.35) (top row). After subtracting the average coverage predicted by the loess model, the residuals were rescaled to the median autosome coverage (bottom row). Because fragment length can also result in coverage bias, this GC correction procedure was performed separately for short (≦ 150 bp) and long (≧ 151 bp) fragments. Although the 100 kb bins (blue dots) on chromosome 19 consistently had less coverage than predicted by the loess model, we did not perform chromosome-specific corrections because such an approach would eliminate the effect of chromosome copy number on coverage. B, Taken together, we found limited correlation between short or long fragment coverage and corrected GC content among healthy subjects and cancer patients with PA scores below 3. [Figure 14]Schematic of the machine learning model. Gradient tree boosting machine learning was used to investigate whether cfDNA could be classified as having characteristics of cancer patients or healthy individuals. The machine learning model included fragmentation size and coverage characteristics in a genome-wide window as well as copy numbers of chromosome arms and mitochondrial DNA. A 10-fold cross-validation approach was used in which each sample was randomly assigned to one fold, and nine of the folds (90% of the data) were used for training and one fold (10% of the data) was used for testing. The prediction accuracy obtained from a single cross-validation is the average over the 10 possible combinations of test and training sets. Since this prediction accuracy may reflect bias from the initial randomization of patients, the entire procedure was repeated 10 times, including randomization of patients into folds. For all cases, feature selection and model estimation were performed on the training data and validated against the test data, and the test data was never used for feature selection. Finally, a DELFI score was obtained that can be used to classify individuals as likely to be healthy or have cancer. [Figure 15] Distribution of AUC over repeated 10-fold cross-validation. The 25th, 50th and 75th percentiles of the 100AUC for a cohort of 215 healthy individuals and 208 patients with cancer are indicated by dashed lines. [Figure 16] Figure 16A and Figure 16B: Genome-wide analysis of chromosome arm copy number changes and mitochondrial genome representation. A, Z-scores for each autosomal arm are shown for healthy individuals (n=215) and patients with cancer (n=208). The vertical axis illustrates normal copies at zero, with positive and negative values ​​indicating gain and loss of arms, respectively. Z-scores above 50 or below -50 are thresholded at the indicated values. B, The percentage of reads mapping to the mitochondrial genome is shown for healthy individuals and patients with cancer. [Figure 17]Figures 17A and 17B. Cancer detection using DELFI. A, Receiver-operator characteristics for detecting cancer using cfDNA fragmentation profiles and other genome-wide features in a machine learning approach are illustrated for a cohort of 215 healthy individuals and 208 patients with cancer (DELFI, AUC=0.94), with specificities of 95% or greater shaded blue. Machine learning analyses of chromosome arm copy number (Chr copy number (ML)) and mitochondrial genome copy number (mtDNA) are shown in the indicated colors. B, Analysis of individual cancer types using the DELFI combined approach had AUCs ranging from 0.86 to over 0.99. [Figure 18] DELFI detection of cancer by stage. Receiver-operator characteristics for detecting cancer using cfDNA fragmentation profiles and other genome-wide features in a machine learning approach are illustrated for a cohort of 215 healthy individuals and 208 patients with cancer at each stage, with specificities of 95% or higher shaded blue. [Figure 19] DELFI tissue of origin prediction. Receiver operator characteristics for DELFI tissue prediction for cholangiocarcinoma, breast cancer, colorectal cancer, gastric cancer, lung cancer, ovarian cancer and pancreatic cancer are illustrated. To increase sample size within the cancer type class, cases detected at 90% specificity were included, and the lung cancer cohort was supplemented with additional baseline cfDNA data from 18 lung cancer patients with prior treatment (see, e.g., Shen et al., 2018 Nature, 563:579-583). [Figure 20] Cancer detection using DELFI and mutation-based cfDNA approaches. DELFI (green) and targeted sequencing for mutation identification (blue) were performed independently in a cohort of 126 patients with breast, cholangiocarcinoma, colorectal, gastric, lung or ovarian cancer. The number of individuals detected by each approach and combination of approaches is shown for DELFI detection with 98% specificity, for targeted sequencing with >99% specificity, and for the combination with 98% specificity. ND indicates not detected. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] This document provides methods and materials for determining a cfDNA fragmentation profile in a mammal (e.g., in a sample obtained from a mammal). As used herein, the terms "fragmentation profile", "position-dependent differences in fragmentation patterns" and "differences in fragment size and coverage in a position-dependent manner across the genome" are equivalent and can be used interchangeably. In some cases, 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 a 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 assessed to determine a cfDNA fragmentation profile. As described herein, the cfDNA fragmentation profile of a mammal with cancer is more heterogeneous (e.g., in fragment length) than the cfDNA fragmentation profile of a healthy mammal (e.g., a mammal without cancer). Thus, this document also provides methods and materials for assessing, monitoring, and / or treating a mammal (e.g., a human) that has or is suspected of having cancer. In some cases, 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 assessed to determine the presence of cancer in the mammal, and optionally the tissue of origin, based at least in part on the cfDNA fragmentation profile of the mammal. In some cases, this document provides methods and materials for monitoring a mammal as having cancer. For example, a sample (e.g., a blood sample) obtained from a mammal can be assessed to determine the presence of cancer in the mammal, based at least in part on the cfDNA fragmentation profile of the mammal. In some cases, this document provides methods and materials for identifying a mammal as having cancer and administering one or more cancer treatments to the mammal to treat the mammal.For example, a sample (e.g., a blood sample) obtained from the mammal can be assessed to determine whether the mammal has cancer based, at least in part, on the mammal's cfDNA fragmentation profile, and the mammal can be administered one or more cancer treatments.

[0016] The cfDNA fragmentation profile can include one or more cfDNA fragmentation patterns. The cfDNA fragmentation pattern can 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 cases, the cfDNA fragmentation pattern includes two or more (e.g., 2, 3, or 4) of median fragment size, fragment size distribution, ratio of small cfDNA fragments to large cfDNA fragments, and cfDNA fragment coverage. In some cases, the cfDNA fragmentation profile can be a whole genome cfDNA profile (e.g., a whole genome cfDNA profile in a window spanning the genome). In some cases, the cfDNA fragmentation profile can be a targeted region profile. The targeted region can be any suitable part of the genome (e.g., a region of a chromosome). As described herein, examples of chromosomal regions for which cfDNA fragmentation profiles can be determined include, but are not limited to, chromosomal portions (e.g., 2q, 4p, 5p, 6q, 7p, 8q, 9q, 10q, 11q, 12q, and / or portions of 14q) and chromosomal arms (e.g., 8q, 13q, 11q, and / or 3p chromosomal arms). In some cases, the cfDNA fragmentation profile can include profiles of two or more targeted regions.

[0017] In some cases, the cfDNA fragmentation profile can be used to identify cfDNA fragment length changes (e.g., alterations). The variations can be genome-wide variations or can be changes in one or more targeted regions / locuses. The targeted region can be any region that contains one or more cancer-specific variations. Examples of cancer-specific variations and their chromosomal locations include, but are not limited to, those shown in Table 3 (Appendix C) and those shown in Table 6 (Appendix F). In some cases, the cfDNA fragmentation profile can be used to identify (e.g., simultaneously identify) about 10 variations to about 500 variations (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 variations).

[0018] In some cases, cfDNA fragmentation profile can be used to detect tumor-derived DNA. For example, cfDNA fragmentation profile can be used to detect tumor-derived DNA by comparing cfDNA fragmentation profile of mammal having or suspected to have cancer with reference cfDNA fragmentation profile (e.g., cfDNA fragmentation profile of healthy mammal and / or nucleosomal DNA fragmentation profile of healthy cell from mammal having or suspected to have cancer). In some cases, reference cfDNA fragmentation profile is a profile previously generated from healthy mammal. For example, the method provided herein can be used to determine reference cfDNA fragmentation profile in healthy mammal, and the reference cfDNA fragmentation profile can be stored (e.g., in computer or other electronic storage medium) for future comparison with test cfDNA fragmentation profile in mammal having or suspected to have cancer. In some cases, the reference cfDNA fragmentation profile (e.g., a stored cfDNA fragmentation profile) of the healthy mammal is determined over the entire genome. In some cases, the reference cfDNA fragmentation profile (e.g., a stored cfDNA fragmentation profile) of the healthy mammal is determined over a subgenomic interval.

[0019] In some cases, the cfDNA fragmentation profile can be used to identify a mammal (e.g., a human) as having cancer (e.g., colorectal cancer, lung cancer, breast cancer, gastric cancer, pancreatic cancer, bile duct cancer and / or ovarian cancer).

