Methods for monitoring cancer using fragmentation profiles - Patents.com

Abstract

The present disclosure provides methods and systems utilizing analysis of a cell-free DNA (cfDNA) fragmentation profile in a sample taken from a patient, determining a short to long fragment ratio and fragment size distribution from the fragmentation profile, determining an information score based on the short to long fragment ratio in the sample correlated with ratios from a sample from a healthy patient, and determining a monitoring score for the sample based on the fragmentation score, the information score, and the model weights by a machine learning model, the monitoring score indicative of a level of tumor-derived nucleic acid in the cfDNA of the sample to detect, monitor, diagnose, or predict a cancer state, determine the likelihood of the presence of cancer, and administer treatment to the patient.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to U.S. Provisional Patent Application No. 63 / 320,906, filed March 17, 2022, the entire contents of which are incorporated herein by reference.

[0002] FIELD OF THEINVENTION The present invention relates generally to genetic analysis, and more specifically to methods and systems for analyzing cell-free DNA (cfDNA) fragments to predict the fraction of tumor-derived DNA modules (ctDNA load) and detect cancer in a subject. [Background technology]

[0003] Background information Much of the morbidity and mortality of cancer in humans worldwide is due to late diagnosis of these diseases, which are less effective in treatment. In addition, after cancer is diagnosed, it is difficult to predict the progression of cancer and the patient's response to treatment, further reducing the success rate of cancer treatment. Unfortunately, clinically proven biomarkers that can be widely used to diagnose and effectively treat cancer patients are not widely available.

[0004] The fraction of tumor-derived DNA molecules in plasma (ctDNA burden) is a useful tool to describe the overall tumor burden in cancer patients. Previous studies have shown that the ctDNA burden in an individual patient is influenced by a variety of factors, including the tumor's tissue of origin and stage, as well as angiogenesis and perfusion. Thus, patients with advanced stage cancer have a higher ctDNA burden than those with early stage cancer. Similarly, patients with tissue cancers that have fast cell turnover and direct access to the bloodstream (such as colorectal cancer) often have a higher ctDNA burden than those with tumors that are less vascular and grow more slowly. ctDNA burden can change over time as tumors are exposed to treatment and die (decreasing), and then acquire resistance mechanisms to treatment and grow (increasing). However, previous studies have lacked the ability to efficiently predict ctDNA burden and utilize the predicted ctDNA burden as a tool for cancer diagnosis, predicting disease progression and treatment response, and determining overall survival in patients diagnosed with cancer. Summary of the Invention

[0005] The present disclosure provides methods and systems for monitoring cancer progression using analysis of cfDNA in samples taken from subjects and predicting overall survival of the subject by scoring a cfDNA fragmentation profile obtained from analysis of cfDNA fragments. The scoring methodology generates features that can be used to train machine learning models to predict biomarkers that can be used to monitor cancer progression, assess patient response to treatment, and predict overall survival of a subject.

[0006] Thus, in one embodiment, the invention provides a method for monitoring cancer, the method comprising: Determining a cell-free DNA (cfDNA) fragmentation profile of a sample from the subject; calculating a score based on the cfDNA fragmentation profile, said score indicating a likelihood of the presence of cancer in the subject; determining a short to long fragment ratio and a fragment size distribution from the fragmentation profile; calculating a diversity score based on the short fragment to long fragment ratio; determining a set of model weights based on the fragment size distribution; training a machine learning model using a set of features extracted from multiple fragmentation profiles of multiple subjects; determining, by the machine learning model, a supervision score for the sample based on the supervision score, the information score, and the model weight, the supervision score indicative of a level of tumor-derived nucleic acid in cfDNA of the sample; Includes.

[0007] In some embodiments, the cfDNA fragmentation profile is determined by obtaining and isolating cfDNA fragments from the subject, sequencing the cfDNA fragments to obtain sequenced fragments, mapping the sequenced fragments to the genome to obtain windows of mapped sequences, and analyzing the windows of mapped sequences to determine lengths of the cfDNA fragments and generate the cfDNA fragmentation profile.

[0008] In another embodiment, the invention provides a method for determining at least one of overall survival, progression free survival, or time to progression in a subject with cancer, the method comprising: determining a cell-free DNA (cfDNA) fragmentation profile of a sample from the subject; calculating a score based on the cfDNA fragmentation profile, said score indicating a likelihood of the presence of cancer in the subject; determining a short to long fragment ratio and a fragment size distribution from the fragmentation profile; calculating an information content score based on the short fragment to long fragment ratio; determining a set of model weights based on the fragment size distribution; training a machine learning model using a set of features extracted from the fragmentation profiles of multiple subjects; determining, by the machine learning model, a monitoring score for the sample based on the score, the information score, and the model weight, the monitoring score being indicative of a level of tumor-derived nucleic acid in cfDNA of the sample, thereby indicating a likelihood of cancer progression in the subject; determining overall survival of the subject by determining at least one of the subject's overall survival probability, progression free survival, or time to progression based on the monitoring score; In yet another aspect, the invention provides a method of treating a subject having cancer.

[0009] In yet another embodiment, the present invention provides a system for monitoring cancer in a subject, the system comprising: Memory, one or more processors coupled to the memory; wherein the one or more processors provide the computer system with: determining a cell-free DNA (cfDNA) fragmentation profile of a sample from the subject; calculating a score based on the cfDNA fragmentation profile, the score indicating a likelihood of the presence of cancer in the subject; and determining a short to long fragment ratio and a fragment size distribution from the fragmentation profile; calculating an information content score based on said short fragment to long fragment ratio; determining a set of model weights based on the fragment size distribution; training a machine learning model using a set of features extracted from the fragmentation profiles of multiple subjects; determining, by the machine learning model, a monitoring score for the sample based on the score, the information score, and the model weight, the monitoring score being indicative of a level of tumor-derived nucleic acid in cfDNA of the sample, thereby indicative of a likelihood of cancer progression in the subject; The method is configured to perform an operation to cause the

[0010] In another embodiment, the present invention provides a non-transitory computer readable storage medium encoded with a computer program, the computer program comprising instructions that, when executed by one or more processors, cause the one or more processors to perform operations for implementing the method of the present invention.

[0011] In yet another embodiment, the present invention provides a computing system comprising a memory and one or more processors coupled to the memory, the one or more processors configured to execute operations implementing the method of the present invention.

[0012] In yet another embodiment, the present invention provides a system for genetic analysis and cancer assessment, comprising: (a) a sequencer configured to generate a whole genome sequencing (WGS) dataset of a sample; and (b) a non-transitory computer readable storage medium and / or a computer system of the present invention. [Brief description of the drawings]

[0013] [Figure 1]

[0023] Figure 1 is a block diagram illustrating a process for training a machine learning model to generate the DELFI Surveillance Score (DMS). The DMS can be used to monitor cancer, determine overall survival, progression-free survival, and time to progression in patients diagnosed with cancer. The DMS can also be used to diagnose cancer in patients and determine which cancer treatment to administer to the patient.

[0014] [Diagram 2]1 is a graphical plot showing a comparison of observed MAF trajectories with trajectories determined based on DMS values.

[0015] [Diagram 3] 1 is a graphical plot showing a comparison of observed MAF trajectories with trajectories determined based on DMS values, regardless of time point or which patient a given sample is from.

[0016] [Figure 4] 1 is a graph plot showing a comparison of observed MAF trajectories to trajectories determined based on DMS values, where the cohort of patients with observed MAF trajectories have a different type of cancer than the cohort of patients for training the model that generates the DMS values.

[0017] [Diagram 5] 1 is a graph plot showing predicted DMS progression free survival data for two different cohorts with different types of cancer.

[0018] [Figure 6] 1 is a graph plot showing overall survival for a cohort of cancer patients by predicted DMS at pre-treatment, the graph plot also showing overall survival for a cohort of cancer patients who underwent incomplete and complete resection.

[0019] [Figure 7] 1 is a graph plot showing progression-free survival for a cohort of cancer patients with clonal variant MAF suppressed to 0% at first post-treatment evaluation. The graph plot also shows a comparison of predicted DMS and MAF by ddPCR.

[0020] [Figure 8] 8 is an example of a computer 800 that can be used to implement the training algorithm shown in FIG. 1 and generate DMS values.

[0021] [Figure 9]Figures 9A-9B are study design and patient flow diagrams. Figure 9A is the CAIRO5 study design flow chart. Once eligibility was confirmed, including unresectable status of liver metastases as defined by the central panel, KRAS (exons 2, 3, and 4), NRAS (exons 2 and 3), and BRAF mutation status were assessed in tissue samples. Patients with RAS / BRAF mutant tumors were randomized to doublet chemotherapy + bevacizumab (group 1) or triplet chemotherapy + bevacizumab (group 2). Patients with RAS / BRAF wild-type tumors were randomized to doublet chemotherapy + bevacizumab (group 3) or panitumumab (group 4). In this translational study, blood collected from patients in group 1 (mutated group) and group 3 (wild-type group) was processed and analyzed. Figure 9B is a patient flow diagram. The number of patients and samples included in the study and reasons for exclusion are shown.

[0022] [Figure 10]Figures 10A-D show that DELFI Tumor Fraction (DELFI-TF) is a mutation-independent approach to monitor metastatic disease. Figure 10A shows that patients with treatment-naïve inoperable liver-only mCRC enrolled in the CAIRO5 Phase III trial had their tumors tested for hotspot mutations in KRAS, NRAS, and BRAF. Blood samples were taken at baseline, on treatment, and at disease progression or last follow-up. Patients with driver mutations were monitored by ddPCR and DELFI-TF assays. Patients with wild-type KRAS, NRAS, and BRAF were monitored with DELFI-TF alone. Figure 10B shows that plasma aliquots from patients with tissue-confirmed RAS / BRAF mutant mCRC were used for cfDNA isolation. From each time point, duplicate cfDNA samples were used for ddPCR and low-coverage WGS. WGS fragment sequencing statistics were calculated for each sample at a given time point. A Bayesian probability model was trained on MAFs interpreted by ddPCR readings of tumor-specific RAS / BRAF variants in all longitudinal cfDNA samples to generate DELFI-TF scores. Figure 10C is a heatmap representation of genomic features, showing the deviations of cfDNA fragment ratios and chromosome arm-level z-scores across baseline and treatment timepoints for 128 patients, along with DELFI-TF scores and clinical and demographic characteristics. Figure 10D shows that cfDNA genome-wide fragmentation profiles in 504 non-overlapping 5Mb genomic regions at baseline and near the second imaging assessment by RECIST1.1 show significant heterogeneity at baseline and for patients who showed disease progression compared to those who experienced stable disease or radiological response after initiating first-line systemic therapy.

[0023] [Figure 11]Figures 11A-11F show how DELFI-TF predicts tumor fraction in blood of patients with progressive disease receiving systemic therapy. Figure 11A shows that mCRC patients show a wider range of DELFI-TF values ​​at baseline compared to non-cancer controls. The 95% CI upper limit in non-cancer controls (grey dotted line) represents the blank DELFI-TF limit value (0.006). Figure 11B shows that DELFI-TF strongly correlates with MAF measured by ddPCR across all study time points (n=692, Pearson correlation, r=0.85, p=3.9e-89). Plasma time points with undetectable MAF (n=60, red) showed a wider range of DELFI-TF values. Figure 11C shows that cfDNA fragmentation profiles in wild-type mCRC patients (bottom) show abnormalities in short-to-long fragment ratios even in the context of tumor copy-neutral regions across 100 kb bins in matched tissue samples (top). Figure 11D shows that plasma tumor fractions assessed by DELFI-TF (blue) and RAS / BRAF MAF (orange) values ​​correlate with cfDNA copy number changes in genomic regions with genes frequently deleted (MBD1) or amplified (PLGC1) in colorectal cancer. Figures 11E and 11F show that for relative coverage at TSS positions of a group of 890 genes highly expressed in colorectal cancer, the valleys in baseline samples (brown) are significantly deeper (Wilcoxon test, p<0.001) than in treatment samples (purple), indicating, on average, lower tumor fractions at time points after treatment initiation.

