Transcription factor profiling

Abstract

The methods and systems disclosed herein can improve the ability to analyze genomic material. The methods provided herein can examine transcription factor binding site accessibility to diagnose disease or monitor disease progression in a subject.

Description

[Technical field]

[0001] cross reference This application claims the benefit of U.S. Provisional Patent Application No. 62 / 742,854, filed October 8, 2018, U.S. Provisional Patent Application No. 62 / 752,270, filed October 29, 2018, and U.S. Provisional Patent Application No. 62 / 849,097, filed May 16, 2019, each of which is incorporated herein by reference. [Background technology]

[0002] Transcription factors (TFs) can regulate the expression of target genes and play key roles in development and differentiation. Genomic modifications can lead to the activation or inactivation of TFs, and the resulting disturbances in gene regulation can contribute to physiological conditions such as aging or be the basis of fundamental diseases such as cancer. To bind regulatory deoxyribonucleic acid (DNA), TFs often must interact with nucleosomes, which can affect both their occupancy and positioning.

[0003] Alterations in transcription factors may be important drivers of cancer tumorigenesis, and TF-nucleosome interactions remain largely unmapped. However, there are no non-invasive assays to assess transcription factor activity.

[0004] Given the role of TFs in controlling chromatin accessibility and transcription, understanding the effects of genetic variation on TF binding can provide insight into the non-coding genetic components of progression and disease. Important insights into the epigenetic information encoded within the nucleoprotein structure of chromatin can also be gained by analyzing chromatin accessibility ("open chromatin"), nucleosome positioning, and transcription factor (TF) occupancy in a high-throughput, genome-wide manner that separately measures chromatin accessibility ("open chromatin"), nucleosome positioning, and transcription factor (TF) occupancy.

[0005] Deregulation of transcription factors (TFs) can be a key driver of tumorigenesis. For TFs to bind to DNA, their binding regions must be accessible. Therefore, TFs and chromatin remodeling complexes translocate and position nucleosomes to increase accessibility. Therefore, there is a need for methods to profile transcription factor binding sites to infer nucleosome positioning and chromatin accessibility. Furthermore, there is a need for methods to use transcription factor binding site profiling and transcription factor binding site signatures to infer disease state, disease progression, and therapeutic response. Summary of the Invention

[0006] The present disclosure provides methods and systems for evaluating (e.g., modeling) transcription factor (TF) binding sites (TFBS) and using TFBS information to detect, assess, diagnose, analyze disease states, and ascertain therapeutic responsiveness.

[0007] Genome-wide assays based on next-generation sequencing can be used to provide TF binding patterns and associated chromatin architecture. Nucleosome- and sequence-specific TFs bind to regulatory deoxyribonucleic acid (DNA) regions in a mutually exclusive manner, so that TFs compete with or interact with nucleosomes, affecting both their occupancy and positioning. In a given cell population, nucleosome occupancy refers to the average number of nucleosomes measured within a particular genomic region, and nucleosome positioning refers to the probability that a reference point on a nucleosome (usually the midpoint of a dyad, e.g., a canonical nucleosome) is present at a particular genomic coordinate.

[0008] Cell-free circulating nucleic acids, such as cell-free DNA (cfDNA), can provide a readily accessible source of nucleic acids for TFBS analysis. Such cfDNA can be the product of digestive processes that preferentially degrade DNA that is not protected by proteins such as histone complexes. Cell-free DNA coverage patterns can reflect nucleosome positioning and occupancy caused by transcription factors that actively bind to the genome. These nucleosome occupancy patterns measured via cfDNA can then be used to infer the activity of TFs in normal and tumor genomes.

[0009] TFs bind preferentially within open chromatin and may influence nucleosome positioning. Circulating cell-free DNA from plasma may represent monobasic nucleosomal DNA, and the plasma footprint of nucleosomes may also provide information about TFBS.

[0010] The present disclosure provides methods and systems for charting nucleosome positions from cfDNA to provide information about TFs for applications related to disease identification, prediction, classification, and / or therapeutic response confirmation. Methods and systems are described herein for using transcription factor information determined from nucleosome footprints in nucleic acid molecules (e.g., cfDNA). Information from nucleosome footprints in nucleic acid molecules can be used to evaluate, predict, detect, and diagnose diseases such as cancer. In some examples, the information may be featured and used as input to machine learning models that are useful in many of these applications, such as disease identification, prediction, classification, and therapeutic response confirmation.

[0011] In one aspect, the disclosure provides a computer-implemented method for determining a transcription factor binding profile in a nucleic acid sample from a subject, the method comprising: (a) providing a set of sequencing reads of deoxyribonucleic acid (DNA) extracted from the subject; (b) using the set of sequence reads to generate a coverage pattern for transcription factors; (c) processing the coverage pattern to provide a signal; and (d) processing the signal with a reference signal, wherein the signal and the reference signal have different frequencies, thereby determining the transcription factor binding profile in the sample.

[0012] In some embodiments, the DNA is cell-free DNA.

[0013] In some examples, (c) includes using a low pass filter. In some examples, (c) includes using a Savitzky-Golay filter.

[0014] In some examples, the subject is a human.

[0015] In another aspect, the disclosure provides a computer-implemented method for detecting the presence or absence of a disease in a subject, the method comprising: (a) providing a set of deoxyribonucleic acid (DNA) sequencing reads extracted from a subject; (b) using the set of sequence reads to generate a coverage pattern for a transcription factor; (c) processing the coverage pattern to provide a signal; and (d) processing the signal with a reference signal, wherein the signal and the reference signal have different frequencies, thereby detecting the presence or absence of the disease in the subject.

[0016] In some instances, the DNA is cell-free DNA.

[0017] In some instances, the disease is cancer.

[0018] In some examples, (b) includes aligning the set of sequence reads to a reference sequence to provide an aligned sequence pattern, selecting regions of the aligned sequence pattern that correspond to binding sites of the transcription factor, and normalizing the aligned sequence pattern in the regions. In some examples, (d) includes calculating an accessibility score for each of the binding sites of the transcription factor.

[0019] In some examples, (c) includes using a low pass filter. In some examples, (c) includes using a Savitzky-Golay filter.

[0020] In some examples, the subject is a human.

[0021] In some examples, the transcription factor is a cancer-specific transcription factor. In some examples, the transcription factor is selected from the group consisting of GRH-L2, ASH-2, HOX-B13, EVX2, PU.1, Lyl-1, Spi-B, FOXA1, HNF-1A, HNF-4A, HNF-4G, and DLX-2.

[0022] In some examples, accessibility scores for at least 2, or at least 5, or at least 10, or at least 15, or at least 20, or at least 25 transcription factor binding sites are determined and input into a machine learning model to train a classifier that can distinguish between healthy subjects and cancer patients, disease progressors and non-progressors, multiple disease subtypes, multiple disease stages, disease treatment responders and non-responders, or any combination thereof.

[0023] In some examples, the transcription factor is selected from the group consisting of GRH-L2, ASH-2, HOX-B13, EVX2, PU.1, Lyl-1, Spi-B, and FOXA1.

[0024] In several instances, patient- and / or tumor-specific patterns are observed, including inferred binding patterns for the transcription factors AR, HOXB13, and NKX3-1.

[0025] In some examples, (d) includes identifying an indication of higher accessibility of a transcription factor. In some examples, the transcription factor is an epithelial transcription factor. In some examples, the transcription factor is GRH-L2.

[0026] In some instances, the transcription factors GRHL2, FOXA1, and ZNF121 are associated with increased accessibility scores or open chromatin accessibility in patients with breast cancer.

[0027] In some instances, open accessibility of the transcription factors GRHL2, FOXA1, and ZNF121 is indicative of breast cancer.

[0028] In some instances, transcription factors EVX2, DLX2, HNF1A, HNF4A, GRHL2, and HNF4G are associated with increased accessibility scores or open chromatin accessibility in patients with colon cancer.

[0029] In some instances, open accessibility of transcription factors EVX2, DLX2, HNF1A, GRH-L2, HNF4A, and HNF4G is indicative of colon cancer.

[0030] In some instances, the transcription factors LYL1 and PU.1 are associated with reduced or closed chromatin accessibility in patients with colon cancer.

[0031] In some instances, closed accessibility of the transcription factors LYL1 and PU.1 is indicative of colon cancer.

[0032] In one example, open accessibility of the transcription factors tbx21 or EOMES regulates exhausted CD8+ Shows T cells.

[0033] In one example, open accessibility of selected transcription factors from Eomesodermin (EOMES), Ybx21, Gata3, Rora, Bcl6, Blimb-1, von Hippel-Lindau tumor suppressor gene (VHL), Foxo1, IRF4, BATF, and NFATc1 was shown to mediate the expression of exhausted CD8 + Shows T cells.

[0034] In some examples, the method further comprises detecting the presence or absence of disease in the subject with at least about 70% accuracy. In some examples, the method further comprises detecting the presence or absence of disease in the subject with at least about 80% accuracy. In some examples, the method further comprises detecting the presence or absence of disease in the subject with at least about 90% accuracy.

[0035] In some examples, the method further comprises detecting the presence of disease in the subject with at least about 70% sensitivity. In some examples, the method further comprises detecting the presence of disease in the subject with at least about 80% sensitivity. In some examples, the method further comprises detecting the presence of disease in the subject with at least about 90% sensitivity.

[0036] In some examples, the method further comprises detecting the absence of disease in the subject with at least about 70% specificity. In some examples, the method further comprises detecting the absence of disease in the subject with at least about 80% specificity. In some examples, the method further comprises detecting the absence of disease in the subject with at least about 90% specificity.

[0037] In some instances, the method further comprises detecting the presence of disease in the subject with a positive predictive value of at least about 70%. In some instances, the method further comprises detecting the presence of disease in the subject with a positive predictive value of at least about 80%. In some instances, the method further comprises detecting the presence of disease in the subject with a positive predictive value of at least about 90%.

[0038] In some examples, the method further comprises detecting the absence of disease in the subject with a negative predictive value of at least about 70%. In some examples, the method further comprises detecting the absence of disease in the subject with a negative predictive value of at least about 80%. In some examples, the method further comprises detecting the absence of disease in the subject with a negative predictive value of at least about 90%.