[0020] The cfDNA fragmentation profile can include a cfDNA fragment size pattern, and the cfDNA fragments can be any suitable size. For example, the cfDNA fragments can be about 50 base pairs (bp) to about 400 bp in length. As described herein, a mammal with cancer can have a cfDNA fragment size pattern that contains a median cfDNA fragment size that is shorter than the median cfDNA fragment size in a healthy mammal. A healthy mammal (e.g., a mammal without cancer) can have a cfDNA fragment size that has a median cfDNA fragment size of about 166.6 bp to about 167.2 bp (e.g., about 166.9 bp). In some cases, a mammal with cancer can have a cfDNA fragment size that is, on average, about 1.28 bp to about 2.49 bp (e.g., about 1.88 bp) shorter than the cfDNA fragment size in a healthy mammal. For example, a mammal with cancer can have a cfDNA fragment size that has a median cfDNA fragment size of about 164.11 bp to about 165.92 bp (e.g., about 165.02 bp).

[0021] The cfDNA fragmentation profile can include a cfDNA fragment size distribution. As described herein, a mammal with cancer can have a cfDNA fragment size distribution that is more variable than a cfDNA fragment size distribution in a healthy mammal. In some cases, the size distribution can be within the targeted region. A healthy mammal (e.g., a mammal without cancer) can have a cfDNA fragment size distribution of the targeted region that is about 1 or less than about 1. In some cases, a mammal with cancer can have a cfDNA fragment size distribution of the targeted region that is longer (e.g., 10, 15, 20, 25, 30, 35, 40, 45, 50 bp or longer, or any number of base pairs between these numbers) than a cfDNA fragment size distribution of the targeted region in a healthy mammal. In some cases, a mammal with cancer can have a cfDNA fragment size distribution of the targeted region that is shorter (e.g., 10, 15, 20, 25, 30, 35, 40, 45, 50 bp or shorter, or any number of base pairs between these numbers) than a cfDNA fragment size distribution of the targeted region in a healthy mammal. In some cases, the mammal with cancer can have a cfDNA fragment size distribution of the targeted region that is about 47 bp smaller to about 30 bp longer than the cfDNA fragment size distribution of the targeted region in a healthy mammal. In some cases, the mammal with cancer can have a cfDNA fragment size distribution of the targeted region that is an average of 10, 11, 12, 13, 14, 15, 15, 17, 18, 19, 20 bp or more difference in cfDNA fragment length. For example, the mammal with cancer can have a cfDNA fragment size distribution of the targeted region that is an average of about 13 bp difference in cfDNA fragment length. In some cases, the size distribution can be a whole genome size distribution. A healthy mammal (e.g., a mammal without cancer) can have a very similar distribution of short and long cfDNA fragments in the whole genome. In some cases, the mammal with cancer can have one or more variations (e.g., increases and decreases) in cfDNA fragment size in the whole genome. The one or more variations can be in any suitable chromosomal region of the genome.For example, variation can be present in a part of chromosome.The example of the part of chromosome that can contain one or more variations of cfDNA fragment size includes but is not limited to 2q, 4p, 5p, 6q, 7p, 8q, 9q, 10q, 11q, 12q and 14q part.For example, variation can be the whole chromosome arm (for example, the whole chromosome arm).

[0022] The cfDNA fragmentation profile can include the ratio of small cfDNA fragments to large cfDNA fragments and the correlation of fragment ratios to reference fragment ratios. As used herein, for the ratio of small cfDNA fragments to large cfDNA fragments, the small cfDNA fragments can be about 100bp to about 150bp in length. As used herein, for the ratio of small cfDNA fragments to large cfDNA fragments, the large cfDNA fragments can be about 151bp to 220bp in length. As described herein, a mammal with cancer can have a lower (2-fold lower, 3-fold lower, 4-fold lower, 5-fold lower, 6-fold lower, 7-fold lower, 8-fold lower, 9-fold lower, 10-fold lower or lower) fragment ratio correlation (e.g., cfDNA fragment ratio correlation to reference DNA fragment ratio, such as DNA fragment ratio obtained from one or more healthy mammals) than in a healthy mammal. A healthy mammal (e.g., a mammal without cancer) can have a fragment ratio correlation (e.g., a correlation of cfDNA fragment ratios to a reference DNA fragment ratio, such as a DNA fragment ratio obtained from one or more healthy mammals) of about 1 (e.g., about 0.96). In some cases, a mammal with cancer can have a fragment ratio correlation (e.g., a correlation of cfDNA fragment ratios to a reference DNA fragment ratio, such as a DNA fragment ratio obtained from one or more healthy mammals) that is on average about 0.19 to about 0.30 (e.g., about 0.25) lower than the fragment ratio correlation in a healthy mammal (e.g., a correlation of cfDNA fragment ratios to a reference DNA fragment ratio, such as a DNA fragment ratio obtained from one or more healthy mammals).

[0023] The cfDNA fragmentation profile can include coverage of all fragments. Coverage of all fragments can include windows of coverage (e.g., non-overlapping windows). In some cases, coverage of all fragments can include windows of small fragments (e.g., fragments about 100 bp to about 150 bp in length). In some cases, coverage of all fragments can include windows of large fragments (e.g., fragments about 151 bp to about 220 bp in length).

[0024] In some cases, the cfDNA fragmentation profile can be used to identify the primary tissue of cancer (e.g., colorectal cancer, lung cancer, breast cancer, gastric cancer, pancreatic cancer, bile duct cancer or ovarian cancer). For example, the cfDNA fragmentation profile can be used to identify localized cancer. When the cfDNA fragmentation profile includes a target region profile, one or more variations described herein (e.g., in Table 3 (Appendix C) and / or Table 6 (Appendix F)) can be used to identify the primary tissue of cancer. In some cases, one or more variations in chromosomal regions can be used to identify the primary tissue of cancer.

[0025] The cfDNA fragmentation profile can be obtained using any suitable method. In some cases, cfDNA obtained from a mammal (e.g., a mammal having or suspected of having cancer) can be subjected to whole genome sequencing (e.g., low-coverage whole genome sequencing) and processed into a sequencing library that can be mapped to the genome and analyzed to determine cfDNA fragment lengths. The mapped sequences can be analyzed in non-overlapping windows encompassing the genome. The windows can be of any suitable size. For example, the windows can be thousands to millions of bases in length. As one non-limiting example, the windows can be about 5 megabases (Mb) long. 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. A cfDNA fragmentation profile can be determined within each window. In some cases, the cfDNA fragmentation profile can be obtained as described in Example 1. In some cases, the cfDNA fragmentation profile can be obtained as shown in FIG. 1.

[0026] In some cases, the methods and materials described herein can also include machine learning.For example, machine learning can be used to identify altered fragmentation profiles (e.g., using cfDNA fragment coverage, cfDNA fragment size, chromosome coverage and mtDNA).

[0027] In some cases, the methods and materials described herein may be the only method used to identify a mammal (e.g., a human) as having cancer (e.g., colorectal cancer, lung cancer, breast cancer, gastric cancer, pancreatic cancer, bile duct cancer, and / or ovarian cancer). For example, determining a cfDNA fragmentation profile may be the only method used to identify a mammal as having cancer.

[0028] In some cases, the methods and materials described herein can be used together with one or more additional methods used to identify a mammal (e.g., a human) as having cancer (e.g., colorectal cancer, lung cancer, breast cancer, gastric cancer, pancreatic cancer, bile duct cancer, and / or ovarian cancer). Examples of methods used to identify a mammal as having cancer include, but are not limited to, identifying one or more cancer-specific sequence variations, identifying one or more chromosomal variations (e.g., aneuploidies and rearrangements), and identifying other cfDNA variations. For example, determining a cfDNA fragmentation profile can be used together with identifying one or more cancer-specific mutations in the genome of the mammal to identify a mammal as having cancer. For example, determining a cfDNA fragmentation profile can be used together with identifying one or more aneuploidies in the genome of the mammal to identify a mammal as having cancer.

[0029] In some embodiments, this document also provides methods and materials for assessing, monitoring, and / or treating a mammal (e.g., a human) that has or is suspected of having cancer. In some cases, 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 assessed to determine whether the mammal has cancer, based at least in part on the cfDNA fragmentation profile of the mammal. In some cases, this document provides methods and materials for identifying the location (e.g., anatomic site or tissue of origin) of a cancer in a mammal. For example, a sample (e.g., a blood sample) obtained from a mammal can be assessed to determine the tissue of origin of a cancer in the mammal, based at least in part on the cfDNA fragmentation profile of the mammal. In some cases, this document provides methods and materials for identifying a mammal as having cancer and 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 assessed 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. In some cases, this document provides methods and materials for treating a mammal with cancer. For example, one or more cancer treatments can be administered to a mammal identified as having cancer (e.g., based at least in part on the cfDNA fragmentation profile of the mammal) to treat the mammal. In some cases, during or after a cancer treatment (e.g., any of the cancer treatments described herein), the mammal can be monitored (or selected for increased monitoring) and / or undergo further diagnostic testing.In some cases, monitoring can include assessing a mammal having or suspected of having cancer, e.g., by assessing a sample (e.g., a blood sample) obtained from the mammal to determine a cfDNA fragmentation profile of the mammal as described herein, and changes in the cfDNA fragmentation profile over time can be used to identify a response to treatment and / or to identify the mammal as having cancer (e.g., residual cancer).