[0024] [Figure 12]Figures 12A-C show cfDNA tumor fractions assessed by whole genome sequencing approaches. Figure 12A shows that patients with metastatic colorectal cancer show a wider range of ichorCNA values ​​at baseline compared to non-cancer controls. The upper 95% confidence interval in non-cancer controls is 0.017 (grey dotted line). Figure 12B shows that DELFI-TF correlates with tumor fractions measured by ichorCNA at time points when mutant allele fractions of RAS / BRAF assessed by ddPCR within the mutant group of the study are undetectable (Spearman correlation, rho=0.54, p<0.001). Figure 12C shows the relative coverage of a group of genes highly expressed in colorectal cancer at transcription start site (TSS) positions at baseline (brown), during systemic treatment (purple), after metastasectomy (red), and at disease recurrence (black) in patient 65. Variation in valley depth reflects dynamic changes in the plasma tumor fraction over longitudinal time points.

[0025] [Figure 13]Figures 13A-G show that DELFI-TF is a non-invasive biomarker for metastatic disease burden and systemic treatment response. Figure 13A shows that baseline DELFI-TF (blue) and RAS / BRAF MAF (orange) values ​​are moderately correlated with SLD of liver target lesions on pretreatment imaging scans (DELFI-TF Spearman correlation, rho=0.49, p<0.001; MAF Spearman correlation, rho=0.48, p<0.001). Figure 13B shows that baseline DELFI-TF (blue) and RAS / BRAF MAF (orange) values ​​are not significantly correlated with baseline levels of carcinoembryonic antigen (DELFI-TF Spearman correlation, rho=0.10, p=0.427; MAF Spearman correlation, rho=0.15, p<0.236). FIG. 13C shows that baseline DELFI-TF and MAF values ​​were significantly lower in patients who later had a confirmed partial response (PR) or complete response (CR) by two consecutive RECIST 1.1 measurements (DELFI-TF, Wilcoxon test, p=0.048; MAF, Wilcoxon test, p=0.017). FIG. 13D shows that baseline DELFI-TF and MAF values ​​were significantly lower in patients who underwent complete resection (orange) or partial resection (green) after systemic treatment (DELFI-TF, Kruskal-Wallis test, p=0.037; MAF, Kruskal-Wallis test, p=0.011). FIG. 13E shows that colorectal cancer patients with metachronous metastases (gray) exhibited a lower tumor fraction assessed by DELFI-TF at baseline compared to patients who presented with synchronous metastases (green) (Wilcoxon test, p<0.001). FIG. 13F shows that liver metastases are highlighted by blue circles in longitudinal imaging scans (top). cfDNA tumor fraction dynamics (DELFI-TF, MAF) and SLD values ​​for study patient 11 are shown (bottom). Treatment is indicated by shaded bars. Purple dotted line represents time of primary tumor resection several weeks after liver metastasis removal. DELFI-TF and ddPCR MAF of KRAS G12D mutation accurately track disease burden dynamics before and after complete resection.FIG. 13G shows that patients who eventually experienced disease progression (top) more frequently exhibited increased DELFI-TF values ​​at the longitudinal time points compared to patients who never exhibited disease progression (bottom).

[0026] [Figure 14] Figures 14A-D show DELFI-TF and clinical characteristics of colorectal cancer. Figure 14A shows that tumor fractions assessed by DELFI-TF and MAF at baseline are comparable between colon cancers from the left (brown) or right (beige) side (DELFI-TF, Wilcoxon test, p=0.329; MAF, Wilcoxon test, p=0.515). Figure 14B shows that tumor fractions assessed by DELFI-TF at baseline are comparable between colon cancers with driver RAS / BRAF mutations (red) or wild-type mutations (pink) (Wilcoxon test, p=1). FIG. 14C shows that tumor fraction assessed by DELFI-TF and MAF at baseline was higher in patients who eventually experienced disease progression (green) compared to patients who never experienced disease progression (red) (DELFI-TF, Wilcoxon test, p=0.03; MAF, Wilcoxon test, p=0.02). Sum of longest diameters at baseline did not differ between progressors (green) and non-progressors (red) (Wilcoxon test, p=0.94). FIG. 14D shows a waterfall plot showing objective clinical response by change in sum of longest diameters (SLD). Red is DELFI-TF slope above the median. Cyan is DELFI-TF slope below the median.

[0027] [Figure 15A] Figures 15A-B show DELFI-TF dynamics in study patients: Figure 15A shows the RAS / BRAF wild type group. [Figure 15B] Figures 15A-B show DELFI-TF dynamics in the study patients. Figure 15B shows the RAS / BRAF mutant group.

[0028] [Figure 16]We show that dynamic changes in DELFI-TF are associated with longitudinal clinical outcomes. DELFI-TF slopes were calculated for all available time points for patients (n=81) with at least baseline and one blood draw within 60 days after disease progression. On the left, DELFI-TF slopes are colored based on outcomes below (cyan) or above (red) the median DELFI-TF slope. On the right, a swimmer plot encompassing RECIST1.1, liquid biopsy, surgery, and patient death events ordered according to study duration in weeks. Each bar represents the interval between study enrollment and death or last follow-up. Bar segments are colored according to RECIST1.1 readings.

[0029] [Figure 17] Figures 17A-D show that baseline DELFI-TF and DELFI-TF slope correlate with progression-free survival (PFS) and overall survival (OS). Figure 17A shows Kaplan-Meier curves of PFS according to baseline DELFI-TF values ​​below (orange) or above (blue) the lower quartile among patients with RAS / BRAF mutant and wild-type metastatic colorectal cancer (n=128) (log-rank p=0.015). Figure 17B shows Kaplan-Meier curves of PFS according to DELFI-TF slope below (orange) or above (blue) the median among patients who had at least one blood draw within 60 days after disease progression (n=81) (log-rank p<0.001). Figure 17C shows Kaplan-Meier curves of PFS according to DELFI-TF slope below (orange) or above (blue) the median among patients who experienced imaging response or stable disease (sustained clinical benefit) lasting >12 months (n=42) (log-rank p<0.001). Figure 17D shows Kaplan-Meier curves of OS according to DELFI-TF slope below (orange) or above (blue) the median among patients who had at least one blood draw within 60 days after disease progression (n=81) (log-rank p<0.001).

[0030] [Figure 18]Figures 18A-D show imaging and plasma biomarkers of survival outcomes in patients with metastatic colorectal cancer. Figure 18A shows that Kaplan-Meier curves of progression-free survival (PFS) according to imaging response assessed by RECIST 1.1 show no survival differences between patients with partial response (PR) (orange) or stable disease (SD) (blue). PD stands for progressive disease. Figure 18B shows Kaplan-Meier curves of PFS according to baseline MAF of RAS / BRAF below (orange) or above (blue) the lower quartile among patients with RAS / BRAF mutant metastatic colorectal cancer (n=65) (log-rank p=0.003). Figure 18C shows Kaplan-Meier curves of PFS according to baseline carcinoembryonic antigen values ​​below (orange) or above (blue) the lower quartile among patients with metastatic colorectal cancer (n=127) (log-rank p=0.067). Figure 18D shows Kaplan-Meier curves of overall survival (OS) according to surgical status and DELFI-TF slope among patients who had at least one blood draw within 60 days after disease progression (n=81) (log-rank p<0.001). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0031] Detailed Description of the Invention Described herein is a non-invasive method for monitoring cancer and predicting overall survival, progression-free survival, and time to progression in subjects with cancer. cfDNA in blood can provide a non-invasive method for monitoring disease in cancer patients. As shown herein, DNA assessment of fragments for early interruption (DELFI) was used to evaluate genome-wide fragmentation patterns of cfDNA in patients with various types of cancer and in healthy individuals. The evaluation of cfDNA involved a scoring methodology. A defined score (also referred to herein as "DELFI monitoring score") was determined based on the cfDNA fragmentation profile obtained with the cfDNA fragments of a given patient sample. Evaluating cfDNA by the methodology described herein can also provide an approach to monitor cancer, which can increase the probability of successful treatment and improved outcomes in patients with cancer.

[0032] Before describing the present compositions and methods, it should be understood that the invention is not limited to the particular methods and systems described, as such methods and systems may vary, and the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting, since the scope of the invention will be limited only by the appended claims.

[0033] As used in this specification and the appended claims, the singular forms "a," "an," and "the" include the plural unless the context clearly indicates otherwise. Thus, for example, reference to "the method" includes one or more methods, and / or steps of the type described herein that will become apparent to those skilled in the art upon reading this disclosure or otherwise.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Although any methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, the preferred methods and materials are described below.

[0035] The present disclosure provides an innovative method and system for analyzing cfDNA to monitor, detect or evaluate cancer.Previous studies have shown that, on average, individuals without cancer have longer cfDNA fragments (average size 167.09bp), while individuals with cancer have shorter cfDNA fragments (average size 164.88bp).The methodology described herein allows for simultaneous analysis of multiple abnormalities in cfDNA by genome-wide analysis of cfDNA fragmentation patterns.

[0036] Thus, in one embodiment, the invention provides a method of monitoring cancer in a subject, the method comprising: Determining a cell-free DNA (cfDNA) fragmentation profile of a sample from the subject; calculating a score based on the cfDNA fragmentation profile, said score indicating a likelihood of the presence of cancer in the subject; determining a short to long fragment ratio and a fragment size distribution from the fragmentation profile; calculating an information content score based on the short fragment to long fragment ratio; determining a set of model weights based on the fragment size distribution; training a machine learning model using a set of features extracted from multiple fragmentation profiles of multiple subjects; determining, by the machine learning model, a supervision score for the sample based on the supervision score, the information score, and the model weight, the supervision score indicative of a level of tumor-derived nucleic acid in cfDNA of the sample; Includes.

[0037] In an embodiment, the present invention provides a method of treating a subject having cancer, the method comprising: a) detecting cancer in said subject by the methodology of the present invention or determining the overall survival rate of said subject by the methodology of the present invention; b) treating the subject by administering a cancer treatment to the subject. Includes.

[0038] In another embodiment, the present invention provides a method for monitoring cancer in a subject, the method comprising: a) detecting cancer in said subject by the methodology of the present invention or determining overall survival of said subject by the methodology of the present invention; b) administering to said subject a cancer treatment; c) monitoring the cancer in said subject by determining the overall survival rate of said subject by the methodology of the present invention after said cancer treatment is administered. Includes.

[0039] The methodology described herein utilizes cfDNA fragmentation profile. As used herein, the term "fragmentation profile" refers to. In some embodiments, determining the cfDNA fragmentation profile in a mammal can be used to identify the mammal as having cancer. For example, the cfDNA fragments obtained from a mammal (e.g., from a sample taken from the mammal) can be subjected to low-coverage whole genome sequencing, and the sequenced fragments can be mapped to the genome (e.g., non-overlapping windows) and evaluated to determine the cfDNA fragmentation profile. 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).