[0039] In some examples, the method further comprises detecting the presence or absence of disease in the subject with an area under the receiver operating characteristic (AUROC) of at least about 0.70. In some examples, the method further comprises detecting the presence or absence of disease in the subject with an area under the receiver operating characteristic (AUROC) of at least about 0.80. In some examples, the method further comprises detecting the presence or absence of disease in the subject with an area under the receiver operating characteristic (AUROC) of at least about 0.90.

[0040] In some examples, the method further comprises applying a trained classifier to the signal to detect the presence or absence of disease in the subject. In some examples, the method further comprises applying a trained classifier to the accessibility score of the binding site of the transcription factor to detect the presence or absence of disease in the subject. In some examples, the trained classifier comprises a trained machine learning classifier. In some examples, the trained machine learning classifier comprises a supervised machine learning algorithm. In some examples, the supervised machine learning algorithm comprises one or more of regression, support vector machine, tree-based method, neural network, and random forest.

[0041] In other aspects, the present disclosure provides methods that allow for classification of patients by tumor type, including tumor subtype (e.g., prostate, colon, breast, lung cancer subtypes) or tumor stage, which may have important clinical implications for patient management, including treatment regimens and responsiveness. Accordingly, provided herein are methods for mapping tumor-specific transcription factors in vivo based on patient samples (e.g., blood, plasma, or serum samples), thereby making key portions of the non-coding genome amenable to clinical analysis.

[0042] In some instances, the method includes identifying a subtype of the disease.

[0043] In some instances, the method includes identifying a subtype of the cancer.

[0044] In some examples, the method includes distinguishing between subtypes of prostate cancer, colon cancer, breast cancer, and lung cancer.

[0045] In some instances, the methods include identifying prostate cancer subtypes, for example, among patients with prostate adenocarcinoma or small cell neuroendocrine prostate cancer.

[0046] In some instances, the method includes identifying a stage of the cancer (eg, from among stages I, II, III, and IV cancer).

[0047] In some instances, the method includes distinguishing stage I and II cancer from stage III and IV cancer.

[0048] In some instances, the transcription factors GRHL2, FOXA1, HOXB13, AR, and NKX3-1 are associated with increased accessibility scores or open chromatin accessibility in patients with prostate adenocarcinoma.

[0049] In some instances, open accessibility of the transcription factors GRHL2, FOXA1, HOXB13, AR, and NKX3-1 is indicative of prostate adenocarcinoma.

[0050] In some instances, the transcription factors REST, GRHL2, FOXA1, HOXB13, AR, and NKX3-1 are associated with reduced or closed chromatin accessibility in patients with small cell neuroendocrine prostate cancer.

[0051] In some instances, decreased accessibility of the transcription factors REST, GRHL2, GRHL3, FOXA1, FOXA2, GATA2, GATA3, HOXB13, AR, and NKX3-1 is indicative of small cell neuroendocrine prostate cancer.

[0052] In some instances, increased accessibility of the transcription factors GLIS1, SOX2, and SOX11 is indicative of small cell neuroendocrine prostate cancer.

[0053] In other aspects, the disclosure provides a system including a computing device including at least one computer processor, an operating system configured to execute executable instructions, a memory, and a computer program including instructions executable by the computing device to provide a computer application for detecting the presence or absence of a disease in a subject, the computer application including: a sequence module programmed to obtain a set of sequencing reads from deoxyribonucleic acid (DNA) extracted from the subject; a coverage module programmed to use the set of sequence reads to generate a coverage pattern for a transcription factor; a signal module programmed to process the coverage pattern to provide a signal; and a detection module programmed to process the signal with a reference signal, where the signal and the reference signal have different frequencies, thereby detecting the presence or absence of a disease in the subject.

[0054] In another aspect, the disclosure provides a non-transitory computer readable medium comprising machine-executable code that, when executed by one or more computer processors, performs a method for detecting the presence or absence of a disease in a subject, the method comprising: (a) providing a set of sequencing reads of deoxyribonucleic acid (DNA) extracted from a subject; (b) using the set of sequencing reads to generate a coverage pattern for a transcription factor; (c) processing the coverage pattern to provide a signal; and (d) processing the signal with a reference signal, wherein the signal and the reference signal have different frequencies, thereby detecting the presence or absence of the disease in the subject.

[0055] In another aspect, the disclosure provides a system for detecting the presence or absence of a disease in a subject, the system comprising: a database comprising a set of sequence reads from deoxyribonucleic acid (DNA) extracted from the subject; and one or more computer processors operably connected to the database, where the one or more computer processors are programmed to, individually or collectively: (a) use the set of sequence reads to generate a coverage pattern for a transcription factor; (b) process the coverage pattern to provide a signal; and (c) process the signal with a reference signal, where the signal and the reference signal have different frequencies, thereby detecting the presence or absence of a disease in the subject.

[0056] In another aspect, the disclosure provides a computer-implemented method for monitoring disease progression or regression in a subject, the method comprising: (a) providing a first set of sequence reads from deoxyribonucleic acid (DNA) extracted from the subject at a first time and a second set of sequence reads from the DNA extracted from the subject at a second time, the second time being later than the first time; (b) using the first set of sequence reads to generate a first coverage pattern for a transcription factor, and (c) using the second set of sequence reads to generate a second coverage pattern for the transcription factor. (c) processing the first coverage pattern to provide a first signal and processing the second coverage pattern to provide a second signal; (d) processing the first signal with a reference signal, where the first signal and the reference signal have different frequencies; (e) processing the second signal with the reference signal, where the second signal and the reference signal have different frequencies; and (f) monitoring disease progression or regression in the subject based on processing the first signal and the second signal with the reference signal.

[0057] In some examples, accessibility scores for at least 2, or at least 5, or at least 10, or at least 15, or at least 20, or at least 25 transcription factor binding sites are determined and input into a machine learning model to train a classifier capable of identifying disease progressors and non-progressors, multiple disease subtypes, multiple disease stages, or any combination thereof.

[0058] In some examples, accessibility scores for at least 2, or at least 5, or at least 10, or at least 15, or at least 20, or at least 25 transcription factor binding sites are determined and input into a machine learning model to train a classifier capable of distinguishing between disease treatment responders and non-responders.

[0059] In some instances, the second coverage pattern indicates a change in the phenotype of the tumor during the course of the disease.

[0060] In some instances, the phenotypic change is from an androgen-dependent cancer to an androgen-independent cancer.

[0061] In some instances, the DNA is cell-free DNA.

[0062] In some instances, the disease is cancer.

[0063] In some examples, (b) includes aligning the first set of sequence reads and the second set of sequence reads to a reference sequence to provide a first aligned sequence pattern and a second aligned sequence pattern, respectively; selecting regions of the first aligned sequence pattern and the second aligned sequence pattern that correspond to binding sites of the transcription factor; and normalizing the first aligned sequence pattern and the second aligned sequence pattern in the regions.

[0064] In some examples, (c) includes using a low pass filter. In some examples, (c) includes using a Savitzky-Golay filter.

[0065] In some examples, the subject is a human.

[0066] In some instances, the transcription factor is a cancer-specific transcription factor.

[0067] In some examples, the transcription factor is selected from the group consisting of GRH-L2, ASH-2, HOX-B13, EVX2, PU.1, Lyl-1, Spi-B, and FOXA1.

[0068] In some examples, the transcription factor is selected from the group consisting of HNF-1a, HNF-4a, HNF-4g, EVX-2, and DLX-2.

[0069] In some examples, the method further includes adjusting a treatment regimen for the subject's disease based on (f).

[0070] In another aspect, the disclosure provides a system including a computing device including at least one computer processor, an operating system configured to execute executable instructions, a memory, and a computer program including instructions executable by the computing device to provide a computer application for monitoring disease progression or regression in a subject, the computer application comprising: a sequence module programmed to obtain a first set of sequence reads from deoxyribonucleic acid (DNA) extracted from the subject at a first time, and a second set of sequence reads from the DNA extracted from the subject at a second time, the second time being later than the first time; using the first set of sequence reads to generate a first coverage pattern for transcription factors, and a second coverage pattern for the transcription factors; a coverage module programmed to use a second set of sequence reads to generate a pattern; a signal module programmed to process the first coverage pattern to provide a first signal and to process the second coverage pattern to provide a second signal; a first processing module programmed to process the first signal with a reference signal, where the first signal and the reference signal have different frequencies; a second processing module programmed to process the second signal with the reference signal, where the second signal and the reference signal have different frequencies; and a detection module programmed to monitor disease progression or regression in the subject based on processing the first signal and the second signal with the reference signal.

[0071] In other aspects, the disclosure provides a non-transitory computer readable medium comprising machine-executable code that, when executed by one or more computer processors, performs a method for detecting disease progression or regression in a subject, the method comprising: (a) providing a first set of sequence reads from deoxyribonucleic acid (DNA) extracted from the subject at a first time, and a second set of sequence reads from the DNA extracted from the subject at a second time, the second time being later than the first time; (b) using the first set of sequence reads to generate a first coverage pattern for a transcription factor, and (c) generating a second coverage pattern for the transcription factor. (c) processing the first coverage pattern to provide a first signal and processing the second coverage pattern to provide a second signal; (d) processing the first signal with a reference signal, where the first signal and the reference signal have different frequencies; (e) processing the second signal with the reference signal, where the second signal and the reference signal have different frequencies; and (f) monitoring disease progression or regression in the subject based on processing the first signal and the second signal with the reference signal.

[0072] In another aspect, the disclosure provides a system for monitoring disease progression or regression in a subject, the system comprising: a database including a first set of sequence reads from deoxyribonucleic acid (DNA) extracted from the subject at a first time and a second set of sequence reads from the DNA extracted from the subject at a second time, the second time being later than the first time; and one or more computer processors operatively connected to the database, wherein the one or more computer processors individually or collectively: (a) use the first set of sequence reads to generate a first coverage pattern for transcription factors, and (b) generate a second coverage pattern for transcription factors. (b) using the second set of sequence reads to generate a coverage pattern; (b) processing the first coverage pattern to provide a first signal and processing the second coverage pattern to provide a second signal; (c) processing the first signal with a reference signal, where the first signal and the reference signal have different frequencies; (d) processing the second signal with the reference signal, where the second signal and the reference signal have different frequencies; and (e) monitoring disease progression or regression in the subject based on processing the first signal and the second signal with the reference signal.