[0030] Any suitable mammal can be assessed, monitored, and / or treated as described herein. The mammal can be a mammal with cancer. The mammal can be a mammal suspected of having cancer. Examples of mammals that can be assessed, monitored, and / or treated as described herein include, but are not limited to, humans, primates such as monkeys, dogs, cats, horses, cows, pigs, sheep, mice, and rats. For example, a human with cancer or suspected of having cancer can be assessed to determine cfDNA fragmentation profile as described herein, and can be treated with one or more cancer treatments as required as described herein.

[0031] Any suitable sample obtained from a mammal can be assessed (e.g., assessed for DNA fragmentation pattern) as described herein. In some cases, the sample can include DNA (e.g., genomic DNA). In some cases, the sample can include cfDNA (e.g., circulating tumor DNA (ctDNA)). In some cases, the sample can be a fluid sample (e.g., liquid biopsy). Examples of samples that can contain DNA and / or polypeptides include, but are not limited to, blood (e.g., whole blood, serum or plasma), amniotic membrane, tissue, urine, cerebrospinal fluid, saliva, sputum, bronchoalveolar lavage fluid, bile, lymphatic fluid, cyst fluid, stool, ascites, Papanicolaou smear, breast milk and exhaled breath condensate. For example, a plasma sample can be assessed to determine a cfDNA fragmentation profile as described herein.

[0032] A sample obtained from a mammal to be assessed (e.g., assessed for DNA fragmentation patterns) as described herein can contain any suitable amount of cfDNA. In some cases, the sample can contain a limited amount of DNA. For example, cfDNA fragmentation profiles can be obtained from samples that contain less DNA than is typically required for other cfDNA analysis methods, such as those described in, e.g., Phallen et al., 2017 Sci Transl Med 9; Cohen et al., 2018 Science 359:926; Newman et al., 2014 Nat Med 20:548; and Newman et al., 2016 Nat Biotechnol 34:547).

[0033] In some cases, the sample can be treated (e.g., to isolate and / or purify DNA and / or polypeptides from the sample). For example, DNA isolation and / or purification can include cell lysis (e.g., with detergents and / or surfactants), protein removal (e.g., with proteases), and / or RNA removal (e.g., with RNases). As another example, polypeptide isolation and / or purification can include cell lysis (e.g., with detergents and / or surfactants), DNA removal (e.g., with DNases), and / or RNA removal (e.g., with RNases).

[0034] A mammal having or suspected of having any suitable type of cancer can be assessed (e.g., to determine a cfDNA fragmentation profile) and / or treated (e.g., by administering one or more cancer treatments to the mammal) using the methods and materials described herein. The cancer can be any stage of cancer. In some cases, the cancer can be an early stage cancer. In some cases, the cancer can be asymptomatic cancer. In some cases, the cancer can be residual disease and / or recurrence (e.g., after surgical resection and / or cancer treatment). The cancer can be any type of cancer. Examples of types of cancer that can be assessed, monitored, and / or treated as described herein include, but are not limited to, colorectal cancer, lung cancer, breast cancer, gastric cancer, pancreatic cancer, bile duct cancer, and ovarian cancer.

[0035] When treating a mammal having or suspected of having cancer as described herein, the mammal may be administered one or more cancer treatments. The cancer treatment may be any suitable cancer treatment. The one or more cancer treatments described herein may be administered to the mammal at any suitable frequency (e.g., once or multiple times over a period ranging from several days to several weeks). Examples of cancer treatments include, but are not limited to, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy (e.g., T cells with chimeric antigen receptors and / or wild-type or modified T cell receptors), targeted therapy, such as kinase inhibitors (e.g., kinase inhibitors that target specific genetic lesions such as translocations or mutations), administration of (e.g., kinase inhibitors, antibodies, bispecific antibodies), signal transduction inhibitors, bispecific antibodies or antibody fragments (e.g., BiTEs), monoclonal antibodies, immune checkpoint inhibitors, surgery (e.g., surgical resection), or any combination of the above. In some cases, treating cancer may reduce the severity of the cancer, reduce the symptoms of the cancer, and / or reduce the number of cancer cells present in the mammal.

[0036] In some cases, the cancer treatment can include an immune checkpoint inhibitor. Non-limiting examples of immune checkpoint inhibitors include nivolumab (Opdivo), pembrolizumab (Keytruda), atezolizumab (Tecentriq), avelumab (Bavencio), durvalumab (Imfinzi), and ipilimumab (Yervoy). For example, Pardoll (2012) Nat. Rev Cancer 12:252-264; Sun et al. (2017) Eur Rev Med Pharmacol Sci 21(6):1198-1205; Hamanishi et al. (2015) J. Clin. Oncol. 33(34):4015-22; Brahmer et al. (2012) N Engl J Med 366(26):2455-65; Ricciuti et al. (2017) J. Thorac Oncol. 12(5):e51-e55; Ellis et al. (2017) Clin Lung Cancer pii:S1525-7304(17)30043-8; Zou and Awad (2017) Ann Oncol 28(4):685-687; Sorscher (2017) N See Engl J Med 376(10:996-7; Hui et al., (2017) Ann Oncol 28(4):874-881; Vansteenkiste et al., (2017) Expert Opin Biol Ther 17(6):781-789; Hellmann et al., (2017) Lancet Oncol. 18(1):31-41; Chen (2017) J. Chin Med Assoc 80(1):7-14.

[0037] In some cases, the cancer treatment can be adoptive T cell therapy (e.g., T cells with chimeric antigen receptors and / or wild-type or modified T cell receptors).See, for example, Rosenberg and Restifo (2015) Science 348(6230):62-68; Chang and Chen (2017) Trends Mol Med 23(5):430-450; Yee and Lizee (2016) Cancer J. 23(2):144-148; Chen et al. (2016) Oncoimmunology 6(2):e1273302; US Patent Application Publication No. 2016 / 0194404; US Patent Application Publication No. 2014 / 0050788; US Patent No. 2014 / 0271635; US Patent No. 9,233,125; each of which is incorporated herein by reference in its entirety.

[0038] In some cases, the cancer treatment can be a chemotherapeutic agent. Non-limiting examples of chemotherapeutic agents include amasacriline, azacitidine, axathioprine, bevacizumab (or an antigen-binding fragment thereof), bleomycin, busulfan, carboplatin, capecitabine, chlorambucil, cisplatin, cyclophosphamide, cytarabine, dacarbazine, daunorubicin, docetaxel, doxifluridine, doxorubicin, epirubicin, erlotinib hydrochloride, etoposide, fiudarabine, floxuridine, fludarabine, and fluorouracil. , gemcitabine, hydroxyurea, idarubicin, ifosfamide, irinotecan, lomustine, mechlorethamine, melphalan, mercaptopurine, methotrexate, mitomycin, mitoxantrone, oxaliplatin, paclitaxel, pemetrexed, procarbazine, all-trans retinoic acid, streptozocin, tafluposide, temozolomide, teniposide, thioguanine, topotecan, uramustine, valrubicin, vinblastine, vincristine, vindesine, vinorelbine and combinations thereof. Further examples of anti-cancer treatments are known in the art; see, for example, the American Society of Clinical Oncology (ASCO), the European Society for Medical Oncology (ESMO) or the National Comprehensive Cancer Network (NCCN) guidelines for treatment.