[0040] The cfDNA fragmentation profile may include one or more cfDNA fragmentation patterns. The cfDNA fragmentation pattern may include any suitable cfDNA fragmentation pattern. Examples of cfDNA fragmentation patterns include, without limitation, fragment size density, median fragment size, fragment size distribution, small cfDNA fragment to large cfDNA fragment ratio, and cfDNA fragment coverage. In some embodiments, the cfDNA fragmentation profile may be a genome-wide cfDNA profile (e.g., a genome-wide cfDNA profile in a genome-wide window). In some embodiments, the cfDNA fragmentation profile may be a targeted region profile. The targeted region may be any suitable portion of the genome (e.g., a chromosomal region). Examples of chromosomal regions for which a cfDNA fragmentation profile can be determined as described herein include, without limitation, a portion of a chromosome (e.g., a portion of 2q, 4p, 5p, 6q, 7p, 8q, 9q, 10q, 11q, 12q, and / or 14q) and a chromosomal arm (e.g., a chromosomal arm of 8q, 13q, 11q, and / or 3p). In some cases, a cfDNA fragmentation profile may include two or more targeted region profiles.

[0041] In various embodiments, the cfDNA obtained from the sample is isolated, and fragments of a certain size range are used for analysis.In some embodiments, the analysis excludes fragment sizes less than about 10, 50, 100, or 105bp, and more than about 220, 250, 300, 350bp, or more.In some embodiments, the analysis excludes fragment sizes less than 105bp and more than 170bp.In some embodiments, the analysis excludes fragment sizes less than about 230, 240, 250, 260bp, and more than about 420, 430, 440, 450bp, or more.In some embodiments, the analysis excludes fragment sizes less than 260bp and more than 440bp.

[0042] In some embodiments, the cfDNA fragmentation profile can be determined by processing a sample from a subject containing cfDNA fragments to create a sequencing library, subjecting the sequencing library to low-coverage whole genome sequencing to obtain sequenced fragments, mapping the sequenced fragments to the genome to obtain mapped sequence windows, and analyzing the mapped sequence windows to determine the lengths of the cfDNA fragments.

[0043] In some embodiments, a cfDNA fragmentation profile can be determined by obtaining and isolating cfDNA fragments from a subject, sequencing the cfDNA fragments to obtain sequenced fragments, mapping the sequenced fragments to the genome to obtain mapped sequence windows, and analyzing the mapped sequence windows to determine the lengths of the cfDNA fragments and generate a cfDNA fragmentation profile.

[0044] The methodology of the present invention is based on low-coverage whole genome sequencing and analysis of isolated cfDNA. In one embodiment, the data used to develop the methodology of the present invention is based on shallow whole genome sequence data (1-2x coverage).

[0045] In some embodiments, the mapped sequences are analyzed in non-overlapping windows covering the genome. Conceptually, the size range of the window can be from thousands to millions of bases, resulting in hundreds to thousands of windows in the genome. 5Mb windows were used to evaluate cfDNA fragmentation patterns, since even a limited amount of 1-2x genome coverage provides more than 20,000 reads per window. Within each window, the coverage and size distribution of cfDNA fragments were examined. In some embodiments, genome-wide patterns from an individual can be compared to a reference population to determine whether the pattern is healthy or cancerous.

[0046] In certain embodiments, the mapped sequences include tens to thousands of genomic windows, e.g., 10, 50, 100 to 1,000, 5,000, 10,000 or more windows. Such windows can be non-overlapping or overlapping and include about one million, 2 million, 3 million, 4 million, 5 million, 6 million, 7 million, 8 million, 9 million, or 10 million base pairs.

[0047] In various embodiments, cfDNA fragmentation profile is determined within each window.Therefore, the present invention provides a method for determining the cfDNA fragmentation profile in a subject (e.g., a sample taken from a subject).

[0048] In some embodiments, cfDNA fragmentation profile can be used to identify cfDNA fragment length changes (e.g., modifications).Modification can be genome-wide modifications or modifications of one or more target regions / locuses.Target regions can be any region that contains one or more cancer-specific modifications. In some embodiments, the cfDNA fragmentation profile can be used to identify (e.g., simultaneously identify) about 10 modifications to about 500 modifications (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 modifications).

[0049] In various embodiments, the cfDNA fragmentation profile may include a cfDNA fragment size pattern. The cfDNA fragments may be of any suitable size. For example, in some embodiments, the cfDNA fragments may be from about 50 base pairs (bp) to about 400 bp in length. As described herein, a subject with cancer may have a cfDNA fragment size pattern that contains a median cfDNA fragment size that is shorter than the median cfDNA fragment size of a healthy subject. A healthy subject (e.g., a subject without cancer) may have a cfDNA fragment size with a median cfDNA fragment size of about 166.6 bp to about 167.2 bp (e.g., about 166.9 bp). In some embodiments, a subject with cancer may 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 subject. For example, a subject with cancer can have a cfDNA fragment size with a median cfDNA fragment size of about 164.11 bp to about 165.92 bp (e.g., about 165.02 bp).

[0050] In some embodiments, the dinucleosomal cfDNA fragments can be about 230 base pairs (bp) to about 450 bp in length. As described herein, subjects with cancer can have a dinucleosomal cfDNA fragment size pattern that contains a median dinucleosomal cfDNA fragment size that is shorter than the median dinucleosomal cfDNA fragment size in healthy subjects. In some embodiments, on average, subjects without cancer have longer cfDNA fragments in the dinucleosomal range (average size 334.75 bp), while subjects with cancer have shorter dinucleosomal cfDNA fragments (average size 329.6 bp). Thus, healthy subjects (e.g., subjects without cancer) can have a dinucleosomal cfDNA fragment size with a median cfDNA fragment size of about 334.75 bp. In some embodiments, subjects with cancer can have a dinucleosomal cfDNA fragment size that is shorter than the dinucleosomal cfDNA fragment size in healthy subjects. For example, a subject with cancer can have a dinucleosomal cfDNA fragment size with a median cfDNA fragment size of about 329.6 bp.

[0051] The cfDNA fragmentation profile may include a cfDNA fragment size distribution. As described herein, a subject with cancer may have a cfDNA fragment size distribution that is more variable than that in a healthy subject. In some embodiments, the size distribution may be within a target region. A healthy subject (e.g., a subject without cancer) may have a target region cfDNA fragment size distribution of about 1 or less than about 1. In some embodiments, a subject with cancer may have a target region cfDNA fragment size distribution that is longer (e.g., 10, 15, 20, 25, 30, 35, 40, 45, 50 bp or more longer, or any number of base pairs between these numbers) than that in a healthy subject. In some embodiments, a subject with cancer may have a target region cfDNA fragment size distribution that is shorter (e.g., 10, 15, 20, 25, 30, 35, 40, 45, 50 bp or more shorter, or any number of base pairs between these numbers) than that in a healthy subject. In some embodiments, a subject with cancer may have a target region cfDNA fragment size distribution that is about 47bp smaller to about 30bp longer than the target region cfDNA fragment size distribution in healthy subjects.In some embodiments, a subject with cancer may have a target region cfDNA fragment size distribution that has an average difference in cfDNA fragment length of 10, 11, 12, 13, 14, 15, 15, 17, 18, 19, 20bp or more.For example, a subject with cancer may have a target region cfDNA fragment size distribution that has an average difference in cfDNA fragment length of about 13bp.In some embodiments, the size distribution may be a genome-wide size distribution.

[0052] The cfDNA fragmentation profile may include a ratio of small cfDNA fragments to large cfDNA fragments, and a correlation of the fragment ratio with a reference fragment ratio. As used herein, for a ratio of small cfDNA fragments to large cfDNA fragments, the small cfDNA fragments may be about 100 bp to about 150 bp in length. As used herein, for a ratio of small cfDNA fragments to large cfDNA fragments, the large cfDNA fragments may be about 151 bp to 220 bp in length. As described herein, a subject with cancer may have a fragment ratio correlation (e.g., a correlation of a cfDNA fragment ratio with a reference DNA fragment ratio, such as a DNA fragment ratio from one or more healthy subjects) that is lower (e.g., 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) than in a healthy subject. Healthy subjects (e.g., subjects without cancer) may have a fragment ratio correlation (e.g., a correlation between cfDNA fragment ratios and a reference DNA fragment ratio, such as a DNA fragment ratio from one or more healthy subjects) of about 1 (e.g., about 0.96). In some embodiments, subjects with cancer may have a fragment ratio correlation (e.g., a correlation between cfDNA fragment ratios and a reference DNA fragment ratio, such as a DNA fragment ratio from one or more healthy subjects) that is, on average, about 0.19 to about 0.30 (e.g., about 0.25) lower than the fragment ratio correlation in healthy subjects (e.g., a correlation between cfDNA fragment ratios and a reference DNA fragment ratio, such as a DNA fragment ratio from one or more healthy subjects).

[0053] The methodology of the present invention further comprises predicting variant allele fraction (MAF) based on cfDNA fragmentation profile. The MAF of a mutation in DNA is a common value reported in oncology diagnostic tests, and represents the fraction of analyzed DNA molecules that contain the variant of interest. For tumor-derived variants identified in circulating cell-free DNA (cfDNA), MAF represents the fraction of all cfDNA that contains the variant. cfDNA is a combination of tumor-derived DNA and normal cell-derived DNA, and therefore the MAF value of clonal somatic variants captures the fraction of tumor-derived cfDNA. This MAF correlates with circulating tumor DNA fraction, and can therefore be used as a proxy for circulating tumor DNA fraction.

[0054] In various embodiments, the present invention can detect cancer in a subject, predict disease prognosis, predict response to treatment, and / or assess overall survival of a subject by the predicted MAF. FIG. 1 is a block diagram showing a process for training a machine learning model to predict MAF. The predicted MAF can be referred to as DELFI surveillance score (DMS). At 102, a DELFI score is calculated based on the cfDNA fragmentation profile. The DELFI score indicates how similar the fragmentation profile is to a typical cancer patient or a cancer-free individual. In some embodiments, the calculation of the DELFI score includes: (i) determining the short cfDNA fragment to long cfDNA fragment ratio of the sample; (ii) determining the Z-score of the cfDNA fragment of the sample by chromosome arm; (iii) quantifying the cfDNA fragment density by computational mixed model analysis; and (iv) processing the output of (i)-(iii) by the machine learning model to define a score. In various embodiments, the score is utilized to determine the probability of overall survival of the subject.

[0055] In one illustrative example (Example 1), in a multi-cancer cohort, we calculated short-to-long fragment ratios by 5MB bins, Z-scores by chromosome arms, and mixture models of cfDNA fragment sizes for each individual from low-coverage whole genome sequencing. Using these features as input, we fitted a cross-validated gradient boosting machine to each individual's cancer status (cancer / non-cancer). The output of this model is a score ranging from 0 to 1, with higher values ​​indicating stronger cancer signals and lower values ​​indicating greater similarity to non-cancer. The scores generated by these techniques can be used as features to train machine learning models that generate DMS.

[0056] At 104, DELFI information may be calculated. In some embodiments, DELFI information may be equivalent to the correlation between the binned and mean-centered short-to-long ratio of a given sample and the binned and mean-centered short-to-long ratio of a healthy sample minus 1. For example, a healthy sample may be equivalent to the median of the binned and mean-centered short-to-long ratio of a reference cohort that contains only healthy samples. As used herein, mean-centered short-to-long ratio is the binned short-to-long ratio minus the overall mean value.

[0057] At 106, a set of weights may be determined for the computational mixture model. In some embodiments, the mixture model may be a vector that includes 11 weights that summarize the fragmentation distribution in the sample. The weights from the mixture model are estimated by a Bayesian mixture of normal distributions of empirical fragment size distributions.

[0058] At 108, a regression model may be trained on the measured MAF of an individual so that the model learns features such as the DELFI score, DELFI information, and mixture model weights of the sample that contribute to the known MAF of the sample. For example, MAF may be correlated to tumor burden, for example, as estimated by MAF. In some embodiments, the regression model may be a hierarchical Bayesian regression model that includes multiple layers, each layer including more predictors. At run-time, the model takes the DELFI score, DELFI information, and mixture model weights as inputs and outputs the predicted MAF. Training is performed by Leave-One-Patient-Out cross-validation. In this cross-validation scheme, data from each patient is provided in turn, the model is trained on the remaining samples, and then the trained model is used to generate predictions for the provided samples. In one example model, the MAF is a beta-distributed random variable, and the model assumes that the expected MAF of a given sample is functionally related to the described features by the inverse logit of the feature matrix multiplied by the vector of regression coefficients, plus a patient-specific random intercept that accounts for within-patient correlation between measurements.