[0073] In another aspect, the disclosure provides a system for determining a transcription factor binding profile in a nucleic acid sample from a subject, the system including a processor configured to: (a) analyze a set of sequencing reads of deoxyribonucleic acid (DNA) extracted from the subject; (b) use the set of sequencing reads to generate a coverage pattern for transcription factors; (c) process the coverage pattern to provide a signal; and (d) process the signal with a reference signal, where the signal and the reference signal have different frequencies, thereby determining a transcription factor binding profile in the nucleic acid sample.

[0074] In some instances, the DNA is cell-free DNA.

[0075] In some examples, (c) includes using a low pass filter. In some examples, (c) includes using a Savitzky-Golay filter.

[0076] In some examples, the subject is a human.

[0077] In another aspect, the disclosure provides a system for detecting the presence or absence of disease in a subject, the system comprising a processor configured to: (i) use a set of sequence reads from deoxyribonucleic acid (DNA) extracted from the subject to generate a coverage pattern for a transcription factor; (ii) process the coverage pattern to provide a signal, the signal having a different frequency from a reference signal; and (iii) process the signal with the reference signal, thereby detecting the presence or absence of disease in the subject.

[0078] In some examples, the disclosure provides a system for classifying tumors by tumor subtype or tumor stage, the system including a processor configured to: (i) use a first set of sequence reads from deoxyribonucleic acid (DNA) extracted from a subject at a first time and a second set of sequence reads extracted from the subject's DNA at a second time, later than the first time, to generate a first coverage pattern for transcription factors and a second coverage pattern for the transcription factors; (ii) process the first coverage pattern to provide a first signal and process the second coverage pattern to provide a second signal, wherein the first signal and the second signal have different frequencies from a reference signal; and (iii) process the first signal with the reference signal and process the second signal with the reference signal to monitor disease progression or regression of the subject.

[0079] In another aspect, the disclosure provides a system for monitoring disease progression or regression in a subject, the system including a processor configured to: (i) use a first set of sequence reads from deoxyribonucleic acid (DNA) extracted from the subject at a first time and a second set of sequence reads extracted from the subject's DNA at a second time, later than the first time, to generate a first coverage pattern for a transcription factor and a second coverage pattern for the transcription factor; (ii) process the first coverage pattern to provide a first signal, and process the second coverage pattern to provide a second signal, wherein the first signal and the second signal have different frequencies from a reference signal; and (iii) process the first signal with the reference signal, and process the second signal with the reference signal, to monitor disease progression or regression in the subject.

[0080] Another aspect of the present disclosure provides a non-transitory computer readable medium comprising machine executable code that, when executed by one or more computer processors, performs any of the methods described above or anywhere herein.

[0081] Another aspect of the present disclosure provides a system including one or more computer processors and a computer memory coupled thereto, the computer memory including machine executable code that, when executed by the one or more computer processors, performs any of the methods described above or anywhere herein.

[0082] In other aspects, the disclosure provides a method for determining a tumor-specific TFBS pattern, the method comprising: (a) providing a first set of sequence reads from deoxyribonucleic acid (DNA) extracted from a subject at a first time and a second set of sequence reads from the DNA extracted from the subject at a second time, the second time being later than the first time; (b) using the first set of sequence reads to generate a first coverage pattern for transcription factors, and using the second set of sequence reads to generate a second coverage pattern for the transcription factors; (c) processing the first coverage pattern to provide a first signal and processing the second coverage pattern to provide a second signal; (d) processing the first signal with a reference signal, where the first signal and the reference signal have different frequencies; (e) processing the second signal with the reference signal, where the second signal and the reference signal have different frequencies; and (f) determining a tumor-specific TFBS pattern based on the processing of the first signal and the second signal with the reference signal.

[0083] In another aspect, the disclosure provides a system including a computing device including at least one computer processor, an operating system configured to execute executable instructions, a memory, and a computer program including instructions executable by the computing device to provide a computer application for determining tumor-specific TFBS patterns, the computer application including: a sequence module programmed to obtain a first set of sequence reads from deoxyribonucleic acid (DNA) extracted from the subject at a first time, and a second set of sequence reads from the DNA extracted from the subject at a second time, the second time being later than the first time; using the first set of sequence reads to generate a first coverage pattern for transcription factors, and using the second coverage pattern for transcription factors; a coverage module programmed to use the second set of sequence reads to generate a tumor-specific TFBS pattern; a signal module programmed to process the first coverage pattern to provide a first signal and to process the second coverage pattern to provide a second signal; a first processing module programmed to process the first signal with a reference signal, where the first signal and the reference signal have different frequencies; a second processing module programmed to process the second signal with the reference signal, where the second signal and the reference signal have different frequencies; and (f) a detection module programmed to determine a tumor-specific TFBS pattern based on processing the first signal and the second signal with the reference signal.

[0084] In other aspects, the disclosure provides a non-transitory computer-readable medium comprising machine-executable code that, when executed by one or more computer processors, performs a method for determining tumor-specific TFBS patterns, the method comprising: (a) providing a first set of sequence reads from deoxyribonucleic acid (DNA) extracted from a subject at a first time, and a second set of sequence reads from the DNA extracted from the subject at a second time, the second time being later than the first time; (b) using the first set of sequence reads to generate a first coverage pattern for transcription factors, and generating a second coverage pattern sequence for the transcription factors. using a second set of sequence reads to generate reads; (c) processing the first coverage pattern to provide a first signal and processing the second coverage pattern to provide a second signal; (d) processing the first signal with a reference signal, where the first signal and the reference signal have different frequencies; (e) processing the second signal with the reference signal, where the second signal and the reference signal have different frequencies; and (f) determining a tumor-specific TFBS pattern based on the processing of the first signal and the second signal with the reference signal.

[0085] In another aspect, the disclosure provides a system for monitoring disease progression or regression in a subject, the system comprising: a database including a first set of sequence reads from deoxyribonucleic acid (DNA) extracted from the subject at a first time and a second set of sequence reads from the DNA extracted from the subject at a second time, the second time being later than the first time; and one or more computer processors operatively connected to the database, wherein the one or more computer processors individually or collectively: (a) use the first set of sequence reads to generate a first coverage pattern for transcription factors, and (b) generate a second coverage pattern for the transcription factors. (b) using the second set of sequence reads to generate a coverage pattern; (b) processing the first coverage pattern to provide a first signal and processing the second coverage pattern to provide a second signal; (c) processing the first signal with a reference signal, where the first signal and the reference signal have different frequencies; (d) processing the second signal with the reference signal, where the second signal and the reference signal have different frequencies; and (e) determining a tumor-specific TFBS pattern based on the processing of the first signal and the second signal with the reference signal.

[0086] Another aspect of the present disclosure provides a non-transitory computer readable medium comprising machine executable code that, when executed by one or more computer processors, performs any of the methods described above or anywhere herein.

[0087] Another aspect of the present disclosure provides a system including one or more computer processors and a computer memory coupled thereto, the computer memory including machine executable code that, when executed by the one or more computer processors, performs any of the methods described above or anywhere herein.

[0088] Further aspects and advantages of the present disclosure will become readily apparent to those skilled in the art from the following detailed description, in which only exemplary embodiments of the present disclosure have been shown and described. As will be realized, the present disclosure is capable of other and different examples, and its various details can be modified in various obvious respects, all without departing from the present disclosure. Accordingly, the drawings and description are to be regarded as illustrative in nature, and not as restrictive.

[0089] INCORPORATION BY REFERENCE All publications, patents, and patent applications mentioned in this specification are incorporated herein by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually incorporated herein by reference. To the extent that a patent publication, patent, or patent application incorporated by reference conflicts with the disclosure contained herein, the present specification is intended to supersede and / or take precedence over such conflicting material. [Brief description of the drawings]