[0039] When monitoring a mammal having or suspected of having cancer as described herein (e.g., based at least in part on the cfDNA fragmentation profile of the mammal), the monitoring can be before, during, and / or after a course of cancer treatment. The monitoring methods provided herein can be used to determine the effectiveness of one or more cancer treatments and / or to select a mammal for increased monitoring. In some cases, the monitoring can include determining a cfDNA fragmentation profile as described herein. For example, a cfDNA fragmentation profile can be obtained prior to administration of one or more cancer treatments to a mammal having or suspected of having cancer, one or more cancer treatments can be administered to the mammal, and one or more cfDNA fragmentation profiles can be obtained during the course of the cancer treatment. In some cases, the cfDNA fragmentation profile can change during the cancer treatment (e.g., any of the cancer treatments described herein). For example, a cfDNA fragmentation profile indicative of the mammal having cancer can change to a cfDNA fragmentation profile indicative of the mammal not having cancer. Such a change in the cfDNA fragmentation profile can indicate that the cancer treatment is working. Conversely, the cfDNA fragmentation profile can remain static (e.g., the same or approximately the same) during the cancer treatment (e.g., any of the cancer treatments described herein). Such a static cfDNA fragmentation profile can indicate that the cancer treatment is not working. In some cases, the monitoring can include conventional techniques that can monitor one or more cancer treatments (e.g., the effectiveness of one or more cancer treatments). In some cases, the mammals selected for increased monitoring can be subjected to diagnostic tests (e.g., any of the diagnostic tests disclosed herein) at an increased frequency compared to mammals not selected for increased monitoring. For example, the mammals selected for increased monitoring can be subjected to diagnostic tests twice daily, daily, twice weekly, weekly, twice monthly, monthly, quarterly, semiannually, annually, or any frequency therein. In some cases, the mammals selected for increased monitoring can be subjected to one or more additional diagnostic tests compared to mammals not selected for increased monitoring. For example, the mammals selected for increased monitoring can be subjected to two diagnostic tests, whereas the mammals not selected for increased monitoring are subjected to only a single diagnostic test (or no diagnostic test).In some cases, the mammal selected for increased monitoring can also be selected for further diagnostic testing. Once the presence of a tumor or cancer (e.g., cancer cells) is identified (e.g., by any of the various methods disclosed herein), the mammal may benefit from both increased monitoring (e.g., to assess the progression of the tumor or cancer in the mammal and / or to assess the development of one or more cancer biomarkers, such as mutations) and further diagnostic testing (e.g., to determine the size and / or exact location (e.g., tissue of origin) of the tumor or cancer). In some cases, one or more cancer treatments can be administered to the mammal selected for increased monitoring after the cancer biomarkers are detected and / or after the cfDNA fragmentation profile of the mammal has not improved or worsened. Any of the cancer treatments disclosed herein or known in the art can be administered. For example, the mammal selected for increased monitoring can be further monitored, and if the presence of cancer cells is maintained throughout the increased monitoring period, the cancer treatment can be administered. Additionally or alternatively, the mammal selected for increased monitoring can be administered a cancer treatment and further monitored as the cancer treatment progresses. In some cases, after the mammal selected for increased monitoring is administered a cancer treatment, the increased monitoring will reveal one or more cancer biomarkers (e.g., mutations). In some cases, such one or more cancer biomarkers will provide a cause for administering a different cancer treatment (e.g., resistance mutations may arise in cancer cells during cancer treatment, and cancer cells harboring this resistance mutation are resistant to the original cancer treatment).

[0040] In the case where a mammal is identified as having cancer as described herein (e.g., based at least in part on the cfDNA fragmentation profile of the mammal), the identification can be before and / or during a course of cancer treatment. The methods of identifying a mammal as having cancer provided herein can be used as a first diagnosis to identify a mammal (e.g., as having cancer before any course of treatment) and / or to select a mammal for further diagnostic testing. In some cases, once it is determined that a mammal has cancer, the mammal can be subjected to further testing and / or selected for further diagnostic testing. In some cases, the methods provided herein can be used to select a mammal for further diagnostic testing in a period prior to the period when conventional techniques are capable of diagnosing a mammal with an early stage cancer. For example, the methods provided herein for selecting a mammal for further diagnostic testing can be used when the mammal has not been diagnosed with cancer by conventional methods and / or when the mammal is not known to have cancer. In some cases, a mammal selected for further diagnostic testing can be subjected to an increased frequency of diagnostic testing (e.g., any of the diagnostic tests disclosed herein) compared to a mammal not selected for further diagnostic testing. For example, the mammals selected for further diagnostic testing can be subjected to diagnostic testing twice daily, daily, twice weekly, weekly, twice monthly, monthly, quarterly, semi-annually, annually, or any frequency therein. In some cases, the mammals selected for further diagnostic testing can be subjected to one or more further diagnostic tests compared to the mammals not selected for further diagnostic testing. For example, the mammals selected for further diagnostic testing can be subjected to two diagnostic tests, whereas the mammals not selected for further diagnostic testing are subjected to only a single diagnostic test (or no diagnostic test).In some cases, the diagnostic test method can determine the presence of a cancer of the same type (e.g., having the same tissue or origin) as the cancer originally detected (e.g., based at least in part on the cfDNA fragmentation profile of the mammal). Additionally or alternatively, the diagnostic test method can determine the presence of a cancer of a different type than the cancer originally detected. In some cases, the diagnostic test method is a scan. In some cases, the scan is a computed tomography (CT), a CT angiogram (CTA), an esophagogram (barium swallow), a barium enema, a magnetic resonance imaging (MRI), a PET scan, an ultrasound (e.g., endobronchial ultrasound, endoscopic ultrasound), an X-ray, a DEXA scan. In some cases, the diagnostic testing method is a physical examination such as anoscopy, bronchoscopy (e.g., autofluorescence bronchoscopy, white light bronchoscopy, navigation bronchoscopy), colonoscopy, digital breast tomosynthesis, endoscopic retrograde cholangiopancreatography (ERCP), esophagogastroduodenoscopy, mammography, Papanicolaou smear, gynecological pelvic exam, positron emission tomography-computed tomography (PET-CT) scan, etc. In some cases, the mammal selected for further diagnostic testing can also be selected for increased monitoring. Once the presence of a tumor or cancer (e.g., cancer cells) has been identified (e.g., by any of the various methods disclosed herein), the mammal may benefit from both increased monitoring (e.g., to assess the progression of the tumor or cancer in the mammal and / or to assess the development of one or more cancer biomarkers, such as mutations) and further diagnostic testing (e.g., to determine the size and / or exact location of the tumor or cancer). In some cases, cancer treatment is performed on the mammal selected for further diagnostic testing after the cancer biomarker is detected and / or the cfDNA fragmentation profile of the mammal is improved or not worsened.Any of the cancer treatments disclosed herein or known in the art can be performed.For example, the mammal selected for further diagnostic testing can be subjected to further diagnostic testing, and if the presence of tumor or cancer is confirmed, a cancer treatment can be performed. Additionally or alternatively, the mammal selected for further diagnostic testing can be subjected to a cancer treatment and further monitored as the cancer treatment progresses. In some cases, after the mammal selected for further diagnostic testing is subjected to a cancer treatment, additional testing will reveal one or more cancer biomarkers (e.g., mutations). In some cases, such one or more cancer biomarkers (e.g., mutations) will provide a reason to perform a different cancer treatment (e.g., resistance mutations may occur in cancer cells during cancer treatment, and cancer cells carrying this resistance mutation are resistant to the original cancer treatment).

[0041] The invention is further described in the following examples, which do not limit the scope of the invention described in the claims. EXAMPLES

[0042] Example 1: Cell-free DNA fragmentation in patients with cancer Analysis of cell-free DNA has primarily focused on targeted sequencing of specific genes. Although such studies allow the detection of a small number of tumor-specific variations in patients with cancer, not all patients, especially those with early stage disease, have detectable changes. Whole genome sequencing of cell-free DNA can identify chromosomal abnormalities and rearrangements in cancer patients, but the detection of such variations has been difficult, in part because it is difficult to distinguish a small number of abnormal chromosomal changes from normal chromosomal changes (Leary et al., 2010 Sci Transl Med 2:20ra14; and Leary et al., 2012 Sci Transl Med 4:162ra154). Other efforts have suggested that nucleosome patterns and chromatin structures may differ between cancer and normal tissues, and that cfDNA in patients with cancer may result in abnormal cfDNA fragment sizes and locations (Snyder et al., 2016 Cell 164:57; Jahr et al., 2001 Cancer Res 61:1659; Ivanov et al., 2015 BMC Genomics 16(Suppl 13):S1). However, the amount of sequencing required for nucleosome footprint analysis of cfDNA is infeasible for routine analysis.

[0043] The sensitivity of any cell-free DNA approach depends on the number of potential variations examined, as well as the technical and biological limits of detecting such variations. A typical blood sample contains approximately 2000 genome equivalents of cfDNA per mL of plasma (Phallen et al., 2017 Sci Transl Med 9), and the theoretical limit of detection of a single variation cannot exceed one variant in thousands relative to the wild-type molecule. Approaches that detect a larger number of variations in the same number of genome equivalents would be more sensitive with respect to detecting cancer in the circulation. Monte Carlo simulations indicate that increasing the number of potential abnormalities detected from just a few to tens or hundreds has the potential to improve the detection limit by orders of magnitude, as has the recent probability analysis of multiple methylation changes in cfDNA (Figure 2).

[0044] This study presents a novel method, called DELFI, for the detection of cancer and further identification of primary tissue using whole genome sequencing (Figure 1). The approach uses cfDNA fragmentation profiles and machine learning to distinguish patterns of healthy blood cell DNA from tumor-derived DNA and identify primary tumor tissue. DELFI was used for retrospective analysis of cfDNA from 245 healthy individuals and 236 patients with breast, colorectal, lung, ovarian, pancreatic, gastric or bile duct cancer, many of whom exhibited localized disease. Assuming that this approach has a sensitivity of 0.80 or greater for discriminating cancer patients from healthy individuals while maintaining a specificity of 0.95, a study of at least 200 cancer patients would allow an estimate of the true sensitivity with a margin of error of 0.06, at the desired specificity of 0.95 or greater.