[0059] At 110, the trained model may be validated to ensure that a desired level of accuracy is reached. In various embodiments, the trained model may be evaluated statistically and clinically. For example, the quality of the generated predictions may be evaluated by assessing the correlation between predicted and observed tumor burden values. Other examples of validation schemes performed to evaluate the trained model include observing longitudinal plots overlaying measured and predicted tumor burden values, and evaluating the relationship between tumor burden predictions and time to patient death.

[0060] The model may be validated in a clinical setting to understand the clinical utility of the predictions. To clinically validate the model, two treatment-naïve metastatic cohorts were acquired. The cohorts were initiated with antineoplastic treatment with chemotherapy or targeted agents, and the ability of the predicted ctDNA burden (represented by MAF) to predict survival for each patient in the cohort was determined at baseline and the first post-treatment blood draw. Observed MAF data were acquired for 76 patients with metastatic colorectal cancer (mCRC) and 17 patients with metastatic non-small cell lung cancer (mNSCLC). All patient samples were analyzed independently with the DELFI surveillance score approach. The first post-treatment blood draw was performed between 4 and 12 weeks post-treatment for mCRC patients and between 1 and 3 weeks post-treatment for mNSCLC patients. MAF of clonal variants was measured by digital droplet PCR (RAS / RAF variants) and deep targeted NGS sequencing (EGFR) for mCRC and mNSCLC patients, respectively.

[0061] The Kaplan-Meier estimator was used to evaluate the predictive value of a single threshold of modeled ctDNA burden. In some embodiments, the DMS threshold may be selected for each cohort by leave-one-patient-out cross-validation. In this analysis, one sample was removed and the threshold that minimized the log-rank p-value was selected. This process was repeated for each patient in the cohort, and the median of all optimized thresholds from the cross-validation was selected as the final threshold of the Kaplan-Meier estimate. When available, Cox proportional hazards models were also used to evaluate the predictive value of continuous modeled ctDNA burden for progression-free survival and overall survival. In alternative embodiments, other approaches to determining thresholds may be used, such as using a reference set of individuals with no or low tumor fraction.

[0062] Model validation results FIG. 2 is a graph plot showing a comparison of observed longitudinal MAF data with predicted tumor burden generated by the model. Specifically, the plot shows the relative MAF trajectories of mCRC patients obtained from observed and predicted MAF data. Each panel plots the observed longitudinal MAF data as a black line, with predictions from the model overlaid as a blue line. The plot also shows the boundaries of the 95% Bayesian confidence intervals overlaid as light blue shaded areas on the black and blue lines. Each panel represents longitudinal data from one subject, and the observed trajectory of a given patient's MAF is in most cases highly consistent with the predicted MAF (i.e., DMS).

[0063] Figure 3 is a graph plot showing a comparison of observed and predicted MAF, regardless of the specific time point or which patient a given sample is derived from. Thus, the plot includes DMS and observed MAF for the entire mCRC cohort. The solid diagonal line included in the plot represents the line of uniformity of DMS and observed MAF. As shown in the figure, the DELFI surveillance score is in very good agreement with the observed MAF in most cases. In the few cases where the prediction is far from the line of uniformity, especially in samples with low measured MAF (i.e., less than 1%), there is evidence that the problem lies in the measurement process rather than the model. For example, tumor and / or metastasis heterogeneity and clonal evolution that may occur during treatment may make it more difficult to evaluate variants of interest.

[0064] FIG. 4 is a graph plot showing a comparison of observed MAF trajectories of mNSCLC patients with DMS values ​​from a model trained on the mCRC cohort. Each panel plots the observed longitudinal MAF data as a black line, with predictions from the model overlaid as a blue line. The plot also shows the boundaries of the 95% Bayesian confidence intervals overlaid as light blue shaded regions on the black and blue lines. Each panel represents longitudinal data from one subject, and the observed trajectory of a given patient's MAF is in most cases closely aligned with the predicted MAF (i.e., DMS).

[0065] The results of the comparison show that a model trained on MAF data from patients with one type of cancer (e.g., mCRC patients) can be successfully applied to patients with a different type of cancer (e.g., mNSCLC cohort). This external applicability is a desirable feature of predictive models because predictions are generally of high quality, regardless of substantial differences between the two cohorts (cancer type, sequencing depth, etc.). The external applicability of the predictive models described herein increases the efficiency of predictive model development and training by allowing one predictive model trained on a particular dataset to be used to generate useful predictions for patients different from those included in the training dataset.

[0066] Figure 5 is a graphical plot showing progression-free survival within a metastatic colorectal cancer cohort (i.e., mCRC cohort) in panel A and progression-free survival within a metastatic lung cancer cohort (i.e., mNSCLC cohort) in panel B. The plot shows results from Kaplan-Meier estimation to distinguish patients with high and low DMS by cohort-specific cross-validation thresholds. In both cohorts, patients with DMS below the cross-validation threshold (DELFI surveillance score (-)) at the first post-treatment time point showed longer progression-free survival than patients with high DMS. According to these results, DMS allows distinguishing subgroups of patients who are more likely to survive without progression from the first post-treatment time point.

[0067] FIG. 6 is a plot showing the overall survival of the mCRC cohort based on DMS in panel A and the overall survival after treatment of mCRC based on DMS in panel B. Overall survival data of the mCRC cohort was obtained to determine the ability of DMS to predict response to treatment. The survival data of each patient was labeled to indicate whether each patient who underwent surgery had a complete or incomplete resection. The DMS at the time point before treatment was started was evaluated against a separate cross-validated discrimination threshold. As shown in panel A, patients with DMS below the threshold had a longer overall survival than patients with DMS above the threshold. In addition, in patients who underwent surgery, DMS was able to further distinguish the overall survival of patients with incomplete or complete resection in panel B from completely resected patients who had a longer predicted overall survival. Taken together, the results show the utility of DMS to predict disease prognosis even before treatment is started.

[0068] Clonal variant MAF correlates with ctDNA burden and may therefore be useful as a quantitative metric to estimate the fraction of plasma DNA derived from the tumor and the patient's overall tumor burden. However, over the course of treatment, the genetic profile of the tumor may change due to the selective pressure of the treatment. Thus, measuring the MAF of only one variant is limited in measuring patient response over time. To evaluate the sensitivity of DMS to changes in tumor DNA during treatment, we determined whether patients treated with clonal variant MAF measured at 0% at the first post-treatment time point would benefit from analysis by DMS.

[0069] Figure 7 is a plot showing progression-free survival for patients in the mCRC cohort with clonal variant MAF measured at 0% at first post-treatment evaluation in panel A and DMS vs. MAF by ddPCR for patients in the mCRC cohort with clonal variant MAF measured at 0% at first post-treatment evaluation in panel B. As shown in panel A, 43 patients with 0% MAF in the mCRC cohort were further divided by DMS with respect to progression-free survival. As shown in panel B, some of these patients had DMS measured at 5%-15%. This data indicates that predicting ctDNA burden based on genome-wide assessment of fragments may be more sensitive than regular MAF.

[0070] In addition, Cox proportional hazards analyses were performed for the mCRC and mNSCLC cohorts to evaluate the predictive value of sequential DMS.At both pretreatment and post-first treatment time points, DMS was able to predict overall survival in the mCRC cohort (HR: 19.2, 95% CI: 2.7-138.5, and HR: 400.4, 95% CI: 11.8-13581.0, respectively) and progression-free survival in the mNSCLC cohort (HR: 67.3, 95% CI: 1.1-4073.6, and HR: 246.5, 95% CI: 2.2-28030.9, respectively).

[0071] Hardware implementation example FIG. 8 illustrates an example of a computer 800 that may be used to implement the training algorithm shown in FIG. 1 and generate DMS values. For example, the computer 800 may, in some embodiments, include a machine learning system that trains a machine learning model to generate the DMS values ​​described above or a portion or combination thereof. The computer 800 may be any electronic device that executes software applications derived from compiled instructions, including, without limitation, a personal computer, a server, a smart phone, a media player, an electronic tablet, a game console, an email device, and the like. In some implementations, the computer 800 may include one or more processors 802, one or more input devices 804, one or more display devices 806, one or more network interfaces 808, and one or more computer-readable media 812. Each of these components may be coupled by a bus 810, and in some embodiments, these components may be distributed across multiple physical locations and coupled by a network.

[0072] The display device 806 may be of any known display technology, including, but not limited to, displays using liquid crystal display (LCD) or light emitting diode (LED) technology. The processor 802 may be of any known processor technology, including, but not limited to, graphics processors and multi-core processors. The input device 804 may be of any known input device technology, including, but not limited to, a keyboard (including a virtual keyboard), a mouse, a trackball, a camera, and a touchpad or display. The bus 810 may be of any known internal or external bus technology, including, but not limited to, ISA, EISA, PCI, PCI Express, USB, Serial ATA, or FireWire. The computer readable medium 812 may be any non-transitory medium that participates in providing instructions to the processor 804 for execution, including, without limitation, a non-volatile storage medium (e.g., optical disks, magnetic disks, flash drives, etc.), or a volatile medium (e.g., SDRAM, ROM, etc.).

[0073] The computer readable medium 812 may include various instructions 814 for implementing an operating system (e.g., Mac OS, Windows, Linux). The operating system may be a multi-user, multi-processing, multi-tasking, multi-threaded, real-time operating system, etc. The operating system may perform basic tasks including, but not limited to, recognizing input from the input device 804, sending output to the display device 806, keeping track of files and directories on the computer readable medium 812, controlling peripherals (e.g., disk drives, printers, etc.) that may be controlled directly or through an I / O controller, and managing traffic on the bus 810. The network communications instructions 816 may establish and maintain network connections (e.g., software for implementing communications protocols such as TCP / IP, HTTP, Ethernet, telephone, etc.).

[0074] Machine learning instructions 818 may include instructions that enable computer 800 to function as a machine learning system and / or train a machine learning model that generates the DMS values ​​described herein. Application(s) 820 may be applications that use or implement the processes described herein and / or other processes. The processes may be implemented in operating system 814. For example, application 820 and / or the operating system may create tasks in the applications described herein.

[0075] The described functionality may be implemented in one or more computer programs executable on a programmable system including at least one programmable processor coupled to receive data and instructions from and transmit data and instructions to the data storage system, at least one input device, and at least one output device. A computer program is a sequence of instructions that can be used directly or indirectly in a computer to perform a particular activity or bring about a particular result. Computer programs may be written in any type of programming language (e.g., Objective-C, Java), including compiled or interpreted languages, and may be deployed in any form, such as as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0076] A processor suitable for executing a program of instructions may include, for example, both general purpose and special purpose microprocessors, and the sole processor or one of multiple processors or cores of any kind of computer. In general, a processor may receive instructions and data from a read-only memory or a random access memory or both. The essential elements of a computer may include a processor for executing instructions and one or more memories for storing instructions and data. In general, a computer may also include, or be operatively coupled to communicate with, one or more mass storage devices for storing data files. Such devices include magnetic disks, such as internal hard disks and removable disks, magneto-optical disks, and optical disks. Storage devices suitable for embodied computer program instructions and data may include any type of non-volatile memory, including, for example, semiconductor memory devices, such as EPROMs, EEPROMs, and flash memory devices, magnetic disks, such as internal hard disks and removable disks, magneto-optical disks, and CD-ROM and DVD-ROM disks. The processor and memory may be supplemented by, or incorporated in, ASICs (Application Specific Integrated Circuits).