[0090] The features of the present invention are set forth with particularity in the appended claims. A better understanding of the features and advantages of the present method and system will be obtained by reference to the following detailed description that sets forth illustrative embodiments in which the principles of the method and system are utilized, and the accompanying drawings (also referred to herein as "Figures" or "FIGs.") in which: [Figure 1] A computer system is shown that is programmed or otherwise configured to perform the methods of the disclosure, such as storing, processing, verifying, or interpreting subject (e.g., patient) data, biological data, biological sequences, reference sequences, transcription factor (TF) binding site (TFBS) data, or TFBS features such as Z-scores or TFBS accessibility scores. [Figure 2A]Establishment of TF-nucleosome interactions from cell-free deoxyribonucleic acid (cfDNA). Figure 2A shows that regions with highly organized, e.g., phased, nucleosomes provide oscillating read depth patterns, with peaks of reads indicating the position of the dyad, e.g., the midpoint of the canonical nucleosome. Poorly defined nucleosome positions result in a fairly flat coverage profile. [Figure 2B] Establishment of TF-nucleosome interactions from cell-free deoxyribonucleic acid (cfDNA). Figure 2B shows that TFBS data for 676 TFs were retrieved from the GTRD and aligned to a curated list of known or likely human TFs. Three different calculations of increasing stringency were performed. [Figure 2C] Establishment of TF-nucleosome interactions from cell-free deoxyribonucleic acid (cfDNA). Figure 2C shows that the coverage pattern of CCCCTC-binding factors (CTCF) is similar across all cfDNA analyzed, which is in agreement with the DNase hypersensitivity data showing nearly equal accessibility in blood (GM12878) and epithelial tissues, e.g., prostate (LNCaP) and colon (HCT116). In this and each subsequent panel, profiles calculated from healthy controls are shown in grey, while patient-derived profiles are displayed in the indicated colors. [Figure 2D] Establishment of TF-nucleosome interactions from cell-free deoxyribonucleic acid (cfDNA). Figure 2D shows that the hematopoietic lineage specificity of TFs (PU.1, LYL1, SPIB) was confirmed by DNA hypersensitivity assays and their amplitudes were decreased in plasma from cancer patients compared to healthy controls. In contrast, the amplitude of epithelial TF GRHL2 was increased in cfDNA from cancer patients. [Figure 2E]Figure 2D shows the establishment of TF-nucleosome interactions from cell-free deoxyribonucleic acid (cfDNA). Figure 2E shows the accessibility plot and DNase hypersensitivity for the TF FOXA1, illustrating preferential amplitude changes in patients with hormone-dependent cancers, e.g., prostate and breast cancer. [Diagram 3]Accessibility scores for TFBS characterization. A shows how TF accessibility is determined. To measure TF accessibility, the observed raw coverage signal (purple in the left panel, black in the right panel) was split into low-frequency (red) and high-frequency (blue) signals using different window sizes by Savitzky-Golay filtering. The right panel shows an overlay of these three signals. The high-frequency signal is used as an index of accessibility. 3B shows that the extent of the high-frequency signal (Y-axis) depends critically on the number of TFBSs (X-axis), since TFs with fewer binding sites have more noise due to less averaging. To correct for this bias, a LOESS model is fitted (blue). C shows a heatmap of the period along the region around the GRHL2:GRHL2 TFBS (left panel). The color code represents the quantiles of the signal power distribution. Average power of the period of the transcription factor GRHL2 (right panel). D shows the detrended original (black) and reconstructed (red) nucleosome coverage profiles of the transcription factor GRHL2 resulting from wavelet analysis.E shows that all tested procedures (left: ≥50%-TFBSs, Savitzky-Golay filtering; center: power sum, wavelet analysis; right: 1,000-ms TFBSs, Savitzky-Golay filtering) showed increased values ​​as a measure of accessibility for transcription factors expressed in blood (>10 FPKM), but not for genes with no or low indication of expression (<0.1 FPKM). F shows that transcription factors with average DNase hypersensitivity coverage >2 in GM12878 DNase data from the ENCODE project have higher regulation ranges and higher summation than factors with average coverage <1 in all three analyses performed (left: ≥50%-TFBS, Savitzky-Golay filtering; center: power sum, wavelet analysis; right: 1,000-ms TFBS, Savitzky-Golay filtering). [Figure 4]Prostate lineage-specific TFs, their plasticity, and suitability for tumor classification are shown. A shows that prostate adenocarcinomas are AR-dependent and accordingly have frequently increased PSA (prostate-specific antigen) levels and normal NSE (neuron-specific enolase) values. In contrast, t-SCNCs are no longer AR-dependent and usually have low PSA and increased NSE values. Several TFs involved in the transdifferentiation process from adenocarcinoma to t-SCNCs were identified and are indicated by arrows. B shows the accessibility profile of the prostate lineage-specific homeobox TF HOXB13 and the respective DNase hypersensitivity assay. In this and the subsequent panels, profiles calculated from healthy controls are shown in grey, while patient-derived profiles are displayed in the indicated colors. C shows the accessibility pattern and DNA hypersensitivity assay of NKX3-1, one of the earliest genes expressed during prostate epithelial maturation. D shows AR accessibility for all AR binding sites in the GTRD in addition to AR binding sites with higher binding strength in tumors (T-ARBSs) and sites with higher binding strength in normal samples (normal AR binding sites, N-ARBSs) (Pomerantz et al., 2015). The well-established lineage specificity of AR was confirmed by DNA hypersensitivity assays. E shows changes in coverage patterns during transdifferentiation from adenocarcinoma to neuroendocrine carcinoma established from two plasma samples from patient P148 for hormone-dependent (AR, FOXA1), tissue identity-specific (HOXB13, NKX3-1), and neuroendocrine reprogramming (REST, N-MYC) TFs. F shows analysis of the same TFs as in A from four plasma samples from a patient with neuroendocrine prostate cancer. [Diagram 5] Somatic copy number alterations (SCNAs) in plasma samples from cancer patients are shown. SCNAs were identified after whole genome sequencing of eight plasma samples from four patients (C2, P40, P147, and P148). [Figure 6]Figure 17 shows TF-nucleosome interaction maps of 676 high-confidence TFs with reliable binding site information. A shows the TFBS-nucleosome coverage profiles of two representative TFs, CREM and GATAD1, established from 24 cfDNA samples from healthy subjects, each shown as an individual blue line. MNase-seq coverage patterns from lymphoblastoid cell line GM12878 obtained from ENCODE are shown in red. Additional MNase plots are shown in Figure 17. B shows a heat map of fragment sizes around CTCF binding sites, plotting the length of each sequencing read (Y axis) as a function of the distance from the fragment midpoint of each annotated feature to the center of the site (X axis). C shows a heat map of individual CTCF binding sites and their surrounding regions. Regions are ordered by coverage within the central 50 base pairs (bp) around the TFBS. The spatial density of cfDNA fragments within 1 kilobase (kb) regions centered on TFBSs was calculated and ranked. D shows a matrix of overlap between TFBSs (left panel: all 676 GRTD TFs, right panel: 505 TFs with 1,000ms-TFBS). Each dot represents the percentage of overlap (within about 50bp) in the definition of a binding site. E shows that TFBS analysis with high molecular weight DNA, but not mononucleosomal DNA, yields a uniform and unshakeable pattern (blue) in contrast to plasma DNA (green). [Figure 7] TF-nucleosome interaction maps for various TFs are shown, along with additional comparison of cfDNA and MNase-seq coverage profiles around transcription factor binding sites. [Figure 8A] The shapes of TFBSs are shown in Figure 8A. Figure 8A shows the coverage profiles of the TFs AP-4 and BCL-3 after calculations performed separately for TFBSs inside and outside the TSS. [Figure 8B]TFBS shapes. Figure 8B shows that the analysis of TFBSs for the TFs ATF1, CREB, CREM, and ATF-3 can result in equidistant or TSS-like coverage patterns depending on whether all tissues in the GTRD are included or, more strictly, whether only peaks supported by 50% or more of the maximum number of samples (>50%-TFBS) are included. [Figure 8C] Figure 8C shows examples of TF-nucleosome profiles calculated for all and >50% TFBSs (upper panel) and for 1,000-ms TFBSs (lower panel), illustrating the variable nucleosome patterns of different TFs in cfDNA. [Figure 8D] Figure 8D shows that measurements of the widths of TFBSs revealed substantial differences between various TFBSs. [Figure 8E] The shapes of TFBSs are shown in Figure 8E. Box plots showing the percentage of overlap between CpG islands (left panel) and TSSs (right panel). [Figure 9] Analysis of pooled shallow coverage cfDNA. Accessibility of epithelial TF GRHL2 and hematopoietic TFs (PU.1, LYL1, and SPIB) is shown for pooled cfDNA samples from prostate (n=69), colon (n=100), and breast (n=60) cancer cases. Accessibility of lineage-specific TFs AR, HOXB13, and NKX3-1 within the prostate cancer cfDNA pool is also shown. [Figure 10] Figure 1 shows transcription factors involved in transdifferentiation of adenocarcinoma to t-CNC. A shows GRHL2 accessibility in plasma samples P148_1 and P148_3 from patient P148. B shows analysis of GLIS1 in two plasma samples from patient P148. [Figure 11]Downsampling of plasma samples P148_1 and P148_3 from patient P148 is shown. Plasma samples P148_1 (819,607,690 reads) and P148_3 (768,763,081 reads) were downsampled to approximately 50 million reads and analyzed for 1,000ms TFBS (left column), total TFBS, and >50% TFBS (right column). Analysis showed that TFs with fewer TFBSs were preferentially affected by increased noise. [Figure 12] FIG. 1 shows a comparison of TFBS accessibility in consecutive analyses. Plot of correlation between consecutive samples from patients C2, P147, P40, and P148. The X-axis represents the first plasma sample and the Y-axis represents the second plasma sample. [Figure 13] Establishment of TF-nucleosome interactions. A shows TFBS-nucleosome profiles for four TFs, e.g., SP1 and SP2, which mostly bind to common sites in the genome and further co-bind with NF-YA and NF-YB. B shows TF-nucleosome interactions depicted as average nucleosome occupancy profiles established from plasma DNA shown for the hematopoietic cell lineage-specific TFs PU.1, LYL1 and SPIB, and the epithelial cell-specific TF GRHL2. The different amplitudes may reflect the contribution of DNA released from hematopoietic and epithelial cells into the circulation. [Figure 14] We show that CTCF is a special example for the characterization of different TFBSs and show the accessibility scores for the characterization of TFBSs. A shows the various binding sites of CTCF associated with TADs or TSSs. The coverage patterns of CTCF are divided into CTCF sites that are overlapping (red) or outside the TAD boundaries (orange), CTCF sites that are proximal (e.g., within about 2 kbp; green) or distal (more than 2 kbp; blue) to the TSS, and ultraconserved CTCF sites (black) for the full GTRD dataset (left panel) and for only peaks supported in more than 50% of the maximum number of analyzed samples (right panel). B shows the TF-nucleosome profiles showing the diversity of the patterns. [Figure 15] Figure 1 shows that the oncogenes c-Jun (upper panel) and JunD (lower panel) showed increased accessibility only in CRC patient C2, with relative colon specificity confirmed by DNA hypersensitivity assays. [Figure 16] Figure 1 shows the accessibility of TFs that changes during prostate cancer transdifferentiation. In particular, A-C shows the change in coverage pattern after neuroendocrine differentiation in sample P148. A shows the prominent nucleosome phasing changes at androgen receptor binding sites at sites defined by GTRD and at tumor-specific AR binding sites (defined by Pomerantz et al.). B shows that other transcription factors of the AR axis also show a dramatic reduction in nucleosome phasing. The phasing is prominent in sample 1 but almost disappears in sample 3. C shows that repressors (ZNF644, REST) ​​that play a role in neurogenesis are almost inactivated in sample 3. [Figure 17] Figure 1 shows TF-nucleosome interaction maps for 676 high-confidence TFs with reliable binding site information. TF-nucleosome profiles are sorted according to their accessibility scores and number of TFBSs. [Figure 18] Analysis of prostate cancer cases with tandem duplicator phenotype or chromothripsis, including four cfDNA samples with tandem duplicator phenotype (P21_2, P111_1, P111_4, and P166_1) and one case with chromothripsis on chromosome 10 (P143_3), shows accessibility of epithelial TFs FOXA1, GRHL1, and GRHL2, and prostate lineage-specific TFs AR, HOXB13, and NKX3-1. [Figure 19] Analysis of AR binding sites in plasma samples from P40, a prostate cancer patient undergoing ADT treatment, resulting in high levels of AR amplification between samples P40_1 and P40_2. [Figure 20]Plot showing how epigenetic regulatory regions affect nucleosome positioning. Histone modifications and enhancers are illustrated. [Figure 21] Nucleosome positioning of selected TFs is shown. [Figure 22] The coverage patterns of selected TFs are shown. [Diagram 23] The overlap of different TFs is shown. These overlap values ​​correspond to the heatmap in Figure 2C. [Figure 24] The effect of TFBS size is shown. [Diagram 25] Nucleosome patterns of REST and KLF16 for samples from 24 healthy individuals. Each line represents a different individual. In the 24 healthy individuals, patterns emerge in a nearly identical configuration for transcription factors active in blood cells. [Figure 26] Nucleosome positions of selected TFs in late-stage cancer samples are shown. CTCF patterns are similar in all samples. Activity of blood-specific TFs, including PU.1, Lyl-1, and Spi-B, is decreased in cancer samples. Cancer-specific TFs, including GRH-L2 (epithelial marker), ASH-2 and HOX-B13 (prostate cancer markers of the androgen receptor axis), and EVX2 (colon cancer marker), are more active. [Figure 27-1] A list of TFs that can be used in the methods and systems provided herein is included. [Figure 27-2] A list of TFs that can be used in the methods and systems provided herein is included. [Figure 28A] Included in the TFs that may be used in the methods and systems provided herein are Figure 28A, which includes TFs with binding sites larger than 300 bp, while Figure 28B, which includes TFs with binding sites closer to dinucleosome size (between 312-352 bp). [Figure 28B]Included in the TFs that may be used in the methods and systems provided herein are Figure 28A, which includes TFs with binding sites larger than 300 bp, while Figure 28B, which includes TFs with binding sites closer to dinucleosome size (between 312-352 bp). [Figure 29] Identification of transcription factors with altered accessibility in plasma samples from cancer patients. A provides a TFBS analysis of a plasma sample from a healthy donor (NPH001). Each dot represents a TF, the Y-axis displays the accessibility value, and the X-axis illustrates the overall Z-score as a measure of deviation of accessibility from a normal control sample. In samples from healthy donors (compared to all remaining healthy donors), few TFs exceeded a z-score of ±3 (grey dotted line), and no TFs exceeded the threshold of a z-score of ±5 (red line). B provides an overall z-score plot as in A, but with a plasma sample derived from a patient with prostate cancer (P40). C provides an overall z-score plot as in A, for plasma sample C2_6. D provides nucleosome position profiles for the TF EVX2 from plasma DNA of a healthy subject (grey profile) and from two plasma samples derived from patient C2 with colon cancer (blue and red). E provides a bar graph of the global Z-score plot of the combined breast, prostate, and colon cancer pools. The left panel displays TFs with increased accessibility in at least one tumor entity; the right panel summarizes the accessibility of hematopoiesis-related TFs. [Diagram 30] We provide graphs showing TF-based plasma separation limits and early cancer detection. A is a graph showing a comparison of the accessibility of selected TFs in subsamples of the COAD cohort based on their tumor fraction. B is a graph showing logistic regression with all 504 TFs for samples from stage I (left panel) and stage II (right panel) colon cancer cohorts, respectively. All presented results are cross-validated test set values. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0091] While embodiments of the present invention have been shown and described herein, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions may occur to those skilled in the art without departing from the invention. It is understood that various alternatives to the embodiments of the invention described herein may be utilized.