[0045] material and method Patient and sample characteristics Plasma samples from healthy individuals and plasma and tissue samples from patients with breast, lung, ovarian, colorectal, bile duct or gastric cancer were obtained from ILSBio / Bioreclamation, Aarhus University, Herlev Hospital, University of Copenhagen, Hvidovre Hospital, University Hospitals of Utrecht University, University Hospitals of Amsterdam, Netherlands Cancer Institute and University of California, San Diego. All samples were obtained under protocols approved by the Institutional Review Boards at the participating institutions with informed consent for research use. Plasma samples from healthy individuals were obtained during routine screening such as colonoscopy or Pap smear. Individuals were considered healthy if they had no history of cancer and a negative screening result.

[0046] Plasma samples from individuals with breast, colorectal, gastric, lung, ovarian, pancreatic and bile duct cancer were obtained at the time of diagnosis, prior to tumor resection or treatment. Nineteen lung cancer patients analyzed for changes in cfDNA fragmentation profiles across multiple time points had been treated with anti-EGFR or anti-ERBB2 therapy (see, e.g., Phallen et al., 2019 Cancer Research 15, 1204-1213). Clinical data for all patients included in the study are listed in Table 1 (Appendix A). Gender was confirmed through genomic analysis of X and Y chromosome representation. Pathological staging of gastric cancer patients was performed after neoadjuvant therapy. Samples with unknown tumor stage were marked as stage X or unknown.

[0047] Nucleosomal DNA purification Live frozen lymphocytes were purified from leukocytes obtained from healthy men (C0618) and women (D0808-L) (Advanced Biotechnologies Inc., Eldersburg, MD). For nucleosomal DNA purification using the EZ Nucleosomal DNA Prep Kit (Zymo Research, Irvine, CA), 1 × 10 6 An aliquot of cells was used. First, cells were treated with 100 μl of Nuclei Prep Buffer and incubated on ice for 5 min. After centrifugation at 200 g for 5 min, the supernatant was discarded and The precipitated nuclei were treated twice with 100 μl of Atlantis Digestion Buffer or with 100 μl of Micrococcal Nuclease (MN) Digestion Buffer. Finally, the nuclear DNA of the cells was fragmented with 0.5 U of Atlantis dsDNase for 20 min at 42°C or with 1.5 U of MNase for 20 min at 37°C. The reaction was stopped with 5X MN Stop Buffer, and DNA was purified using a Zymo-Spin™ IIC Column. The concentration and quality of the eluted nuclear DNA was analyzed using a Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA).

[0048] Sample preparation and sequencing of cfDNA Whole blood was collected in EDTA tubes and processed immediately or within 1 day after storage at 4°C, or, for the three cancer patients who were part of the monitoring analysis, in Streck tubes and processed within 2 days of collection. Plasma and cellular components were separated by centrifugation at 800g for 10 min at 4°C. Plasma was centrifuged again at 18,000g at room temperature to remove any remaining cellular debris and stored at -80°C until the time of DNA extraction. DNA was isolated from plasma using the Qiagen Circulating Nucleic Acids Kit (Qiagen GmbH) and eluted into LoBind tubes (Eppendorf AG). The concentration and quality of cfDNA were assessed using a Bioanalyzer 2100 (Agilent Technologies).

[0049] NGS cfDNA libraries were prepared for whole genome sequencing and targeted sequencing using 5–250 ng of cfDNA as described elsewhere (see, e.g., Phallen et al., 2017 Sci Transl Med 9:eaan2415). Briefly, genomic libraries were prepared using the NEBNext DNA Library Prep Kit for Illumina [New England Biolabs (NEB)] with four major modifications to the manufacturer's instructions; (i) to minimize sample loss during the elution and tube transfer steps, the library purification step used an on-bead AMPure XP approach (see, e.g., Fisher et al., 2011 Genome Biol 12;R1); (ii) the NEB Next End Repair, A-tailing and adapter ligation enzymes and buffer volumes were adjusted appropriately to accommodate the on-bead AMPure XP purification strategy; (iii) a pool of eight unique Illumina dual index adapters with 8 base pair (bp) barcodes was used in the ligation reaction instead of the standard Illumina single or dual index adapters with 6 or 8 bp barcodes, respectively; (iv) cfDNA libraries were amplified using Phusion Hot Start Polymerase.

[0050] Whole genome libraries were sequenced directly. For targeted libraries, capture was performed using Agilent SureSelect reagents and a custom set of hybridization probes targeting 58 genes according to the manufacturer's guidelines (see, e.g., Phallen et al., 2017 Sci Transl Med 9:eaan2415). Phusion Hot Start Polymerase (NEB) was used to amplify captured libraries. The DNA 1000 Kit (Agilent Technologies) was used to assess the concentration and quality of captured cfDNA libraries on a Bioanalyzer 2100. Targeted libraries were sequenced using a 100bp paired-end run on an Illumina HiSeq 2000 / 2500 (Illumina).

[0051] Analysis of targeted sequencing data from cfDNA Analysis of targeted NGS data for cfDNA samples was performed as described elsewhere (see, e.g., Phallen et al., 2017 Sci Transl Med 9:eaan2415). Briefly, primary processing, including demultiplexing and masking of dual-index adapter sequences, was completed using Illumina CASAVA (Consensus Assessment of Sequence and Variation) software (version 1.8). Sequence reads were aligned to the human reference genome (version hg18 or hg19) using NovoAlign, and selected regions were further realigned using the Needleman-Wunsch method (see, e.g., Jones et al., 2015 Sci Transl Med 7:283ra53). The location of sequence variations was not affected by the different genome builds. Candidate mutations consisting of point mutations, small insertions and deletions were identified using VariantDx (see, e.g., Jones et al., 2015 Sci Transl Med 7:283ra53) (Personal Genome Diagnostics, Baltimore, MD) across the targeted regions of interest.

[0052] To analyze the fragment length of cfDNA molecules, each read pair from a cfDNA molecule was required to have a Phred quality score of 30 or greater. All duplicated ctDNA fragments, defined as having the same start, end and index barcode, were removed. For each mutation, only fragments in which one or both of the read pairs contained the mutated (or wild-type) base at the given position were included. The analysis was performed using the R packages Rsamtools and Genomic Alignments.

[0053] For each genomic locus where a somatic mutation was identified, the length of the fragment containing the mutant allele was compared to that of the wild-type allele. In cases where more than 100 mutant fragments were identified, a Welch two-sample t-test was used to compare the mean fragment lengths. For loci with fewer than 100 mutant fragments, a bootstrap procedure was performed. Specifically, restored N fragments containing the wild-type allele were sampled, where N represents the number of fragments with a mutation. For each bootstrap replicate of the wild-type fragment, the median of its length was calculated. The p-value was estimated as the proportion of bootstrap replicates with a wild-type fragment median length equal to or more extreme than the observed mutant fragment median length.

[0054] Analysis of whole genome sequencing data from cfDNA Primary processing of whole-genome NGS data for cfDNA samples, including demultiplexing and masking of dual-index adapter sequences, was performed using Illumina CASAVA (Consensus Assessment of Sequence and Variation) software (version 1.8.2). Sequence reads were aligned against the human reference genome (version hg19) using ELAND.

[0055] Read pairs and PCR duplicates with a MAPQ score of less than 30 for either read were removed. hg19 autosomes were tiled into 26,236 contiguous non-overlapping 100 kb bins. Low mappability regions, indicated by the 10% bins with the lowest coverage, were removed because the read data fell into Duke blacklist regions (see, e.g., hgdownload.cse.ucsc.edu / goldenpath / hg19 / encodeDCC / wgEncodeMapability / ). Using this approach, 361 Mb (13%) of the hg19 reference genome was excluded, including centromere and telomere regions. Short fragments were defined as having a length between 100 and 150 bp, and long fragments were defined as having a length between 151 and 220 bp.

[0056] To account for coverage bias due to GC content of the genome, locally weighted loess smoothing with a span of 3 / 4 was applied to the scatter plot of mean fragment GC versus coverage calculated for each 100 kb bin. This loess regression was performed separately for short and long fragments to account for possible differences in GC effects on coverage in plasma depending on fragment length (see, e.g., Benjamini et al., 2012 Nucleic Acids Res 40:e72). Predictions for short and long coverage explained by GC from the loess model were subtracted to obtain residuals for short and long fragments that were not correlated with GC. Residuals were brought back to the original scale by adding back the genome-wide median short and long estimates of coverage. This procedure was repeated for each sample to account for possible differences in GC effects on coverage between samples. To further reduce feature space and noise, GC-adjusted total coverage in 5 Mb bins was calculated.

[0057] To compare the variability of fragment lengths from healthy subjects to those in patients with cancer, the standard deviation of the short-to-long fragmentation profile for each individual was calculated. Standard deviations in the two groups were compared by Wilcoxon rank-sum test.