[0077] To provide for interaction with a user, the functionality may be implemented on a computer having a display device, such as an LED or LCD monitor, for displaying information to the user, and a keyboard and pointing device, such as a mouse or trackball, by which the user can provide input to the computer.

[0078] The functionality may be implemented in a computer system that includes back-end components such as data servers, or middleware components such as application servers or Internet servers, or front-end components such as client computers having a graphical user interface or Internet browser, or any combination thereof. The components of the system may be connected by any form or medium of digital data communication, such as a communications network. Examples of communications networks include, for example, telephone networks, LANs, WANs, and the computers and networks forming the Internet.

[0079] A computer system may include clients and servers. Clients and servers may generally be remote from each other and typically interact through a network. The relationship of client and server may arise by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

[0080] One or more functions or steps of the disclosed embodiments may be implemented in an API, which may define one or more parameters passed between a calling application and other software code (e.g., an operating system, library routines, functions) that provides a service, provides data, or performs an operation or calculation.

[0081] An API may be implemented as one or more calls in program code that send or receive one or more parameters through a parameter list or other structure based on calling conventions defined in an API specification. A parameter may be a constant, a key, a data structure, an object, an object class, a variable, a data type, a pointer, an array, a list, or another call. API calls and parameters may be implemented in any programming language. A programming language may define the vocabulary and calling conventions that a programmer uses to access the functions that support the API.

[0082] In some implementations, the API calls may report to the application the capabilities of the device running the application, such as input capabilities, output capabilities, processing capabilities, power capabilities, communication capabilities, etc.

[0083] While various embodiments have been described above, it should be understood that they are presented by way of example and not limitation. It will be apparent to one skilled in the art that various changes in form and detail can be made without departing from the spirit and scope of the invention. Indeed, upon reading the above description, it will be apparent to one skilled in the art how to implement alternative embodiments. For example, other steps may be provided or removed from the described flows, and other components may be added or removed from the described systems. Accordingly, other implementations are within the scope of the following claims.

[0084] Additionally, it should be understood that any features emphasizing functionality and advantages have been presented for illustrative purposes only, and that the disclosed methodologies and systems are each sufficiently flexible and configurable so that they can be utilized in ways other than those shown.

[0085] The term "at least one" may be used frequently in the specification, claims, and drawings, but the terms "a," "an," "the," "said," and the like also refer to "at least one" or "the at least one" in the specification, claims, and drawings.

[0086] Finally, Applicant intends that only those claims which expressly contain the phrase "means for" or "step for" be construed under 35 U.S.C. 112(f). Any claim which does not expressly contain a "means for" or "step for" will not be construed under 35 U.S.C. 112(f).

[0087] The methods and systems described herein are useful for detecting, predicting, treating, and / or monitoring a cancer condition in a subject. Any suitable subject, e.g., a mammal, can be evaluated, monitored, and / or treated as described herein. Some examples of mammals that can be evaluated, monitored, and / or treated as described herein include, without limitation, humans, primates such as monkeys, dogs, cats, horses, cows, pigs, sheep, mice, and rats. For example, a human having or suspected of having cancer can be evaluated with the methods described herein and, optionally, treated with one or more cancer therapies described herein.

[0088] A subject having or suspected of having any suitable type of cancer can be monitored, evaluated, and / or treated (e.g., by administering one or more cancer therapies to the subject) with the methods and systems described herein. The cancer can be any stage of cancer. In some embodiments, the cancer can be early stage cancer. In some embodiments, the cancer can be asymptomatic cancer. In some embodiments, the cancer can be residual disease and / or recurrence (e.g., after surgical resection and / or after cancer therapy). The cancer can be any type of cancer. Examples of types of cancer that can be evaluated, monitored, and / or treated as described herein include, without limitation, lung cancer, colorectal cancer, prostate cancer, breast cancer, pancreatic cancer, bile duct cancer, liver cancer, CNS cancer, gastric cancer, esophageal cancer, gastrointestinal stromal tumor (GIST), uterine cancer, and ovarian cancer. Other types of cancer include, without limitation, myeloma, multiple myeloma, B-cell lymphoma, follicular lymphoma, lymphocytic leukemia, leukemia, and myeloid leukemia. In some embodiments, the cancer is a solid tumor. In some embodiments, the cancer is a sarcoma, carcinoma, or lymphoma. In some embodiments, the cancer is lung cancer, colorectal cancer, prostate cancer, breast cancer, pancreatic cancer, bile duct cancer, liver cancer, CNS cancer, gastric cancer, esophageal cancer, gastrointestinal stromal tumor (GIST), uterine cancer, or ovarian cancer. In some embodiments, the cancer is a hematological cancer. In some embodiments, the cancer is a myeloma, multiple myeloma, B-cell lymphoma, follicular lymphoma, lymphocytic leukemia, leukemia, or myeloid leukemia.

[0089] When treating a subject having or suspected of having cancer as described herein, the subject may be administered one or more cancer therapies. The cancer treatment may be any suitable cancer treatment. The one or more cancer therapies described herein may be administered to the subject at any suitable frequency (e.g., once or several times over a period of several days to several weeks). Examples of cancer therapies include, without limitation, surgical intervention, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy (e.g., chimeric antigen receptors and / or T cells with wild-type or modified T cell receptors), targeted therapy (e.g., kinase inhibitors, antibodies, bispecific antibodies), such as administration of kinase inhibitors (e.g., kinase inhibitors that target specific genetic lesions such as translocations or mutations), 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 embodiments, treating cancer can reduce the severity of cancer, reduce the symptoms of cancer, and / or reduce the number of cancer cells in a subject.

[0090] In some embodiments, the cancer treatment can be a chemotherapy drug.Non-limiting examples of chemotherapy drugs include amsacrine, azacitidine, azathioprine, 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, fluorouracil, gemcitabine, Include 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.Other examples of anticancer therapy are known in the art.See, for example, therapy guidelines from the American Society of Clinical Oncology (ASCO), the European Society for Medical Oncology (ESMO), or the National Comprehensive Cancer Network (NCCN).

[0091] When monitoring a subject having or suspected of having cancer as described herein, monitoring may occur 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 subjects for increased monitoring.

[0092] In some embodiments, monitoring can include conventional techniques capable of monitoring one or more cancer treatments (e.g., the effectiveness of one or more cancer treatments). In some embodiments, subjects selected for enhanced monitoring can be administered diagnostic tests (e.g., any diagnostic test disclosed herein) at an increased frequency compared to subjects not selected for enhanced monitoring. For example, subjects selected for enhanced monitoring can be administered diagnostic tests twice daily, daily, biweekly, weekly, bimonthly, monthly, quarterly, semi-annually, annually, or any frequency thereof.

[0093] In various embodiments, DNA is present in a biological sample taken from a subject and is used in the methodology of the present invention.Biological sample can be almost any kind of biological sample that contains DNA.Biological sample is typically liquid, for example, whole blood or a part thereof that has circulating cfDNA. In embodiments, the sample includes DNA from tumor or liquid biopsies, including but not limited to amniotic fluid, aqueous humor, vitreous humor, blood, whole blood, fractionated blood, plasma, serum, breast milk, cerebrospinal fluid (CSF), earwax, chyle, chyme, endolymph, perilymph, feces, exhaled air, gastric acid, gastric juice, lymph, mucus (including nasal mucus and sputum), pericardial fluid, peritoneal fluid, pleural fluid, pus, mucosal secretions, saliva, exhaled breath condensate, sebum, semen, sputum, sweat, synovial fluid, tears, vomit, prostatic fluid, nipple aspirate, tears, perspiration, oral mucosal swabs, cell lysates, gastrointestinal fluid, biopsy tissue, and urine or other biological fluids. In one embodiment, the sample includes DNA from circulating tumor cells.

[0094] As disclosed above, the biological sample can be a blood sample. Blood samples can be collected by methods known in the art, such as fingerstick sampling or phlebotomy. Preferably, the blood sample is about 0.1-20 ml, or alternatively about 1-15 ml, with a blood volume of about 10 ml. Smaller volumes may be used, as well as free DNA circulating in blood. Microsampling and sampling by needle biopsy, catheter, excretion or production of bodily fluids containing DNA are also potential sources of biological samples.

[0095] The disclosed methods and systems may include any method or sequencing device for performing nucleic acid sequencing, including nucleic acid amplification, polymerase chain reaction (PCR), nanopore sequencing, 454 sequencing, and insertion tagged sequencing, to utilize nucleic acid sequence information. In some embodiments, the disclosed methodology or system is implemented using Illumina, Inc. (HiSeq™ X10, HiSeq™ 1000, HiSeq™ 2000, HiSeq™ 2500, Genome Analyzers™, MiSeq™, etc.). 、 Systems such as those provided by Oxford Nanopore Technologies (GridiON™, MinION™) or Pacific Biosciences (Pacbio™ RS II or Sequel I or II) are utilized.

[0096] The present invention includes systems for performing the steps of the disclosed methods, which are described, in part, in terms of functional components and various process steps. Such functional components and process steps may be realized by any number of components, operations, and techniques configured to perform the specified functions and achieve various results. For example, the present invention may employ various biological samples, biomarkers, elements, materials, computers, data sources, storage systems and media, information collection techniques and processes, data processing standards, statistical analyses, regression analyses, and the like, which may perform various functions.

[0097] Accordingly, the present invention further provides a system for monitoring, detecting, analyzing, and / or assessing cancer, in various embodiments, the system comprising (a) a sequencer configured to generate a low-coverage whole genome sequencing dataset of a sample, and (b) a computer system and / or processor capable of performing the methods of the present invention.

[0098] In some embodiments, the computer system further comprises one or more additional modules. For example, the system may comprise one or more of the following extraction and / or isolation units operable to perform suitable genetic content analysis, for example, to select cfDNA fragments of a particular size.

[0099] In some embodiments, the computer system further comprises a visual display device. The visual display device may be operable to display the curve fit line, the reference curve fit line, and / or a comparison of both.

[0100] The detection and analysis methods according to various aspects of the present invention can be implemented in any suitable manner, for example, using a computer program running on a computer system. As discussed herein, the exemplary system according to various aspects of the present invention can be implemented in combination with a computer system, for example, a conventional computer system including a processor and random access memory, such as a remotely accessible application server, a network server, a personal computer, or a workstation. The computer system also preferably includes additional memory devices or information storage systems, such as a mass storage system, and a user interface, such as a conventional monitor, keyboard, and tracking device. However, the computer system can include any suitable computer system and associated equipment and can be configured in any suitable manner. In one embodiment, the computer system includes a stand-alone system. In another embodiment, the computer system is part of a network of computers including a server and a database.

[0101] The software necessary to receive, process, and analyze the information may be implemented on a single device or on multiple devices. The software may be accessible over a network such that storage and processing of information is performed remotely to a user. The system and its various elements according to various aspects of the present invention provide functions and operations that facilitate detection and / or analysis, such as data collection, processing, analysis, reporting, and / or diagnosis. For example, in this aspect, the computer system executes a computer program that may receive, store, retrieve, analyze, and report information about the human genome or regions thereof. The computer program may include multiple modules that perform various functions or operations, such as a processing module for processing raw data and generating supplemental data, and an analysis module for analyzing the raw data and supplemental data to generate a disease state model and / or a quantitative assessment of diagnostic information.

[0102] The procedures performed by the system may include any suitable process that facilitates analysis and / or cancer diagnosis. In one embodiment, the system is configured to establish a disease state model and / or determine a disease state of a patient. Determining or identifying a disease state may include, for example, generating any useful information regarding a patient's condition related to a disease, such as performing a diagnosis, providing information useful for a diagnosis, assessing the stage or progression of a disease, identifying symptoms that may indicate susceptibility to a disease and whether further testing is recommended, predicting and / or assessing the effectiveness of one or more treatment programs, or assessing the disease state, likelihood of disease, or other aspects of the patient's health.