[0092] Where values ​​are listed as ranges, such disclosure will be understood to include disclosure of all possible subranges within such ranges, as well as specific numerical values ​​falling within such ranges, whether or not a specific numerical value or subrange is explicitly stated.

[0093] As used herein, the term "accessibility score" generally refers to a measure of the accessibility of each transcription factor (TF) binding site. The ranking value is called "accessibility score" because the binding of a transcription factor may open or "prime" its target enhancer without necessarily activating it itself. The accessibility score can be used to objectively compare the accessibility of TFBSs in consecutive analyses from the same person or in different individuals. This score reliably evaluates the accessibility of TFBSs and is particularly useful when cfDNA is used for clinical diagnosis, cancer detection, and treatment monitoring.

[0094] As used herein, the term "aligned sequence pattern" generally refers to the spatial pattern of sequence reads after alignment to a reference genome.

[0095] The term "circulating free DNA" or "cell-free DNA" (cfDNA) as used herein generally refers to deoxyribonucleic acid (DNA) that was first detected in human plasma in 1948. (Mandel, P. Metais, P., CR Acad. Sci. Paris, 142, 241-243 (1948)) Since then, its association with disease has been established in several areas. (Tong, YKLo, YM, Clin Chim Acta, 363, 187-196 (2006)) Research has shown that much of the circulating nucleic acid in blood originates from necrotic or apoptotic cells (Giacona, MB, et al., Pancreas, 17, 89-97 (1998)), and in diseases such as cancer, a significant increase in the level of nucleic acid from apoptosis has been observed. (Giacona, MB, et al., Pancreas, 17, 89-97 (1998); Fournie, GJ, et al., Cancer Lett, 91, 221-227 (1995)). In particular, in the case of cancer, circulating DNA carries characteristic signs of the disease, including mutations in cancer genes, microsatellite alterations, and in the case of certain cancers, viral genomic sequences, DNA or RNA in plasma are increasingly being investigated as possible biomarkers of the disease. 16266-16271 (2008)).

[0096] The cell-free fraction may be blood serum or plasma. As used herein, the term "cell-free fraction" of a biological sample generally refers to a fraction of the biological sample that is substantially free of cells. As used herein, the term "substantially free of cells" generally refers to a preparation from a biological sample that contains less than about 20,000 cells per mL, less than about 2,000 cells per mL, less than about 200 cells per mL, or less than about 20 cells per mL. Genomic DNA may not be excluded from a cell-free sample and typically comprises about 50% to about 90% of the nucleic acid present in the sample.

[0097] As used herein, the term "coverage pattern" generally refers to the spatial arrangement of sequencing reads after alignment to a reference genome. The coverage pattern specifies the extent and depth of coverage of a next-generation sequencing method.

[0098] As used herein, the term "derived from" generally refers to an origin or source and may include naturally occurring, recombinant, unpurified, or purified molecules. A nucleic acid derived from an original nucleic acid may contain some or all of the original nucleic acid, or may be a fragment or variant of the original nucleic acid. A nucleic acid derived from a biological sample may be purified from the sample.

[0099] As used herein, the term "diagnosing" or "diagnosis" of a condition or outcome generally refers to predicting or diagnosing a condition or outcome, determining a predisposition to a condition or outcome, monitoring the treatment of a subject (e.g., a patient), diagnosing a subject's (e.g., a patient's) response to treatment, and the prognosis, progression, and response to a particular treatment of a condition or outcome.

[0100] As used herein, the term "nucleic acid" generally refers to a polynucleotide comprising two or more nucleotides. It may be DNA or RNA. A nucleic acid may be a polymeric form of nucleotides of any length, deoxyribonucleotides (dNTPs), or ribonucleotides (rNTPs), or analogs thereof. A nucleic acid may have any three-dimensional structure and may perform any function, unknown or known. Non-limiting examples of nucleic acids include deoxyribonucleic acid (DNA), ribonucleic acid (RNA), coding or non-coding regions of a gene or gene fragment, loci defined from linkage analysis, exons, introns, messenger RNA (mRNA), transfer RNA, ribosomal RNA, small interfering RNA (siRNA), short hairpin RNA (shRNA), microRNA (miRNA), ribozymes, cDNA, recombinant nucleic acids, branched nucleic acids, plasmids, vectors, isolated DNA of any sequence, isolated RNA of any sequence, nucleic acid probes, and primers. A nucleic acid may also include one or more modified nucleotides, such as methylated nucleotides and nucleotide analogs. If present, modifications to the nucleotide structure may be made before or after assembly of the nucleic acid. The sequence of nucleotides of the nucleic acid may be interrupted by non-nucleotide moieties. The nucleic acid may be further modified after polymerization, such as by conjugation or binding to a reporter agent. A "variant" nucleic acid is a polynucleotide having nucleotides identical to the nucleotide sequence of its original nucleic acid, except that at least one nucleotide has been modified, e.g., deleted, inserted, or substituted, respectively. A variant may have a nucleotide sequence having at least about 80%, 90%, 95%, or 99% identity to the nucleotide sequence of the original nucleic acid.

[0101] As used herein, the term "target nucleic acid" generally refers to a nucleic acid molecule in a starting population of nucleic acid molecules having a nucleotide sequence whose presence, amount, and / or sequence, or one or more changes therein, are desired to be determined. The target nucleic acid may be any type of nucleic acid, including DNA, RNA, and analogs thereof. As used herein, "target ribonucleic acid (RNA)" generally refers to a target nucleic acid that is RNA. As used herein, "target deoxyribonucleic acid (DNA)" generally refers to a target nucleic acid that is DNA.

[0102] As used herein, the terms "amplify" and "amplification" generally refer to increasing the size or amount of a nucleic acid molecule. A nucleic acid molecule can be single-stranded or double-stranded. Amplification can include generating one or more copies of a nucleic acid molecule or "amplified product". Amplification may be performed, for example, by extension (e.g., primer extension) or ligation. Amplification can include performing a primer extension reaction to generate a strand complementary to a single-stranded nucleic acid molecule, and in some cases, generating one or more copies of that strand and / or the single-stranded nucleic acid molecule. The term "DNA amplification" generally refers to generating one or more copies of a DNA molecule or "amplified DNA product". The term "reverse transcription amplification" generally refers to generating deoxyribonucleic acid (DNA) from a ribonucleic acid (RNA) template through the action of reverse transcriptase.

[0103] The term "transcription factor" generally refers to a protein that controls the rate of transcription of genetic information from DNA to messenger RNA by binding to specific DNA sequences. Transcription factors are proteins that bind to DNA regulatory sequences (e.g., enhancers and silencers) that are typically located in the 5' upstream region of a target gene to regulate the rate of gene transcription. This can result in increased or decreased gene transcription, protein synthesis, and subsequent changes in cellular function (e.g., cells change in response to their environment (normal or pathological), such as during atrophy, hypertrophy, hyperplasia, metaplasia, or dysplasia). As used herein, specific transcription factors are referred to by nomenclature, although other synonyms may also be used for the transcription factors described herein.

[0104] The term "transcription factor binding profile" generally refers to the multifactorial information profile of a given transcription factor, including both its contribution to tissues and biological processes. TFBPs further include an "accessibility score" and a z-score statistic for objectively comparing significant changes between plasma samples with different TFBS accessibility. This profile allows the identification of lineage-specific TFs suitable for both tissue-of-origin and tumor-of-origin identification.