[0058] Analysis of chromosome arm copy number alterations To develop arm-level statistics for copy number alterations, we adopted an approach for aneuploidy detection in plasma as described elsewhere (see, e.g., Leary et al., 2012 Sci Transl Med 4:162ra154). This approach divides the genome into non-overlapping 50KB bins, and GC-corrected log2 read depths were obtained for the bins after loess correction with a span of 3 / 4. This loess-based correction was equivalent to the approach outlined above, but was evaluated on a log2 scale to increase robustness to outliers in smaller bins and did not stratify by fragment length. To obtain arm-specific Z-scores for copy number alterations, the GC-adjusted mean read depths (GRs) for each arm were centered and scaled by the mean and standard deviation of the GR scores obtained from an independent set of 50 healthy samples, respectively.

[0059] Analysis of mitochondrial aligned read data from cfDNA Whole genome sequence reads that originally mapped to the mitochondrial genome were extracted from the bam files and realigned to the hg19 reference genome in end-to-end mode using Bowtie2 as described elsewhere (see, e.g., Langmead et al., 2012 Nat Methods 9:357-359). The resulting aligned reads were filtered so that both mates aligned to the mitochondrial genome with a MAPQ of 30 or higher. The number of fragments mapping to the mitochondrial genome was counted and converted to a percentage of the total number of fragments in the original bam file.

[0060] Predictive Models for Cancer Classification To distinguish healthy from cancer patients using fragmentation profiles, a stochastic gradient boosting model was used (gbm; see, e.g., Friedman et al., 2001 Ann Stat 29:1189-1232; and Friedman et al., 2002 Comput Stat Data An 38:367-378). For each sample, GC-corrected total and short fragment coverage for all 504 bins were centered and scaled to have a mean of zero and unit standard deviation. Additional features included Z-scores for each of the 39 autosomal arms and mitochondrial representation (log10 transformed proportion of reads mapping to mitochondria). To estimate the prediction error of this approach, 10-fold cross-validation was used as described elsewhere (see, e.g., Efron et al., 1997 J Am Stat Assoc 92,548-560). Feature selection, performed only on the training data in each cross-validation run, removed bins that were highly correlated (correlation > 0.9) or had near-zero variance. Stochastic gradient boosting machine learning was performed using the R package, gbm package, with parameters n.trees=150, interaction.depth=3, shrinkage=0.1 and n.minobsinside=10. The 10-fold cross-validation procedure was repeated 10 times to average the prediction error resulting from randomization of patients into folds. Sensitivity fixed at 98% and confidence intervals for 95% specificity were obtained from 2000 bootstrap iterations.

[0061] Predictive models for tumor tissue of origin classification For samples correctly classified as cancer patients with 90% specificity (n=174), a separate stochastic gradient boosting model was trained to classify the tissue of origin. To account for the small number of lung samples used for prediction, 18 cfDNA baseline samples from late stage lung cancer patients were included from the monitoring analysis. The performance characteristics of the model were evaluated by 10 iterations of 10-fold cross-validation. The gbm model was trained with the same features as in the cancer classification model. As previously mentioned, features that correlated with each other more than 0.9 or had near-zero variance within each training dataset during cross-validation were removed. Tissue class probabilities were averaged over 10 iterations for each patient, and the class with the highest probability was considered as the predicted tissue.

[0062] Analysis of nucleosomal DNA and cfDNA from human lymphocytes Fragment sizes were analyzed in 5 Mb bins from nuclease-treated lymphocytes as described for whole genome cfDNA analysis. A genome-wide map of nucleosome positions was constructed from nuclease-treated lymphocyte cell lines. This approach identified local biases in the coverage of circulating fragments, indicating regions protected from degradation. A "window positioning score" (WPS) was used to score each base pair in the genome (see, for example, Snyder et al., 2016 Cell 164:57). Using a sliding window of 60 bp centered around each base, WPS was calculated as the number of fragments completely spanning the window minus the number of fragments with only one end present in the window. Since fragments arising from nucleosomes have a median length of 167 bp, a high WPS suggested possible nucleosome positioning. Using a moving median, WPS scores were centered at zero and smoothed using a Kolmogorov-Zurbenko filter (see, e.g., Zurbenko, The spectral analysis of time series. North-Holland series in statistics and probability; Elsevier, New York, NY, 1986). For spans of positive WPS from 50 to 450 bp, nucleosome peaks were defined as pairs of base pairs with WPS above the median in that window. Calculations of nucleosome positions for cfDNA from 30 healthy individuals at 9× sequence coverage were determined similarly to lymphocyte DNA. To ensure that nucleosomes in healthy cfDNA were representative, a nucleosome consensus track was defined consisting of only nucleosomes identified in two or more individuals. The median distance between adjacent nucleosomes was calculated from the consensus track.

[0063] Monte Carlo simulation of detection sensitivity Monte Carlo simulations were used to estimate the probability of detecting molecules with tumor-derived variations. Briefly, 1 million molecules were generated from a multinomial distribution. For simulations with m variations, a wild-type molecule was simulated with probability p, and each of the m tumor variations was simulated with probability (1-p) / m. Then, g*m molecules were randomly drawn with replacement, where g represents the number of genome equivalents in 1 ml of plasma. If a tumor variation was drawn s or more times, the sample was classified as cancer-derived. The simulations were repeated 1000 times to estimate the probability that the cancer index would correctly classify a computer-simulated sample as cancer. The number of tumor variations was varied by powers of two from 1 to 256, and the proportion of tumor-derived molecules was varied from 0.0001% to 1%, with g=2000 and s=5.

[0064] statistical analysis All statistical analyses were performed using R version 3.4.3. To perform the classification of healthy vs. cancer and tissue of origin, the R packages caret (version 6.0-79) and gbm (version 2.1-4) were used. Confidence intervals from the model output were obtained using the pROC (version 1.13) R package (see, for example, Robin et al., 2011 BMC bioinformatics 12:77). Assuming a high prevalence of undiagnosed cancer cases in this population (1 or 2 cases per 100 healthy individuals), a genomic assay with a specificity of 0.95 and a sensitivity of 0.8 would have useful operating characteristics (a positive predictive value of 0.25 and a negative predictive value of approximately 1). Power calculations suggest that analysis of more than 200 cancer patients and an approximately equal number of healthy controls would allow an estimate of sensitivity with a margin of error of 0.06, with the desired specificity of 0.95 or greater.

[0065] Availability of data and code Sequence data used in this study have been deposited in the European Genome-phenome Archive under study accession numbers EGAS00001003611 and EGAS00001002577. Code for the analyses is available at github.com / Cancer-Genomics / delfi_scripts.

[0066] result DELFI allows for the simultaneous analysis of multiple abnormalities in cfDNA through genome-wide analysis of fragmentation patterns. The method is based on low-coverage whole genome sequencing and analysis of isolated cfDNA. Mapped sequences are analyzed in non-overlapping windows covering the genome. Conceptually, the size of the windows can range from thousands to millions of bases, resulting in hundreds to thousands of windows in the genome. 5 Mb windows were used to evaluate cfDNA fragmentation patterns, since even a limited amount of 1-2× genome coverage gives more than 20,000 reads per window. Within each window, the coverage and size distribution of cfDNA fragments were examined. This approach was used to evaluate the variation of whole genome fragmentation profiles in healthy and cancer populations (Table 1; Appendix A). Whole genome patterns from an individual can be compared to a reference population to determine whether the patterns are likely to be healthy or cancerous. These patterns may also suggest the tissue source of the cfDNA, since whole genome profiles reveal positional differences associated with specific tissues that may be missed in the overall fragment size distribution.

[0067] We focused on cfDNA fragmentation size, since it was found that cancer-derived cfDNA molecules can be more variable in size than cfDNA derived from non-cancerous cells. We first examined cfDNA fragments from targeted regions captured and sequenced at high coverage (43,706 total coverage, 8,044 differential coverage) from patients with breast, colorectal, lung or ovarian cancer (Table 1 (Appendix A), Table 2 (Appendix B) and Table 3 (Appendix C)). Analysis of loci containing 165 tumor-specific variations (range 1-7 variations per patient) from 81 patients revealed a mean absolute difference of 6.5 bp (95% CI, 5.4-7.6 bp) between the median mutant cfDNA fragment length and the median wild-type cfDNA fragment length (Figure 3, Table 3 (Appendix C)). The median mutant cfDNA fragment size was The mutated and non-mutated fragments ranged from 30 bases smaller than the corresponding region in the wild-type sequence at position 41,266,124 on chromosome 3 to 47 bases larger at position 108,117,753 on chromosome 11 (Table 3; Appendix C). The GC content was similar for the mutated and non-mutated fragments (Figure 4a), and no correlation existed between GC content and fragment length (Figure 4b). A similar analysis of 44 germline variations obtained from 38 patients identified a median cfDNA size difference of less than 1 bp between the fragment lengths of the different alleles (Figure 5, Table 3 (Appendix C)). In addition, 41 variations associated with clonal hematopoiesis were identified through previous sequence comparison of DNA obtained from plasma, buffy coat, and tumors of the same individuals. Unlike the tumor-derived fragments, there were no significant differences between the fragments with hematopoietic variations and the wild-type fragments (Figure 6, Table 3 (Appendix C)). Taken together, cancer-derived cfDNA fragment lengths were significantly more variable in certain genomic regions compared to non-cancer cfDNA fragments (p<0.001, variance ratio test). It was hypothesized that these differences could be due to alterations in higher-order chromatin structure as well as other genomic and epigenomic abnormalities in cancer, and that cfDNA fragmentation in a location-specific manner could therefore serve as a unique biomarker for cancer detection.