[0103] The following examples are provided to further illustrate the advantages and features of the present invention and are not intended to limit the scope of the invention. The examples are typical of those that might be used, but other procedures, methodologies, or techniques known to those skilled in the art may alternatively be used. EXAMPLES

[0104] Example 1 Cancer detection and prediction of overall survival In this example, the methodology of the present disclosure is utilized to detect cancer and predict overall patient survival. Exhibit A shows the studies and results.

[0105] This study of prospectively enrolled individuals demonstrated the ability of the cfDNA fragmentation assay to distinguish between individuals with and without cancer. The assay of the present invention showed high performance in a multi-cancer setting using only fragmentation-related information obtained from low-coverage WGS.

[0106] The results suggest that the machine learning model can distinguish cancer from non-cancer by cfDNA fragmentation profiles, regardless of the presence of common non-malignant conditions (including cardiovascular, autoimmune, or inflammatory diseases).In addition, individuals with higher DELFI scores had a worse prognosis, independent of other characteristics.

[0107] These data support the development of genome-wide cfDNA fragmentation assays for the noninvasive detection of both single and multiple cancers.

[0108] Example 2 Cell-free DNA fragmentation profiling for monitoring treatment response in metastatic cancer The fraction of circulating tumor DNA (ctDNA) molecules in plasma (ctDNA burden) has become an available measure to describe the overall tumor burden in cancer patients. ctDNA burden can change over time, decreasing with treatment response and increasing when tumors become resistant to therapy. Monitoring ctDNA dynamics throughout treatment allows physicians to make timely treatment decisions. Ideally, a fast, inexpensive and generally applicable monitoring test is needed to predict treatment success and patient prognosis. Plasma ctDNA from liquid biopsies has great potential as a minimally invasive biomarker for tumor detection and (targeted) treatment response monitoring. Due to its short half-life, plasma ctDNA is a dynamic tumor marker, allowing earlier detection of recurrence than imaging and clinical parameters.

[0109] There are various techniques for ctDNA profiling. Targeted next-generation sequencing (NGS) is a sensitive approach that can provide information on somatic abnormalities and detect genomic alterations in tumors. However, the high incidence of clonal hematopoietic mutations in the elderly population limits this approach. To prevent these variants from obscuring the detection of tumor-specific alterations, tissue- or leukocyte-derived approaches must be used. On the other hand, single-nucleotide variants can be tracked longitudinally using less expensive ctDNA hotspot mutation approaches such as droplet digital PCR (ddPCR). These hotspot mutation assays detect a limited number of somatic tumor alterations, making them generally inapplicable to diverse tumors within a patient population and narrowing the scope of tumor genetic makeup they can capture. For example, in patients with metastatic colorectal cancer (mCRC), trackable RAS / BRAF driver mutations are present in only half of patients.

[0110] Previous studies have shown that the size distribution of cell-free DNA (cfDNA) across the genome can be used to elucidate its origin. The proportion of shorter fragments is higher in cancer patients than in healthy individuals. As the survival rate of cancer patients is inversely proportional to the stage of the disease, the size composition of cfDNA fragments, i.e., the ratio of short to long cfDNA fragments, was exploited to develop a tool for early disease detection in cancer patients. This approach, called DELFI (DNA Evaluation of Fragment for Early Interruption), can distinguish cancer from non-cancer and indicate the origin of the tumor. Due to the minimally invasive nature of this technique, cfDNA fragmentomics may also have increased clinical value in monitoring disease progression. Therefore, the inventors developed a machine learning classifier, the DELFI Tumor Fraction Score (DELFI-TF), that can detect tumor dynamics without requiring information about the tumor of origin. In the study, the inventors evaluate this DELFI-TF classifier for treatment response monitoring in mCRC patients.

[0111] DELFI-TF model development by genome-wide cfDNA fragmentation profiling - 692 consecutive plasma samples from patients with mCRC and RAS / BRAF-mutated (n=79) or RAS / BRAF-wild-type (n=74) disease participating in a prospective phase III clinical trial were processed and analyzed (CAIRO5) (Table 1, Figures 9A and 9B). After study eligibility confirmation, including unresectable status of liver metastases defined by a central panel of experts, KRAS (exons 2, 3, and 4), NRAS (exons 2 and 3), and BRAF (codon 600) mutation status was assessed in available tissue samples, followed by blood draws at pretreatment baseline and at serial time points during treatment (Figures 9A and 10). During the first 2 months, patients in the mutant and wild-type groups were treated with fluoropyrimidine-based first-line regimens (FOLFOX or FOLFIRI) + bevacizumab (Figures 9A and 10A). Therefore, planned radiological evaluation to assess the resectability of liver metastases was performed by central tumor board review. Tumor-informative cfDNA analysis using droplet digital PCR (ddPCR) was performed retrospectively on 312 samples from patients in the mutant group. Tumor-independent DELFI-TF analysis was successfully performed on 692 samples from patients in both the mutant and wild-type groups (Figure 9B and Figure 10A). DELFI-TF failure rates associated with library generation and whole genome sequencing (WGS) were 0.42% and 0.29%, respectively (Figure 9B). Participants were planned to be followed until death or study discontinuation.

[0112] Table 1. Demographic and clinical characteristics of study participants TIFF2025512761000002.tif223143TIFF2025512761000003.tif123143OS stands for overall survival, and PFS stands for progression-free survival.

[0113] To perform a mutation-independent assessment of cancer-specific alterations in cfDNA, we first designed a DELFI-TF model (Figure 10B). Tumor burden was first quantified as the mutant allele frequency (MAF) of tumor-tissue-proven RAS / BRAF variants measured by ddPCR for all cfDNA samples of patients in the mutation group (n=312). Using duplicate aliquots of cfDNA samples, genome-wide fragment-sequencing statistics were obtained by low-coverage WGS of cfDNA libraries (Figure 10B). A hierarchical Bayesian regression model was trained and cross-validated against the MAF of tumor-specific driver RAS / BRAF variants measured by ddPCR in all longitudinal cfDNA samples sequenced within the mutation group. To generate a predicted DELFI-TF value for each sample, the model considered the DELFI score, plasma aneuploidy (PA) score, and weight components from a mixture model utilizing cfDNA fragment size density (Figure 10B). We performed unsupervised clustering analysis using short-to-long fragment size ratios across 504 5Mb bins and arm-level copy number z-scores across baseline and treatment timepoints for 39 chromosome arms in 128 mCRC patients (Figure 10C). Notably, differences in fragmentation profiles were observed in multiple regions across the genome in the majority of mCRC patients across several clinical and demographic characteristics at baseline, most of which corresponded to high DELFI-TF values. Similarly, we observed that the majority of timepoints associated with progressive disease by imaging assessment also showed significant heterogeneity and high tumor fractions, in contrast to the majority of timepoints associated with stable disease or radiological response after initiation of first-line systemic therapy, which are associated with fewer genetic abnormalities and lower DELFI-TF values ​​(Figure 10C and Figure 10D). These findings suggest that the model of the present invention designed to predict tumor fraction in cfDNA can identify systemic treatment response in real time in a non-invasive manner.

[0114] DELFI-TF accurately reflects cfDNA variant allele frequency and copy number alterations - An independent analysis of the DELFI-TF model was performed with non-cancer control samples (n=155) from a Danish cohort of symptomatic patients with previously negative cancer diagnoses. Compared to treatment-naïve samples (n=128) from the CAIRO5 cohort, the non-cancer control samples showed significantly lower DELFI-TF values, with an upper 95% confidence interval (CI) of 0.006. Of note, all treatment-naïve samples from mCRC patients had DELFI-TF values ​​significantly higher than 0.006 (Figure 11A). We observed a similar distribution when evaluating samples from the two cohorts with ichorCNA, a tool designed to estimate tumor fraction in ultra-low-pass WGS data (Figure 12A). However, the non-cancer controls showed a slightly wider range of tumor fraction values, with a higher upper 95% CI (0.017). In addition, ichorCNA values ​​of some CRC samples overlapped with the estimated tumor fraction of non-cancer controls, suggesting that DELFI-TF more accurately reflects the disease burden detected in cfDNA samples.

[0115] Next, the analytical performance of DELFI-TF compared to mutation-based tumor burden assessment was evaluated by ddPCR for RAS / BRAF MAF quantification (Figure 11B). A strong correlation was observed between MAF and DELFI-TF values ​​(Pearson r=0.85, p=3.92e-89). Interestingly, a wide range of positive DELFI-TF values ​​was observed in 60 (20.5%) treatment time points with undetectable RAS / BRAF MAF by ddPCR, implying that the DELFI-TF approach may be more sensitive for measuring tumor fraction in patients who received effective antitumor treatment. Furthermore, it was confirmed that these discordant time points with undetectable RAS / BRAF MAF showed DELFI-TF values ​​that correlated with tumor fraction predicted by ichorCNA (Spearman rho=0.54, p<0.001) (Figure 12B). We examined the association of cfDNA fragments with copy number alterations in tissue (Figure 11C) and plasma samples (Figure 11D). Formalin-fixed paraffin-embedded patient-matched (FFPE) tumor tissues from 104 of 153 patients were analyzed by low-pass WGS (mean coverage 0.2x). We consistently observed that aberrant cfDNA fragmentation profiles in tissue samples were present in copy-neutral regions of the genome and were further affected in regions with copy number alterations (Figure 11C). In addition, we demonstrated comparable analytical performance and significant correlation between tumor fractions assessed by DELFI-TF and ddPCR MAF for cfDNA copy numbers of MBD1 (Pearson, DELFI-TF r=0.64, p<0.001, ddPCR MAF r=-0.67, p<0.001) and PLGC1 (Pearson, DELFI-TF r=-0.84, p<0.001. ddPCR MAF r=0.55, p<0.001), genes commonly deleted and amplified in mCRC, respectively (Figure 11D). We investigated the ability to detect nucleosome-depleted regions using relative coverage at transcription start sites (TSSs) as a surrogate marker of gene expression in CRC.The relative coverage in TSS of a group of approximately 900 genes highly expressed in CRC was observed to be significantly lower in baseline samples than in treatment samples (Wilcoxon, p<0.001) (Figure 11E and Figure 11F), reflecting the dynamic changes detected during disease response and progression (Figure 12C). In summary, DELFI-TF accurately captures MAF-associated cancer-specific alterations and copy number abnormalities in cfDNA.

[0116] DELFI-TF correlates with clinical characteristics and standard imaging assessments - We compared the DELFI-TF approach with clinical patient characteristics. At the treatment-naive time point, a moderate correlation was observed between DELFI-TF and ddPCR MAF and the sum of the longest diameters (SLD) of target metastatic lesions in the liver (DELFI-TF Spearman rho=0.49, p<0.001; ddPCR MAF Spearman rho=0.48, p<0.001) (Figure 13A). In contrast, no correlation was found with serum carcinoembryonic antigen (CEA) levels measured before treatment initiation (DELFI-TF Spearman rho=0.1, p=0.43; ddPCR MAF Spearman rho=0.15, p=0.24) (Figure 13B). Baseline DELFI-TF and ddPCR MAF tumor fractions robustly correlated with clinical response, with significantly lower pretreatment levels in patients who subsequently achieved partial or complete response compared with patients with stable disease or progressive disease on two consecutive scans (Wilcoxon, DELFI-TF p<0.05; ddPCR MAF p<0.05) (Figure 13C). Furthermore, patients deemed resectable after systemic induction therapy had significantly lower baseline tumor fractions (Kruskal-Wallis, DELFI-TF p<0.05; ddPCR MAF p<0.05) (Figure 13D), as did patients with metachronous disease (Wilcoxon, p<0.05) (Figure 13E). Notably, DELFI-TF and ddPCR MAF tumor fractions did not differ by tumor sideness (Figure 14A) or RAS / BRAF mutation status (Figure 14B). At baseline, DELFI-TF and ddPCR MAF tumor fractions were significantly lower in patients who did not experience disease progression at any time during the CAIRO5 study (non-progressors) compared to patients who experienced disease progression at some time points during treatment (progressors) (Figure 14C). In contrast, analysis using SLD of target liver lesions at baseline could not distinguish between non-progressors and progressors (Figure 14C). We also demonstrated that DELFI-TF can accurately track longitudinal disease burden dynamics at later time points in patients who underwent curative-intent hepatic metastases resection (Figure 13C). Overall, we observed that progressors more frequently exhibited elevated DELFI-TF values ​​at earlier time points and disease progression at later time points than non-progressors (Figure 13G).