[0105] As used herein, the term "subject" generally refers to an individual, entity, or medium that has or is suspected of having testable or detectable genetic information or material. A subject may be a human, individual, or patient. A subject may be a vertebrate, such as a mammal. Non-limiting examples of mammals include humans, monkeys, livestock animals, sport animals, rodents, and pets. A subject may exhibit symptoms indicative of the subject's health or physiological state or condition, such as the subject's cancer or stage of cancer. Alternatively, a subject may be asymptomatic with respect to such health or physiological state or condition.

[0106] As used herein, the term "sample" generally refers to a biological sample obtained or derived from one or more subjects. A biological sample may be a cell-free biological sample or a substantially cell-free biological sample, or may be processed or fractionated to generate a cell-free biological sample. For example, a cell-free biological sample may include cell-free ribonucleic acid (cfRNA), cell-free deoxyribonucleic acid (cfDNA), cell-free protein, and / or cell-free polypeptide. A biological sample may be tissue (e.g., tissue obtained by biopsy), blood (e.g., whole blood), plasma, serum, sweat, urine, saliva, or derivatives thereof. A cell-free biological sample may be obtained or derived from a subject using an ethylenediaminetetraacetic acid (EDTA) collection tube, a cell-free RNA collection tube (e.g., Streck), or a cell-free DNA collection tube (e.g., Streck). A cell-free biological sample may be obtained from a whole blood sample by fractionation. A biological sample or derivatives thereof may include cells. For example, the biological sample may be a blood sample or a derivative thereof (eg, blood collected by a blood collection tube or blood drop), a tumor sample, a tissue sample, a urine sample, or a cell (eg, tissue) sample.

[0107] The present disclosure provides methods and systems for transcription factor (TF) binding site (TFBS) modeling and using TFBS information to detect, evaluate, diagnose, and analyze disease states. cfDNA represents a unique analyte generated by endogenous physiological processes to create an in vivo map of nucleosome occupancy by whole genome sequencing. Nucleosome occupancy at transcription factor binding sites (TFBS) can be exploited to infer genes expressed by cells releasing DNA into circulation. Nucleosome occupancy in cfDNA can also reflect the footprint of TFs.

[0108] I. Transcription factor binding site / nucleosome occupancy analysis Although next-generation sequencing can provide important information about TFs, non-invasive methods to measure TF activity and its changes during treatment (e.g., from blood) are needed. Cell-free DNA (cfDNA) (e.g., from plasma) may also contain circulating tumor DNA (ctDNA) in cancer patients, providing an opportunity for non-invasive diagnostic strategies in cancer patients. cfDNA may circulate mostly as mononucleosomal DNA, as it is released after enzymatic degradation from apoptotic cells. Whole-genome sequencing of cfDNA fragments may therefore enable the generation of nucleosome maps in which dyads at sites with high nucleosomal preference, e.g., the midpoint of the canonical nucleosome, result in strong peaks of reads, while dyads with less preferentially placed nucleosomes have reduced or no peaks.

[0109] Since inference of TF binding from cfDNA has enormous diagnostic potential in cancer and beyond, an improved and optimized bioinformatics pipeline was developed. This process can resolve those components involved in nucleosome signatures in TFBSs to objectively assess and compare TFBS accessibility in different plasma samples. Deep whole genome sequencing (WGS) data can be obtained from plasma samples from healthy donors and from patients with cancer (e.g., metastatic prostate, colon, breast cancer). In some instances, cfDNA also contains circulating tumor DNA (ctDNA). In addition, shallow WGS data can also be obtained from plasma samples from patients with tumor bodies as mentioned above. This approach can be used to profile individual TFs instead of establishing general tissue-specific patterns using a mixture of cfDNA signals from multiple cell types and analysis by Fourier transform, as in other approaches. The methods and systems provided herein provide a more nuanced view of both tissue contributions and biological processes, which can identify lineage-specific TFs suitable for both tissue- and tumor-origin analysis.

[0110] Certain lineage-specific TFs may be suitable for determining the tissue origin of plasma DNA. However, to determine which TFs may be useful for such applications, it is necessary to assess, for example, the accessibility of the TF at its binding site in cfDNA. Traditional methods may not be able to assess the accessibility of the TF at its binding site in cfDNA as a proxy for its activity. Separate calculations are performed on TFBSs inside and outside the transcription start site (TSS). The average TFBS pattern contains two signals: TSS-proximal (within about 2 kb of the TSS resulting in a "low frequency pattern") and TSS-distal (more than 2 kb away from the TSS peak resulting in a "high frequency pattern"), which correspond to more evenly spaced peak signals. To suppress coverage effects not brought about by preferential nucleosome positioning and to remove local biases from the nucleosome data, a filter may be used for detrending (e.g., Savitzky-Golay filter). The resulting low-frequency signals can then be used to normalize the high-frequency signals, after which the data ranges of the high-frequency signals (maximum minus minimum data values ​​corresponds to amplitude) can be recorded. As the ranges of the high-frequency signals depend on the number of TFBSs (except for the 1000ms TFBS), these range values ​​can be corrected by smoothing as they depend on the number of TFBSs, and then used to calculate a rank as an index of the accessibility of each TFBS.

[0111] The metric developed for this analysis (called the "accessibility score") can be used to objectively compare the accessibility of TFBSs in consecutive analyses of samples obtained from the same or different individuals. The rank value is sometimes referred to as the "accessibility score" because TF binding opens or "primes" its target enhancer without necessarily activating the target enhancer itself. These results demonstrate a robust approach to assess TFBS accessibility, particularly useful for utilizing cfDNA in clinical diagnostics.

[0112] In contrast to other analyses that may use a common tissue-specific pattern using a mixture of cfDNA signals from multiple cell types and analysis by Fourier transform, the disclosed method and system can profile individual TFs, thereby establishing lineage-specific TFs for clinical application.The improved resolution of TFBS analysis allows monitoring the accessibility of TFBS from cfDNA, and in some cases has been proven useful for revealing TF plasticity during disease course, for example, reprogramming into different cell lineages.

[0113] Figure 20 is a plot showing how epigenetic control regions affect nucleosome positioning. Histone modifications and enhancers are illustrated.

[0114] FIG. 21 shows the nucleosomal positioning of selected TFs.

[0115] Figure 22 shows the coverage patterns of selected TFs.

[0116] Figure 23 shows the overlap of different TFs. These overlap values ​​correspond to the heatmap in Figure 2C.

[0117] Figure 24 shows the effect of TFBS size.

[0118] Figure 25 shows the nucleosome patterns of REST and KLF16 for samples from 24 healthy individuals. Each line represents a different individual. In the 24 healthy individuals, the patterns appear in a nearly identical set of transcription factors that are active in blood cells.

[0119] Figure 26A-C shows the nucleosome positions of selected TFs in late-stage cancer samples. The CTCF patterns are similar in all samples. The activity of blood-specific TFs, including PU.1, Lyl-1, and Spi-B, is reduced in cancer samples. Cancer-specific TFs, including GRH-L2 (epithelial marker), ASH-2 and HOX-B13 (prostate cancer markers of the androgen receptor axis), and EVX2 (colon cancer marker), are more active.

[0120] II. Transcription factor binding sites Transcription factor binding sites are identified from the Gene Transcription Regulation Database (GTRD: Gene Transcription Regulation Database - 2019 Update. ISYevshin, RNSharipov. SKKolmykov, YVKondrakhin, FA Kolpakov. Nucleic Acids Res. 2019 Jan 8;47(D1):D100-D105) using statistical thresholds for use in the present methods and systems and are informative for machine learning model and classifier generation. In some instances, associated pathways and classes of transcription factors are similarly useful and informative for machine learning model and classifier generation.

[0121] A statistical threshold is used to identify differential TFs between two or more patient groups for analysis (e.g., healthy subjects versus cancer patients, progressors versus non-progressors, stages among multiple stages (e.g., I, II, III, or IV), subtypes among multiple subtypes, or treatment responders versus non-responders).

[0122] In some examples, transcription factors such as those described in Figures 27-1 and 27-2, and Figures 28A and 28B, can be analyzed using the methods and systems described herein.

[0123] In some examples, the transcription factor is selected from the group consisting of GRH-L2, ASH-2, HOX-B13, EVX2, PU.1, Lyl-1, Spi-B, and FOXA1.

[0124] In several instances, patient- and tumor-specific patterns are observed, including inferred binding patterns for the transcription factors AR, HOXB13, and NKX3-1.

[0125] In some instances, the transcription factor is an epithelial transcription factor. In some instances, the transcription factor is GRHL2.

[0126] In some instances, the transcription factors GRHL2, FOXA1, and ZNF121 are associated with increased accessibility scores or open chromatin accessibility in patients with breast cancer.

[0127] In some instances, open accessibility of at least one transcription factor from GRHL2, FOXA1, and ZNF121 is indicative of breast cancer.

[0128] In some instances, the transcription factors EVX2, DLX2, HNF1A, HNF4A, GRHL2, and HNF4G are associated with increased accessibility scores or open chromatin accessibility in patients with colon cancer.

[0129] In some instances, open accessibility of at least one transcription factor from EVX2, DLX2, HNF1A, GRHL2, HNF4A, and HNF4G is indicative of colon cancer.

[0130] In some instances, the transcription factors LYL1, EVI1, TAL1, Spi-B, TBX21, and PU.1 are associated with decreased accessibility scores or closed chromatin accessibility in patients with colon cancer.

[0131] In some instances, closed accessibility of at least one transcription factor selected from LYL1, EVI1, TAL1, Spi-B, TBX21, and PU.1 is indicative of colon cancer.

[0132] In some instances, the transcription factors GRHL2, FOXA1, HOXB13, AR, and NKX3-1 are associated with increased accessibility scores or open chromatin accessibility in patients with prostate adenocarcinoma.

[0133] In some instances, open accessibility of at least one transcription factor selected from GRHL2, FOXA1, HOXB13, AR, and NKX3-1 is indicative of prostate adenocarcinoma.

[0134] In some instances, the transcription factors REST, GRHL2, FOXA1, HOXB13, AR, and NKX3-1 are associated with reduced or closed chromatin accessibility in patients with small cell neuroendocrine prostate cancer.