[0068] As targeted sequencing only analyzes a limited number of loci, we investigated a larger-scale whole-genome analysis to detect further abnormalities in cfDNA fragmentation. cfDNA was isolated from approximately 4 ml of plasma from eight lung cancer patients with stage I-III disease as well as 30 healthy individuals (Table 1 (Appendix A), Table 4 (Appendix D) and Table 5 (Appendix E)). A high-efficiency approach was used to convert cfDNA into next-generation sequencing libraries and whole-genome sequencing was performed at approximately 9× coverage (Table 4; Appendix D). Overall, cfDNA fragment lengths were larger in healthy individuals with a median fragment size of 167.3 bp, whereas patients with cancer had a median fragment size of 163.8 (p<0.01, Welch's t-test) (Table 5; Appendix E). To investigate differences in fragment size and coverage in a position-dependent manner across the genome, sequenced fragments were mapped to their genomic origin and fragment length was assessed in 504 windows of 5 Mb in size covering approximately 2.6 Gb of the genome. For each window, the proportion of small cfDNA fragments (length 100-150 bp) to larger cfDNA fragments (151-220 bp) as well as the overall coverage were determined and used to obtain a whole genome fragmentation profile for each sample.

[0069] Healthy individuals had highly similar fragmentation profiles throughout the genome (Figures 7 and 8). To investigate the origin of the fragmentation patterns typically observed in cfDNA, nuclei were isolated from purified lymphocytes of two healthy individuals and treated with DNA nuclease to obtain nucleosomal DNA fragments. Analysis of the observed cfDNA patterns in healthy individuals revealed a high correlation to lymphocyte nucleosomal DNA fragmentation profiles (Figures 7b and 7d) and nucleosomal distances (Figures 7c and 7f). Using Hi-C methods to investigate the three-dimensional structure of the genome (see, e.g., Lieberman-Aiden et al., 2009 Science 326:289-293; and Fortin et al., 2015 Genome Biol 16:180), the median distance between nucleosomes in lymphocytes correlated with open (A) and closed (B) compartments of lymphoblastoid cells (Figure 7c). These analyses suggest that the fragmentation pattern of normal cfDNA is the result of a nucleosomal DNA pattern that largely reflects the chromatin structure of normal blood cells.

[0070] Unlike healthy cfDNA, patients with cancer had multiple distinct genomic differences with increased and decreased fragment sizes in different regions (Figures 7a and 7b).Similar to our observations from targeted analysis, there was also greater fragment length diversity across the genome for patients with cancer compared to healthy individuals.

[0071] To determine whether cfDNA fragment length patterns could be used to distinguish patients with cancer from healthy individuals, we performed genome-wide correlation analysis for the ratio of short to long cfDNA fragments compared to median fragment length profiles calculated from healthy individuals for each sample (Figures 7a, 7b and 7e). While cfDNA fragment profiles were highly consistent among healthy individuals (median correlation of 0.99), the median correlation of genome-wide fragment ratios among cancer patients was 0.84 (0.15 lower, 95% CI 0.07-0.50, p<0.001, Wilcoxon rank sum test; Table 5 (Appendix E)). Similar differences were observed when comparing fragmentation profiles of cancer patients with fragmentation profiles or nucleosome distances in healthy lymphocytes (Figures 7c, 7d and 7f). To account for potential bias in the fragmentation profiles due to GC content, locally weighted smoothing was applied to each sample independently and we found that after this adjustment, differences in fragmentation profiles between healthy individuals and cancer patients remained (median correlation of cancer patients to healthy = 0.83) (Table 5; Appendix E).

[0072] We performed subsampling analyses of whole genome sequence data at 9× coverage from cfDNA of patients with cancer at approximately 2×, 1×, 0.5×, 0.2×, and 0.1× genome coverage, and determined that even at 0.5× genome coverage, altered fragmentation profiles were readily identified (Figure 9). Based on these observations, we performed whole genome sequencing at 1-2× coverage to evaluate whether fragmentation profiles may change during targeted therapy in a manner similar to monitoring sequence variation. We evaluated cfDNA obtained during anti-EGFR or anti-ERBB2 treatment from 19 patients with non-small cell lung cancer, including five with partial radiographic response, eight with stable disease, four with progressive disease, and two with non-measurable disease (Table 6; Appendix F). As shown in Figure 10, the degree of abnormality in the fragmentation profile during treatment was well matched with the level of EGFR or ERBB2 mutant allele frequency determined using targeted sequencing (Spearman correlation of mutant allele frequency to fragmentation profile = 0.74). This correlation is noteworthy because whole genome and mutation-based methods are not statistically related and examine different cfDNA variations that may be suppressed in these patients by previous treatment. Notably, all cases with 6 months or more of progression-free survival showed a decline in ctDNA or had very low levels of ctDNA after the start of treatment, as determined by fragmentation profile, whereas cases with poor clinical outcomes had an increase in ctDNA. These results demonstrate the feasibility of fragmentation analysis to detect the presence of tumor-derived cfDNA and suggest that such analysis may also be useful for quantitative monitoring of cancer patients during treatment.

[0073] Parallel analyses of tumor tissue were performed to examine fragmentation profiles in the context of known copy number alterations in tumor patients. These analyses demonstrated that altered fragmentation profiles were present in regions of the genome that were copy neutral, and that these could be further affected in regions with copy number alterations (Fig. 11a and Fig. 12a). Differences in position-dependent fragmentation patterns could be used to distinguish cancer-derived cfDNA from healthy cfDNA in these regions (Fig. 12a, b), whereas global cfDNA fragment size measurements would have missed such differences (Fig. 12a).

[0074] These analyses were extended to an independent cohort of cancer patients and healthy individuals. Whole genome sequencing at 1-2× coverage was performed on cfDNA from a total of 208 patients with cancer, including breast (n=54), colorectal (n=27), lung (n=12), ovarian (n=28), pancreatic (n=34), gastric (n=27) or cholangiocarcinoma (n=26), and 215 individuals without cancer (Table 1 (Appendix A) and Table 4 (Appendix D)). All cancer patients were treatment naive, and the majority had resectable disease (n=183). After GC adjustment of short and long cfDNA fragment coverage (Figure 13a), we investigated the coverage and size characteristics of fragments in genome-wide windows (Figure 11b, Table 4 (Appendix D) and Table 7 (Appendix G)). Genome-wide correlations of coverage to GC content were limited, and no differences in these correlations were observed between cancer patients and healthy individuals (Figure 13b). Healthy individuals had highly concordant fragmentation profiles, whereas patients with cancer had high variability and reduced correlation to the median healthy profile (Table 7: Appendix G). Among cancer patients, analysis of the most commonly altered fragmentation windows in the genome revealed a median of 60 affected windows across the cancer types analyzed, highlighting the existence of a large amount of position-dependent variation in cfDNA fragmentation among individuals with cancer (Figure 11c).

[0075] To determine whether position-dependent fragmentation changes can be used to detect individuals with cancer, a gradient tree boosting machine learning model was implemented to classify cfDNA as having features of cancer patients or healthy individuals, and performance characteristics of this approach were estimated by ten repeated 10-fold cross-validation (Figures 14 and 15). The machine learning model included GC-adjusted short and long fragment coverage features in windows throughout the genome. A machine learning classifier for copy number changes from chromosome arm-dependent features rather than a single score was also developed (Figure 16a and Table 8 (Appendix H)), and mitochondrial copy number changes were included as they may also be useful in distinguishing cancer from healthy individuals (Figure 16b). With this implementation of DELFI, a score was obtained that can be used to classify patients as healthy or having cancer. 152 of 208 cancer patients were detected (73% sensitivity, 95% CI 67%-79%), while 4 of 215 healthy individuals were misclassified (98% specificity) (Table 9). At the 95% specificity threshold, 80% of patients with cancer were detected (95% CI, 74%-85%), including 79% (145 of 183) of resectable (stage I-III) patients and 82% (18 of 22) of metastatic (stage IV) patients (Table 9). Receiver operator characteristic analysis for detecting patients with cancer had an AUC of 0.94 (95% CI 0.92-0.96), ranging among cancer types from 0.86 for pancreatic cancer to 0.99 or greater for lung and ovarian cancer (Figures 17a and 17b), with an AUC of 0.92 or greater across all stages (Figure 18). DELFI classifier scores did not differ with age, either among cancer patients or healthy individuals (Table 1; Appendix A). [Table 1]

[0076] To assess the contribution of fragment size and coverage, chromosome arm copy number or mitochondrial mapping to the model's predictive accuracy, a repeated 10-fold cross-validation procedure was performed to assess the performance characteristics of these features in isolation. We observed that the fragment coverage feature alone (AUC=0.94) was nearly identical to the classifier of all features combined (AUC=0.94) (Figure 17a). In contrast, the analysis of chromosome copy number changes had a lower performance (AUC=0.88) but was more predictive than copy number changes based on individual scores (AUC=0.78) or mitochondrial mapping (AUC=0.72) (Figure 17a). These results suggest that fragment coverage is the main contributing factor for our classifier. Since all features can be obtained from the same genome sequence data, including all features in the predictive model may contribute in a complementary manner for the detection of patients with cancer.