[0117] Once the analytical equivalence of the DELFI-TF model and the ddPCR assay for RAS / BRAF MAF assessment was verified, we decided to further explore the association between the dynamic changes of DELFI-TF and clinical outcomes. To summarize the longitudinal evolution of the continuous DELFI-TF values ​​into a single score, we calculated the DELFI-TF slope, which is defined as the slope of the line fitted to the DELFI-TF values ​​by linear regression, starting from the time of the first blood biopsy after the start of treatment and ending at the time of disease progression confirmed by RECIST1.1. We then observed that the DELFI-TF slope of patients who experienced partial or complete response as the best overall response tended to be lower (Fisher's exact test, p=0.1) (Figure 14E). Overall, the analysis of DELFI-TF and ddPCR MAF over time showed comparable tumor dynamics (Figure 15A). For patients in the wild-type group, the analysis over time could only be performed using DELFI-TF values ​​(Figure 15B). Patients with DELFI-TF gradients below the median had higher objective radiological response rates to first-line treatment and longer follow-up durations than patients with DELFI-TF gradients above the median ( Figure 16 ).

[0118] We then correlated baseline DELFI-TF and DELFI-TF slope with survival outcomes. At baseline, patients with DELFI-TF values ​​below the first quartile had a longer median progression-free survival (PFS) than patients with DELFI-TF values ​​above the first quartile (13.4 vs. 10.2 months, hazard ratio [HR] = 1.77, 95% CI 1.12-2.78, log-rank p = 0.013) (Figure 17A). In the RAS / BRAF mutant group, baseline MAF for tumor fraction assessment showed a similar difference in median PFS (14.4 vs. 8.3 months, HR = 2.56, 95% CI 1.36-4.83, log-rank p = 0.00272) (Figure 18A). Serum CEA levels at baseline were not predictive of disease progression or death (Figure 18B). In the study, patients were evaluated 2 months after initiation of therapy by an expert panel to assess the resectability of liver metastases. This initial clinical response assessment did not distinguish between survival differences between patients with partial responses and those with stable disease (11.3 vs. 11.2 months, HR = 1.13, 95% CI 0.79-1.61, log-rank p = 0.52) (Figure 18C). Patients with DELFI-TF slopes below the median demonstrated longer PFS in the overall study population (13.4 vs. 10.4 months, HR=2.03, 95% CI 1.247-3.318, log-rank p=3.76e-3) (Figure 17B), as did patients who experienced sustained clinical benefit defined as objective response or stable disease for >12 months (16.7 vs. 13.3 months, HR=2.235, 95% CI 1.097-4.553, log-rank p=.023) (Figure 17C). Patients with DELFI-TF slopes below the median also experienced significantly longer overall survival (OS; 59.4 vs. 29.1 months, HR=3.05, 95% CI 1.58-5.90, log-rank p=5.135e-4) (Figure 17D). The inventors could further stratify by resection status and observed that patients who underwent complete resection of the primary tumor and liver metastases had a better OS outcome than those with incomplete or no resection (Figure 17D).

[0119] Liquid biopsy cfDNA analysis is a new and promising clinical tool in cancer research. The DELFI-TF score was developed as a fragmentation approach to quantitatively measure tumor burden and showed its potential for longitudinal disease monitoring of mCRC patients.

[0120] Currently, liquid biopsy ctDNA testing for the presence of cancer mostly relies on the detection of one or more somatic tumor alterations. Different research advantages exploit the properties of cfDNA fragmentation methods as alternative functions. In vitro and in silico size selection of cfDNA molecules, i.e., selecting short cfDNA fragments over long cfDNA fragments, can enrich ctDNA and enhance the identification of genetic alterations in ctDNA. Alternatively, genome-wide fragmentation profiles can facilitate tumor detection and the identification of the tumor of origin. The novelty of the cfDNA fragmentation approach of the present invention is that it allows longitudinal monitoring of patient response by low-coverage whole genome sequencing of minute amounts of cfDNA without the need to detect driver mutations.

[0121] Despite advances in diagnosis and treatment, most mCRC patients will relapse, so there is a clinical need for biomarkers to guide the treatment process. However, currently available follow-up methods, such as clinical imaging and serum CEA, have limited accuracy in detecting tumor tissue viability, making it difficult to evaluate treatment efficacy immediately after therapy initiation. The current study showed that DELFI-TF could predict PFS better than serum CEA measurement and clinical computed tomography (CT) after treatment initiation, making DELFI-TF potentially more sensitive than traditional approaches for treatment response monitoring. Identification of treatment response or progression provides physicians with the opportunity to adapt patients' treatment regimens.

[0122] Besides DELFI-TF, ddPCR MAF after treatment initiation was also prognostic for disease recurrence. However, the ability to detect differences in PFS in patients with undetectable ddPCR MAF suggests that DELFI-TF may be more sensitive for treatment response monitoring, although fragmentation monitoring approaches that can be combined with targeted sequencing approaches cannot track treatment-induced genomic alterations in tumors. Furthermore, both pre-treatment ddPCR MAF and DELFI-TF could show the success rate and OS of complete resection of liver metastases. However, DELFI-TF has conceptual advantages over hotspot mutation assays such as ddPCR. DELFI-TF does not require prior knowledge of driver alterations in the tumor, making it generally applicable to samples from patients with any type of cancer. The low-coverage WGS required for fragmentation profiles is less costly than targeted sequencing. As tumor burden fluctuates over time, decreasing with treatment response and increasing as tumors become resistant to therapy, DELFI-TF can be utilized as a tool to highlight the appropriate timing for more precise panel sequencing analysis.

[0123] Although the number of patients with blood samples taken after liver resection was limited, postoperative DELFI-TF positivity appeared to indicate disease recurrence with moderate sensitivity. However, blood testing immediately before surgery may have been a confounding factor in the training cohort. Measurements within 48 hours after surgery showed a sharp increase in DELFI-TF. Because surgery is an invasive procedure, samples taken immediately after surgery may represent wound healing rather than tumor-derived cfDNA. Therefore, the cutoff for a positive DELFI-TF result in samples taken after complete resection, i.e., the minimal residual disease setting, needs to be further investigated.

[0124] Here, we evaluated DELFI-TF using samples from mCRC patients collected in a well-controlled clinical trial and applied an orthogonal validation at the sample level to a single-nucleotide variant genotyping approach. This defined DELFI-TF and demonstrated its potential prognostic ability to detect disease progression in the training set, which outperforms traditional treatment response monitoring approaches. It should be noted that these results need to be confirmed in a validation cohort and then evaluated for other types of cancer or earlier stage disease before clinical application. These results are not directly transferable to other body fluids such as urine and cerebrospinal fluid due to different distribution of cfDNA fragments. In conclusion, we developed a novel quantitative measure of ctDNA burden due to cfDNA fragmentation. Within the training cohort, DELFI-TF appears to be a useful non-invasive approach to monitor treatment success in mCRC patients.

[0125] Study design and population - This study was a retrospective analysis of liquid biopsies collected from a homogenous group of mCRC patients participating in the CAIRO5 prospective clinical trial ( NCT02162563). The randomized phase III CAIRO5 trial investigates the optimal first-line systemic therapy for patients with histologically proven CRC with isolated, untreated, initially unresectable liver metastases. Patients treated with doublet chemotherapy (FOLFOX or FOLFIRI) and bevacizumab, with at least one blood draw before and after treatment, are included in the study. All patients were considered unresectable at enrollment, i.e., R0 resection could not be achieved in one procedure with a single surgical intervention. After treatment with doublet chemotherapy and bevacizumab, patients were evaluated every 2 months by an expert panel of liver surgeons and abdominal radiologists for the possibility of local treatment of colorectal liver metastases according to current clinical practice. Clinical follow-up was performed according to standard of care, including clinical review every 3 months and CT imaging and serum CEA every 6 months. If liver metastases remained unresectable, chemotherapy was continued without targeted agents for a total of 6 months of preoperative and postoperative treatment, and patients were continuously evaluated with serum CEA and CT imaging every 2 months until disease progression. Follow-up was recorded until September 1, 2021. The study was approved by the Medical Ethics Committee and conducted in accordance with the Declaration of Helsinki, and patients signed written informed consent for study participation and blood sampling for translational research.

[0126] Blood collection and cfDNA extraction - Liquid biopsy samples were collected at the medical center of enrollment before study treatment (baseline), before surgery, after surgery, and every 3 months during follow-up until disease progression or completion of treatment. Blood samples were collected in 10 mL cell-free DNA BCT® tubes (Streck, La Vista, USA) and collected centrally at the Netherlands Cancer Institute (Amsterdam, the Netherlands). Cell-free plasma was isolated by a two-stage centrifugation process at 1700×g for 10 min and 20000×g for 10 min. Cell-free plasma was stored at -80°C until further use. cfDNA isolation was performed by QIAsymphony (Qiagen, Germany) with an elution volume of 60 μL. cfDNA concentration was assessed by Qubit™ dsDNA High-Sensitivity Assay (ThermoFisher, Waltham, MA, USA). Aliquots of up to 15 ng were made as input for library construction, and the volume was adjusted to 51 μL with TE buffer as needed. The cfDNA aliquots were shipped to the laboratory at Delfi Diagnostics (Baltimore, MD, USA).

[0127] Library construction and cfDNA sequencing - After arriving at the laboratory, extracted cfDNA was quality checked with a TapeStation 4200 (Agilent Technologies, Santa Clara, CA, USA). NGS libraries were constructed with a maximum cfDNA input of 15 ng by NEBNext DNA Library Prep kit (New England Biolabs, Ipswich, MA, USA) as previously described (19), with four main modifications to the manufacturer's guidelines: (1) the library purification step used an on-beads AMPure XP (Beckman Coulter, Brea, CA, USA) approach to minimize sample loss during the elution and tube transfer steps; (2) the NEBNext End Repair, dA tailing, and adapter ligation enzymes and buffer volumes were appropriately adjusted to match the on-beads AMPure XP strategy; (3) Illumina dual index adapters were utilized for the ligation reactions; (4) cfDNA libraries were amplified for four cycles with Phusion HotStart Polymerase (ThermoFisher, Waltham, MA, USA). The quality of the WGS libraries was determined by 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) or TapeStation 4200 (Agilent Technologies, Santa Clara, CA, USA). A total of 96 dual-indexed cfDNA libraries containing samples with different barcodes were then pooled together in a single lane of an S4 flow cell and subjected to 100 bp paired-end (200 cycles) WGS sequencing on a NovaSeq 6000 (Illumina, San Diego, CA, USA) aiming for 8× coverage per genome. To limit batch effects, libraries of all samples taken from the same individual were generated in the same batch, including duplicate libraries as inter-batch controls and technical replicates of nucleosomal DNA obtained from nuclease-digested human peripheral blood mononuclear cells as intra-batch controls.