[0135] In some instances, decreased accessibility of at least one transcription factor selected from REST, GRHL2, FOXA1, HOXB13, AR, and NKX3-1 is indicative of small cell neuroendocrine prostate cancer.

[0136] In one example, the correlation between hematopoietic transcription factor accessibility and tissue-specific TFs is associated with the presence of disease, such as cancer.

[0137] In one example, the hematopoietic transcription factor is selected from LYL1, SCL, Bcl11a, Hhex, Lmo2, Spi1, and PU.1 In one example, the hematopoietic transcription factor is selected from LYL1 or PU.1.

[0138] In some instances, low accessibility of hematopoietic transcription factors such as LYL1, SPIB, and EVI1 (transcriptional regulator ecotropic virus integration site 1) is associated with prostate cancer.

[0139] In some examples, the transcription factor is selected from the group consisting of GRH-L2, ASH-2, HOX-B13, EVX2, PU.1, Lyl-1, Spi-B, and FOXA1.

[0140] In some examples, the transcription factor is selected from the group consisting of HNF-1a, HNF-4A, HNF-4G, EVX-2, and DLX-2.

[0141] In some instances, low accessibility of hematopoietic-related TFs, such as LYL1, TAL1 (SCL / TAL1 (Stem Cell Leukemia / T-cell Acute Lymphoblastic Leukemia [T-ALL]1), EVI1, TBX21 (T-bet), and PU.1, is associated with cancer.

[0142] Persistent exposure to antigens in chronic viral infections and cancer induces the expression of effector CD8 + T cells adopt an alternative differentiation fate, called T cell exhaustion. They are unable to undergo antigen-independent self-renewal like memory cells and lose effector function hierarchically, which leads to the elimination of these antigen-specific CD8 + Prevents T cell-mediated viral clearance and tumor control.

[0143] In one example, open accessibility of the transcription factors tbx21 or EOMES regulates exhausted CD8 + Shows T cells.

[0144] In one example, open accessibility of the transcription factors eomesodermin (EOMES), Blimb-1, von Hippel-Lindau tumor suppressor gene (VHL), Foxo1, IRF4, BATF, and NFATc1 regulates the expression of exhausted CD8 + Shows T cells.

[0145] III. Machine Learning Systems and Models A. Sample characteristics In some examples, TFBS accessibility scores are used as input features in machine learning models to discover correlations between sequence composition and groups of subjects (e.g., patients). Examples of such patient groups include the presence of a disease or condition, stage, subtype, responders vs. non-responders, and progressors vs. non-progressors. In some examples, feature matrices are generated to compare samples obtained from individuals with known conditions or characteristics. In some examples, samples are obtained from healthy individuals or individuals with no known indications, and samples from patients known to have cancer.

[0146] As used herein, as it relates to machine learning and pattern recognition, the term "feature" refers to an individual measurable characteristic or feature of an observed phenomenon. Features are typically numerical, although structural features such as strings and graphs may be used in syntactic pattern recognition. The concept of "feature" is related to the concept of explanatory variables used in statistical methods such as, for example, but not limited to, linear regression. In some examples, the feature is a transcription factor binding profile. In some examples, the feature is an accessibility score calculated from the transcription factor binding profile.

[0147] In some examples, the features are input into a feature matrix for machine learning analysis.

[0148] In some examples, accessibility scores for at least 2, or at least 5, or at least 10, or at least 15, or at least 20, or at least 25 transcription factor binding sites are determined and input into a machine learning model to train a classifier capable of distinguishing between healthy subjects and cancer patients, or disease progressors and non-progressors.

[0149] In some examples, accessibility scores for at least 2, or at least 5, or at least 10, or at least 15, or at least 20, or at least 25 transcription factor binding sites are determined and input into a machine learning model to train a classifier capable of identifying multiple disease subtypes or multiple disease stages.

[0150] In some examples, accessibility scores for at least 2, or at least 5, or at least 10, or at least 15, or at least 20, or at least 25 transcription factor binding sites are determined and input into a machine learning model to train a classifier capable of distinguishing between disease treatment responders and non-responders.

[0151] For the plurality of assays, the system identifies a feature set that is received as input to a machine learning model. The system runs an assay for each molecular class and forms a feature vector from the measurements. The system receives the feature vector as input to the machine learning model and generates an output classification of whether the biological sample has the specified characteristic.

[0152] In some examples, the machine learning model generates a classifier that can distinguish between two or more groups or classes of features of an individual or a population of individuals or features of a population. For example, the classifier may be a binary classifier that can distinguish between two groups or classes of features of an individual or a population of individuals or features of a population. As another example, the classifier may be a multi-class classifier that can distinguish between two groups or classes of features of an individual or a population of individuals or features of a population. In some examples, the classifier is a trained machine learning classifier.

[0153] In some examples, informative loci or features of biomarkers in cancer tissues are assayed to form a profile. In the case of binary classifiers, a receiver operating characteristic (ROC) curve can be generated to plot the performance of a particular feature (e.g., any of the biomarkers described herein and / or any item of additional biomedical information) in distinguishing two populations (e.g., individuals who respond to a therapeutic drug and those who do not). In some examples, the feature data of the entire population (e.g., cases and controls) is sorted in ascending order based on the value of a single feature.

[0154] In some examples, the specified characteristic is selected from healthy subjects versus cancer patients, a disease subtype among a plurality of disease subtypes, a disease stage among a plurality of disease stages, progressors versus non-progressors, responders versus non-responders, or a combination thereof.

[0155] B. Data Analysis In some examples, the disclosure provides a system, method, or kit having data analysis implemented in software applications, computing hardware, or both. In some examples, the analysis application or system includes at least a data receiving module, a data pre-processing module, a data analysis module (which can operate on one or more types of genomic data), a data interpretation module, or a data visualization module. In some examples, the data receiving module can include a computer system that connects laboratory hardware or instrumentation to a computer system that processes laboratory data. In some examples, the data pre-processing module can include a hardware system or computer software that performs operations on the data in preparation for analysis. Examples of operations that can be applied to the data in the pre-processing module include affine transformations, noise removal operations, data cleaning, reformatting, or sub-sampling. The data analysis module can specialize in the analysis of genomic data from one or more genomic materials, for example, taking assembled genomic sequences and performing probabilistic and statistical analysis to identify abnormal patterns associated with a disease, pathology, state, risk, condition, or phenotype. The data interpretation module can use analytical methods derived from, for example, statistics, mathematics, or biology to support an understanding of the association between the identified abnormality patterns and health status, functional status, prognosis, or risk. The data visualization module can use methods of mathematical modeling, computer graphics, or rendering to create visual representations of the data that can facilitate understanding or interpretation of the results (e.g., by a subject (e.g., a patient), or a user such as a physician or other health care provider).

[0156] In some instances, machine learning methods are applied to distinguish between samples in a population of samples, hi some instances, machine learning methods are applied to distinguish between healthy and cancerous (e.g., advanced adenoma) samples.

[0157] In some examples, the one or more machine learning operations used to train the prediction engine include one or more of a generalized linear model, a generalized additive model, a nonparametric regression operation, a random forest classifier, a spatial regression operation, a Bayesian regression model, a time series analysis, a Bayesian network, a Gaussian network, a decision tree learning operation, an artificial neural network, a recurrent neural network, a reinforcement learning operation, a linear or nonlinear regression operation, a support vector machine, a clustering operation, and a genetic algorithm operation.

[0158] In some examples, the computational method is selected from logistic regression, multiple linear regression (MLR), dimensionality reduction, partial least squares (PLS) regression, principal component regression, autoencoder, variational autoencoder, singular value decomposition, Fourier basis, wavelets, discriminant analysis, support vector machines, decision trees, classification and regression trees (CART), tree-based methods, random forests, gradient boosted trees, logistic regression, matrix decomposition, multidimensional scaling (MDS), dimensionality reduction methods, t-SNE, multilayer perceptron (MLP), network clustering, neuro-fuzzy, and artificial neural networks.

[0159] In some examples, the methods disclosed herein can include computational analysis of nucleic acid sequence data of a sample from an individual, or from multiple individuals.

[0160] C. Classifier generation In one embodiment, the present disclosure provides a system and method, comprising a classifier that is generated based on the feature information derived from sequence analysis of cfDNA biological sample.The classifier forms part of a prediction engine for distinguishing groups in a population based on the sequence features identified in biological sample such as cfDNA.

[0161] In some examples, the classifier is created by normalizing the sequence information by formatting similar portions of the sequence information into a unified format and a unified scale; storing the normalized sequence information in a columnar database; training a prediction engine by applying one or more machine learning techniques to the stored normalized sequence information, where the prediction engine maps one or more feature combinations to a particular population; applying the prediction engine to the accessed field information to identify individuals associated with a group; and classifying the individuals into a group.

[0162] The trained classifier may be configured to receive a plurality of input variables and generate one or more output values ​​based on the plurality of input variables. The plurality of input variables may include one or more data sets indicative of a disease, disorder, or abnormal condition (e.g., cancer). For example, the input variables may include a number of nucleic acid sequences that correspond to or match a set of disease-associated genomic loci. The plurality of input variables may further include clinical health data of the subject.

[0163] For example, clinical health data may include one or more quantitative measurements of a subject, such as age, weight, height, body mass index (BMI), blood pressure, heart rate, and glucose level. As another example, clinical health data may include one or more categorical measures, such as race, ethnicity, history of medication or other clinical treatments, history of tobacco use, history of alcohol consumption, daily activity or fitness level, genetic test results, blood test results, and imaging results.