[0077] Because fragmentation profiles reveal regional differences in fragmentation that may vary between tissues, we used a similar machine learning approach to examine whether cfDNA patterns could identify the primary tissue of these tumors. This approach was found to have an accuracy of 61% (95% CI 53%-67%), including 76% for breast cancer, 44% for cholangiocarcinoma, 71% for colorectal cancer, 67% for gastric cancer, 53% for lung cancer, 48% for ovarian cancer, and 50% for pancreatic cancer (Figure 19, Table 10). When considering that patients with abnormal cfDNA were assigned to one of two primary sites, the accuracy increased to 75% (95% CI 69%-81%) (Table 10). For all tumor types, classification of primary tissue by DELFI was significantly higher than that determined by random assignment (p<0.01, binomial test, Table 10). [Table 2]

[0078] Because cancer-specific sequence variations can be used to identify patients with cancer, we evaluated whether combining DELFI with this approach could increase the sensitivity of cancer detection (Figure 20). Analysis of cfDNA from a subpopulation of treatment-naive cancer patients using both DELFI and targeted sequencing revealed that 82% (103 of 126) of patients had fragmentation profile variations, whereas 66% (83 of 126) had sequence variations. More than 89% of cases with a variant allele frequency of more than 1% were detected by DELFI, whereas for cases with a variant allele frequency of less than 1%, the frequency detected by DELFI was 80%, including cases that could not be detected using targeted sequencing (Table 7; Appendix G). Using these approaches together, the combined detection sensitivity increased to 91% (115 of 126 patients) with a specificity of 98% (Figure 20).

[0079] Taken together, whole genome cfDNA fragmentation profiles differ between cancer patients and healthy individuals. The variability of fragment length and coverage in a position-dependent manner throughout the genome may explain the seemingly contradictory observations of previous analyses of cfDNA at specific loci or of fragment size in general. In patients with cancer, the heterogeneous fragmentation patterns in cfDNA appear to be the result of a mixture of nucleosomal DNA from both blood and neoplastic cells. These studies provide a method for simultaneously analyzing dozens or potentially hundreds of tumor-specific abnormalities from minute amounts of cfDNA, overcoming limitations that have hindered the possibility of more sensitive analysis of cfDNA. DELFI analysis detected a higher percentage of cancer patients than traditional cfDNA analysis methods that have focused on sequence or overall fragmentation size (see, e.g., Phallen et al., 2017 Sci Transl Med 9:eaan2415; Cohen et al., 2018 Science 359:926; Newman et al., 2014 Nat Med 20:548; Bettegowda et al., 2014 Sci Transl Med 6:224ra24; Newman et al., 2016 Nat Biotechnol 34:547). As demonstrated in this example, combining DELFI with analysis of other cfDNA variations can further increase the sensitivity of detection. Since fragmentation profiles appear to be associated with nucleosomal DNA patterns, DELFI can be used to determine the primary source of tumor-derived cfDNA. Identification of the source of circulating tumor DNA in more than half of the patients analyzed can be further improved by including clinical features, other biomarkers including methylation changes and additional diagnostic approaches (Ruibal Morell, 1992 The International journal of biological markers 7:160; Galli et al., 2013 Clinical chemistry and laboratory medicine 51:1369; Sikaris, 2011 Heart, lung & circulation 20:634; Cohen et al., 2018 Science 359:926).Finally, this approach requires only a small amount of whole genome sequencing, without the need for deep sequencing typical of approaches that focus on specific variations. This performance characteristic and the limited amount of sequencing required for DELFI suggest that our approach is broadly applicable for the screening and management of patients with cancer.

[0080] These results demonstrate that the genome-wide cfDNA fragmentation profiles differ between cancer patients and healthy individuals. Therefore, the cfDNA fragmentation profiles may have important implications for further research and application of non-invasive approaches to detect human cancer. Other embodiments

[0081] While the invention has been described in conjunction with its detailed description, it should be understood that the foregoing description is intended to be illustrative and not limiting of the scope of the invention, which is defined by the appended claims. Other aspects, advantages, and modifications are within the scope of the following claims. [Table 3-1] [Table 3-2] [Table 3-3] [Table 3-4] [Table 3-5] [Table 3-6] [Table 4-1] [Table 4-2]

Table 4-3

Table 4-4

Table 4-5

Table 4-6

Table 5-1

Table 5-2

Table 5-3

Table 5-4

Table 5-5

Table 5-6

Table 5-7

Table 5-8

Table 6-1

Table 6-2

Table 6-3

Table 6-4

Table 6-5

Table 6-6

Table 6-7

Table 6-8

Table 7

Table 8-1

Table 8-2

Table 9-1

Table 9-2

Table 9-3

Table 9-4

Table 9-5

Table 9-6

Claims

1. A method for identifying a subject as having cancer, comprising: determining a cell-free DNA (cfDNA) fragmentation profile of sequenced fragments in a sample obtained from said subject, wherein the sequenced fragments were obtained through whole genome sequencing (WGS); analyzing said windows of mapped sequences to determine a cfDNA fragmentation profile; analyzing the cfDNA fragmentation profile against a reference cfDNA fragmentation profile obtained from a healthy subject, where the cfDNA fragmentation profile comprises a ratio of small cfDNA fragments to large cfDNA fragments; identifying a cfDNA fragmentation profile obtained from the subject as being more variable than a reference cfDNA fragmentation profile, where increased variability in the fragmentation profile obtained from the subject is indicative that the subject has cancer; (excluding medical procedures on humans).

2. The method according to claim 1, wherein the cancer is selected from the group consisting of colorectal cancer, lung cancer, breast cancer, gastric cancer, pancreatic cancer, bile duct cancer and ovarian cancer.

3. The method of claim 1, wherein the reference cfDNA fragmentation profile is generated by determining a cfDNA fragmentation profile in a sample obtained from the healthy subject.

4. The method described in claim 1, wherein the reference DNA fragmentation pattern is a reference nucleosome cfDNA fragmentation profile.

5. The method of claim 1, wherein determining the cfDNA fragmentation profile includes determining a median fragment size, and the median fragment size of the cfDNA fragmentation profile is shorter than the median fragment size of the reference cfDNA fragmentation profile.

6. The method of claim 1, wherein determining the cfDNA fragmentation profile includes determining a fragment size distribution, and the fragment size distribution of the cfDNA fragmentation profile differs by at least 10 nucleotides compared to the fragment size distribution of the reference cfDNA fragmentation profile.

7. The method of claim 1, wherein the cfDNA fragmentation profile comprises a ratio of small cfDNA fragments to large cfDNA fragments in a window of mapped sequences, the small cfDNA fragments being approximately 100 base pairs (bp) to 150 bp in length and the large cfDNA fragments being approximately 151 bp to 220 bp in length.

8. The method of claim 1, wherein the cfDNA fragmentation profile comprises small cfDNA fragments in a genome-spanning window.

9. The method of claim 1, wherein the cfDNA fragmentation profile includes large cfDNA fragments in a genome-spanning window.

10. The method of claim 1, wherein the cfDNA fragmentation profile comprises small and large cfDNA fragments in a genome-spanning window.

11. The method of claim 1, wherein analyzing the cfDNA fragmentation profile against a reference cfDNA fragmentation profile obtained from a healthy subject comprises analyzing the cfDNA fragmentation profile relative to the reference cfDNA fragmentation profile across a subgenomic interval.

12. The method of claim 8, wherein the genomic coverage of the mapped sequences is from about 2x, 1x, 0.5x, 0.2x or 0.1x.

13. The method of claim 1, wherein each window is thousands to millions of bases in length.

14. The method described in claim 8, wherein the correlation of fragment ratios between small cfDNA fragments and large cfDNA fragments in a cfDNA fragmentation profile of a subject with cancer compared to a reference cfDNA fragmentation profile is lower than the correlation of fragment ratios between small cfDNA fragments and large cfDNA fragments in a healthy subject compared to a reference cfDNA fragmentation profile.

15. The method of claim 1, wherein the sample is selected from the group consisting of blood, serum, plasma, amniotic membrane, tissue, urine, cerebrospinal fluid, saliva, sputum, bronchoalveolar lavage fluid, bile, lymphatic fluid, cyst fluid, stool, ascites, Papanicolaou smear, breast milk and exhaled breath condensate.

16. The method of claim 1, wherein the cell-free DNA (cfDNA) fragmentation profile is determined during and / or after the treatment.