[0128] RAS / BRAF Mutation Analysis-RAS and BRAF V600E mutation analysis was performed on tumor tissue DNA according to routine clinical practice for all patients. A subset of patients with RAS / BRAF tumor tissue mutations underwent longitudinal liquid biopsy hotspot mutation analysis and fragmentation analysis by ddPCR (Bio-Rad, Hercules, CA, USA). ddPCR™ KRAS G12 / G13 (#1863506), ddPCR™ KRAS Q61 (#12001626), ddPCR™ KRAS A146T (#10049550), and ddPCR™ BRAF V600 (#12001037) Screening Kits were used according to the manufacturer's instructions, with 9 μL of sample, 11 μL of ddPCR supermix for probes (without dUTP), 1 μL of multiplex analyte, and 1 μL of nuclease-free water. All measurements were performed in duplicate, including blanks (nuclease-free water) and positive controls. Patients with RAS / BRAF not traceable by ddPCR were excluded (Figure 9B). Data were analyzed using QuantaSoft™ software version 1.6.6 (Bio-Rad, Hercules, CA, USA) and the autocorrection algorithm described above.

[0129] Analysis of cfDNA sequencing data - For each sample, paired-end sequenced reads were aligned to the reference genome (hg19) by paired-end alignment with Bowtie (version 2.3.0). Aligned reads were sorted and converted to BAM and then BED formats by Samtools (version 1.3.1) and Bedtools (version 2.26.0), respectively. Fragment lengths were calculated based on start and end coordinates, and fragments were divided into 504 5 Mb bins covering approximately 2.6 Gb of the genome. The number of short fragments (100-150 bp) and the number of long fragments (151-220 bp) per bin were then calculated by R / Bioconductor (version 3.6.2), and these counts were corrected for GC content as described by Benjamini and Speed. The corrected counts of short fragments were divided by the corrected counts of long fragments per bin to obtain a fragmentation profile per individual.

[0130] Four statistics were calculated for each sample to generate the DELFI-TF score, DELFI score, DELFI information, mixture model components, and group-level aneuploidy score. The DELFI score was calculated similarly to the method described by Cristiano et al. (Genome-wide cell-free DNA fragmentation in patients with cancer. Nature 570, 385-389 (2019)) and indicates how similar the fragmentation profile is to individuals with or without cancer. DELFI information is defined as the correlation between the binned and mean-centered short-to-long ratio of a given sample and the binned and mean-centered short-to-long ratio of a "moderately healthy" sample from a reference cohort containing only healthy samples minus one. The mixture model summarizes the fragment-size distribution, and the weight statistic from this model is evaluated during DELFI-TF generation.

[0131] Using these statistics calculated for each sample, a hierarchical Bayesian regression model was trained in R on the allele frequencies of tumor-specific driver RAS / BRAF variants measured by ddPCR in longitudinal cfDNA samples. The model takes DELFI scores, DELFI information, mixture model weights, and aneuploidy scores as inputs and outputs a predicted MAF. The model assumes that MAFs are beta-distributed random variables and assumes that the expected MAF of a given sample is functionally related to the described features by the inverse logit of the feature matrix multiplied by the vector of regression coefficients, with a patient-specific random intercept that accounts for the within-patient correlation between measurements. To generate unbiased predictions, avoid overfitting, and evaluate generalizability, training is performed by leave-one-patient-out cross-validation. In this cross-validation scheme, data from each patient is provided in turn, the model is trained on the remaining samples, and the trained model is then used to generate predictions for the provided samples. DELFI-TF was defined as the predicted MAF from this cross-validation scheme. We evaluate the quality of the generated predictions by determining the correlation of these predictions with the observed ddPCR MAF values ​​and by assessing the relationship of these predictions with time to progression or death.

[0132] DELFI-TF Dynamics Analysis - To capture the molecular dynamics of tumor burden over time, we calculated the DELFI-TF slope, i.e. the slope of the regression line fitted to the DELFI-TF values ​​from time T1 onwards, before progression for PFS analysis and 60 days after progression for OS analysis. In this exercise, we selected patients who had at least three samples taken before progression and at least one of these samples taken within the progression window, which was 120 days before progression for PFS analysis (79 patients) and 120 days before progression to 60 days after progression for OS analysis (80 patients). The regression lines were calculated by Python / scikit-learn (version 3.9.13 / 1.1.1).

[0133] Relative coverage calculation for gene expression analysis - For this analysis, we selected a set of 854 transcripts known to be highly expressed in colon adenocarcinoma identified from the Broad GDAC Firehose Pipeline and extracted their transcription start site (TSS) coordinates. Fragment coverage was calculated for all genes in these TSSs + 1,500 bp flanking regions on each side, only for the 126 patients who had plasma samples taken at both T0 and T1 time points. The TSS coordinates and the list of aligned fragments were in BED format, and coverage calculations were performed with pybedtools (version 0.9.0), a python interface to Bedtools.

[0134] Statistical analysis. Correlations between DELFI-TF and ddPCR MAF were calculated by Pearson's correlation coefficient. Similarly, correlations between DELFI-TF / ddPCR MAF and copy number ratio were calculated by Pearson's correlation. Spearman correlation test was used between DELFI-TF / ddPCR MAF and SLD and serum CEA. All two-sample hypothesis tests, except survival analysis, were performed by Wilcoxon rank sum test. Tumor fractions based on DELFI-TF and MAF between resection status were compared by Kruskal-Wallis test. Survival analysis was performed by Mantel-Cox log-rank test. Analyses were performed with R Statistical Software (version 4.2.1 Foundation for Statistical Computing, Vienna, Austria). Unless otherwise stated, hypothesis tests were two-sided with 5% type 1 error to determine statistical significance.

[0135] Although the invention has been described with reference to the presently preferred embodiments, it should be understood that various modifications can be made without departing from the spirit of the invention. Accordingly, the invention is limited only by the scope of the following claims.

Claims

1. determining a cell-free DNA (cfDNA) fragmentation profile of a sample from the subject; calculating a fragmentation score based on the cfDNA fragmentation profile, said score indicating a likelihood of the presence of cancer in the subject; determining a short to long fragment ratio and a fragment size distribution from the fragmentation profile; calculating an information score based on the short to long fragment ratio in the sample relative to a ratio from a sample from a healthy subject; determining a set of model weights based on the fragment size distribution; training a machine learning model using a set of features extracted from multiple fragmentation profiles of multiple subjects; determining, by the machine learning model, a supervision score for the sample based on the fragmentation score, the information score, and the model weight, the supervision score indicative of a level of tumor-derived nucleic acid in the cfDNA of the sample; 23. A method for monitoring cancer, comprising:

2. The method of claim 1 , wherein the information score indicates a correlation between the fragmentation profile and a median fragmentation profile observed in healthy subjects.

3. The method of claim 1 , wherein the monitoring score has a range of 0 to 1.

4. 10. The method of claim 1, further comprising determining at least one of the subject's likelihood of overall survival, likelihood of progression-free survival, or time to progression based on the monitoring score.

5. 5. The method of claim 4, wherein the subject's likelihood of overall survival, likelihood of progression-free survival, or time to progression decreases with increasing monitoring score value.

6. further comprising classifying the monitoring score as a high score or a low score; A high score indicates at least one of a decreased overall survival, a decreased progression-free survival, or a decreased time to progression in the subject. The method according to claim 5.

7. 7. The method of claim 6, wherein the monitoring score is classified before a cancer treatment is administered to the subject to provide a baseline classification of the subject.

8. 7. The method of claim 6, wherein the monitoring score is classified at the first time point after a cancer treatment is administered to the subject to provide a post-treatment classification of the subject.

9. 2. The method of claim 1, wherein the likelihood of cancer progression in the subject increases with increasing surveillance score value.

10. 10. The method of claim 1, wherein the likelihood of cancer progression in the subject decreases with a decrease in the surveillance score value.

11. 2. The method of claim 1, wherein the likelihood of a positive response to a cancer treatment administered to the subject increases with a decrease in the monitoring score value.

12. 10. The method of claim 1, wherein the likelihood of a positive response to a cancer treatment administered to the subject decreases with increasing monitoring score value.

13. 10. The method of claim 1, wherein the surveillance score predicts a fraction of tumor-derived nucleic acid or a clonal variant allele fraction of a mutation associated with cancer.

14. The method of claim 1 , wherein the cancer is a solid tumor.

15. 10. The method of claim 1, wherein the cancer is a sarcoma, carcinoma, or lymphoma.

16. 2. The method of claim 1, wherein the cancer is selected from the group consisting of colorectal cancer, lung cancer, kidney cancer, brain cancer, prostate cancer, breast cancer, pancreatic cancer, bile duct cancer, liver cancer, CNS cancer, gastric cancer, esophageal cancer, gastrointestinal stromal tumors (GIST), uterine cancer, and ovarian cancer.

17. 10. The method of claim 1, wherein the cancer is a hematological cancer.

18. 2. The method of claim 1, wherein the cancer is selected from the group consisting of myeloma, multiple myeloma, B-cell lymphoma, follicular lymphoma, lymphocytic leukemia, leukemia, and myeloid leukemia.

19. 10. The method of claim 1, further comprising administering to the subject a cancer treatment.

20. 18. The method of claim 17, wherein the cancer treatment is selected from the group consisting of surgery, adjuvant chemotherapy, neoadjuvant chemotherapy, radiation therapy, hormonal therapy, cytotoxic therapy, immunotherapy, adoptive T cell therapy, targeted therapy, or any combination thereof.

21. The cfDNA fragmentation profile may include: obtaining and isolating cfDNA fragments from the subject; Sequencing the cfDNA fragments to obtain sequenced fragments; mapping the sequenced fragments to a genome to obtain a window of mapped sequences; analyzing the windows of mapped sequences to determine cfDNA fragment lengths and generate the cfDNA fragmentation profile; The method of claim 1 , wherein the

22. 1. A method for determining at least one of overall survival, progression free survival, or time to progression, comprising: determining a cell-free DNA (cfDNA) fragmentation profile of a sample from the subject; calculating a fragmentation score based on the cfDNA fragmentation profile, said score indicating a likelihood of the presence of cancer in the subject; determining a short to long fragment ratio and a fragment size distribution from the fragmentation profile in the sample correlating with ratios from a sample from a healthy subject; calculating an information content score based on the short fragment to long fragment ratio; determining a set of model weights based on the fragment size distribution; training a machine learning model using a set of features extracted from the fragmentation profiles of multiple subjects; determining, by the machine learning model, a monitoring score for the sample based on the fragmentation score, the information score, and the model weights, the monitoring score being indicative of a level of tumor origin in the cfDNA of the sample, thereby indicative of a likelihood of cancer progression in the subject; determining at least one of overall survival, progression-free survival, or time to progression based on the monitoring score; A method comprising:

23. 1. A system for monitoring cancer in a subject, comprising: Memory, one or more processors coupled to the memory; Equipped with The one or more processors in the computer system include: determining a cell-free DNA (cfDNA) fragmentation profile of a sample from the subject; and calculating a fragmentation score based on the cfDNA fragmentation profile, said score indicating a likelihood of the presence of cancer in the subject; and determining a short to long fragment ratio and a fragment size distribution from the fragmentation profile in the sample, correlating with ratios from a sample from a healthy subject; calculating an information content score based on said short fragment to long fragment ratio; determining a set of model weights based on the fragment size distribution; training a machine learning model using a set of features extracted from the fragmentation profiles of multiple subjects; determining, by the machine learning model, a monitoring score for the sample based on the fragmentation score, the information score, and the model weight, the monitoring score being indicative of a level of tumor origin in the cfDNA of the sample, the level of tumor origin in the cfDNA being indicative of a likelihood of cancer progression in the subject; configured to perform an operation to cause system.

Images