[0164] The trained algorithms provided herein may include classifiers, where each of the one or more output values ​​includes one of a fixed number of possible values ​​(e.g., linear classifiers, logistic regression classifiers, etc.) that indicate the classification of the sample by the classifier. The trained algorithms may include binary classifiers, where each of the one or more output values ​​includes one of two values ​​(e.g., {0,1}, {positive, negative}, or {high risk, low risk}) that indicate the classification of the sample by the classifier. The trained algorithms may be another type of classifier, where each of the one or more output values ​​includes one of more than two values ​​(e.g., {0,1,2}, {positive, negative, or indeterminate}, or {high risk, medium risk, or low risk}) that indicate the classification of the sample by the classifier. The output values ​​may include descriptive labels, numerical values, or combinations thereof. Some of the output values ​​may include descriptive labels. Such descriptive labels may provide an identification or indication of the subject's disease, disorder, or abnormal condition assessment, and may include, for example, positive, negative, high risk, medium risk, low risk, or indeterminate. Such descriptive labels may provide an identification of a treatment for the subject's disease, disorder, or abnormal condition assessment, and may include, for example, a therapeutic intervention, a duration of the therapeutic intervention, and / or a dosage of the therapeutic intervention suitable for treating the disease, disorder, or abnormal condition. Such descriptive labels may provide an identification of a secondary clinical test suitable for performing on the subject, and may include, for example, an imaging test, a blood test, a computed tomography (CT) scan, a magnetic resonance imaging (MRI) scan, an ultrasound scan, a chest x-ray, a positron emission tomography (PET) scan, a PET-CT scan, a cytology assay, or any combination thereof. For example, such descriptive labels may provide a prognosis of the subject's disease, disorder, or abnormal condition. As another example, such descriptive labels may provide a relative assessment of the subject's disease, disorder, or abnormal condition. Some descriptive labels may be mapped to numerical values, for example by mapping "positive" to 1 and "negative" to 0.

[0165] Some of the output values ​​may include numerical values, such as binary values, integer values, or continuous values. Such binary output values ​​may include, for example, {0, 1}, {positive, negative}, {high risk, low risk}. Such integer output values ​​may include, for example, {0, 1, 2}. Such continuous output values ​​may include, for example, probability values ​​at least 0 and less than or equal to 1. Such continuous output values ​​may include, for example, a non-normalized probability value of at least 0. Such continuous output values ​​may also indicate a prognosis of a disease, disorder, or abnormal condition in a subject. Some numerical values ​​may be mapped to descriptive labels, for example, by mapping 1 to "positive" and 0 to "negative."

[0166] Some of the output values ​​may be assigned based on one or more cutoff values. For example, a binary classification of a sample may assign an output value of "positive" or 1 if the sample indicates that the subject has a 50% or greater probability of having a disease, disorder, or abnormal condition. For example, a binary classification of a sample may assign an output value of "negative" or "0" if the sample indicates that the subject has a less than 50% probability of having a disease, disorder, or abnormal condition. In this case, a single cutoff value of 50% is used to classify the sample into one of two possible binary output values. Examples of single cutoff values ​​are about 1%, about 2%, about 5%, about 10%, about 15%, about 20%, about 25%, about 30%, about 35%, about 40%, about 45%, about 50%, about 55%, about 60%, about 65%, about 70%, about 75%, about 80%, about 85%, about 90%, about 91%, about 92%, about 93%, about 94%, about 95%, about 96%, about 97%, about 98%, and about 99%.

[0167] As another example, a classification of a sample can be assigned an output value of "positive" or 1 if the sample indicates that there is at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, at least about 99%, or more, probability that the subject has a disease, disorder, or abnormal condition. The classification of the sample can be assigned an output value of "positive" or 1 if the sample indicates a probability that the subject has a disease, disorder, or abnormal condition greater than about 50%, greater than about 55%, greater than about 60%, greater than about 65%, greater than about 70%, greater than about 75%, greater than about 80%, greater than about 85%, greater than about 90%, greater than about 91%, greater than about 92%, greater than about 93%, greater than about 94%, greater than about 95%, greater than about 96%, greater than about 97%, greater than about 98%, or greater than about 99%.

[0168] A classification of a sample can be assigned an output value of "negative" or 0 if the sample indicates that there is less than about 50%, less than about 45%, less than about 40%, less than about 35%, less than about 30%, less than about 25%, less than about 20%, less than about 15%, less than about 10%, less than about 9%, less than about 8%, less than about 7%, less than about 6%, less than about 5%, less than about 4%, less than about 3%, less than about 2%, or less than about 1% probability that the subject has a disease, disorder, or abnormal condition. A classification of a sample can be assigned an output value of "negative" or 0 if the sample indicates that the subject has about 50% or less, about 45% or less, about 40% or less, about 35% or less, about 30% or less, about 25% or less, about 20% or less, about 15% or less, about 10% or less, about 9% or less, about 8% or less, about 7% or less, about 6% or less, about 5% or less, about 4% or less, about 3% or less, about 2% or less, or about 1% probability that the subject has a disease, disorder, or abnormal condition.

[0169] Classification of a sample may involve assigning an output value of "indeterminate" or "2" if the sample cannot be classified as "positive", "negative", "1", or "0". In this case, a set of two cutoff values ​​is used to classify the sample into one of three possible output values. Example sets of cutoff values ​​may include: {1%, 99%}, {2%, 98%}, {5%, 95%}, {10%, 90%}, {15%, 85%}, {20%, 80%}, {25%, 75%}, {30%, 70%}, {35%, 65%}, {40%, 60%}, and {45%, 55%}. Similarly, a set of n cutoff values ​​may be used to classify a sample into one of n+1 possible output values, where n is any positive integer.

[0170] The trained classifier may be trained using multiple independent training samples. Each of the independent training samples may include a sample from a subject, an associated dataset (described elsewhere herein) obtained by assaying the sample, and one or more known output values ​​corresponding to the sample (e.g., clinical diagnosis, prognosis, absence, or treatment effect of a disease, disorder, or abnormal condition in the subject). The independent training samples may include samples and associated datasets and outputs obtained or derived from multiple different subjects. The independent training samples may include samples and associated datasets and outputs obtained from the same subject at multiple different time points (e.g., periodically, such as weekly, biweekly, or monthly). The independent training samples may be associated with the presence of a disease, disorder, or abnormal condition (e.g., the training samples include samples and associated datasets and outputs obtained or derived from multiple subjects known to have a disease, disorder, or abnormal condition). Independent training samples may be associated with the absence of a disease, disorder, or abnormal condition (e.g., training samples include samples and associated datasets and output obtained or derived from multiple subjects who are known to have no previous diagnosis of a disease, disorder, or abnormal condition or who have received a negative test result for a disease, disorder, or abnormal condition).

[0171] The trained classifier may be trained with at least about 5, at least about 10, at least about 15, at least about 20, at least about 25, at least about 30, at least about 35, at least about 40, at least about 45, at least about 50, at least about 100, at least about 150, at least about 200, at least about 250, at least about 300, at least about 350, at least about 400, at least about 450, or at least about 500 independent training samples. The independent training samples may include samples associated with the presence of a disease, disorder, or abnormal condition and / or samples associated with the absence of a disease, disorder, or abnormal condition. The trained classifier may be trained with about 500 or less, about 450 or less, about 400 or less, about 350 or less, about 300 or less, about 250 or less, about 200 or less, about 150 or less, about 100 or less, or about 50 or less independent training samples associated with the presence of a disease, disorder, or abnormal condition. In some embodiments, the samples are independent of the samples used to train the trained classifier.

[0172] The trained classifier may be trained using a first number of independent training samples associated with the presence of a disease, disorder, or abnormal condition and a second number of independent training samples associated with the absence of a disease, disorder, or abnormal condition. The first number of independent training samples associated with the presence of a disease, disorder, or abnormal condition may be less than or equal to the second number of independent training samples associated with the absence of a disease, disorder, or abnormal condition. The first number of independent training samples associated with the presence of a disease, disorder, or abnormal condition may be equal to the second number of independent training samples associated with the absence of a disease, disorder, or abnormal condition. The first number of independent training samples associated with the presence of a disease, disorder, or abnormal condition may be greater than the second number of independent training samples associated with the absence of a disease, disorder, or abnormal condition.

[0173] The trained classifier may provide at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 80%, at least about 90%, at least about 100%, at least about 150, at least about 200, at least about 250, at least about 300, at least about 350, at least about 400, at least about 450, or at least about 500 independent training samples. The trained algorithm may be configured to identify the presence or absence of a disease, disorder, or abnormal condition with an accuracy of at least about 75%, at least about 80%, at least about 81%, at least about 82%, at least about 83%, at least about 84%, at least about 85%, at least about 86%, at least about 87%, at least about 88%, at least about 89%, at least about 90%, at least about 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, at least about 99%, or more. The accuracy of identifying the presence or absence of a disease, disorder, or abnormal condition by the trained algorithm may be calculated as the percentage of independent test samples (e.g., subjects known to have a disease, disorder, or abnormal condition or subjects with a negative laboratory test result for a disease, disorder, or abnormal condition) that are correctly identified or classified as having or not having the disease, disorder, or abnormal condition.

[0174] The trained classifier can be configured to identify the presence of a disease, disorder, or abnormal condition with a positive predictive value (PPV) of at least about 5%, at least about 10%, at least about 15%, at least about 20%, at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 81%, at least about 82%, at least about 83%, at least about 84%, at least about 85%, at least about 86%, at least about 87%, at least about 88%, at least about 89%, at least about 90%, at least about 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, at least about 99% or more. The PPV of identifying the presence of a disease, disorder, or abnormal condition using a trained classifier can be calculated as the proportion of samples identified or classified as having the disease, disorder, or abnormal condition that correspond to subjects who truly have the disease, disorder, or abnormal condition.

[0175] The trained classifier can be configured to identify the absence of a disease, disorder, or abnormal condition with a negative predictive value (NPV) of at least about 5%, at least about 10%, at least about 15%, at least about 20%, at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 81%, at least about 82%, at least about 83%, at least about 84%, at least about 85%, at least about 86%, at least about 87%, at least about 88%, at least about 89%, at least about 90%, at least about 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, at least about 99% or more. The NPV of identifying a disease, disorder, or abnormal condition using a trained classifier can be calculated as the proportion of samples identified or classified as not having the disease, disorder, or abnormal condition, corresponding to subjects who truly do not have the disease, disorder, or abnormal condition.The trained classifier may be at least about 5%, at least about 10%, at least about 15%, at least about 20%, at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 81%, at least about 82%, at least about 83%, at least about 84%, at least about 85%, at least about 86%, at least about 87%, at least about 88%, at least about 89%, at least about 90%, at least about 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, at least about 99%, at least about 100%, at least about 101%, at least about 102%, at least about 103%, at least about 104%, at least about 105%, at least about 106%, at least about 107%, at least about 108%, at least about 109%, at least about 110%, at least about 111%, at least about 112%, at least about 113%, at least about 114%, at least about 115%, at least about 116%, at least about 117%, at least about 118%, at least about 119%, at least about 120%, at least about 122%, at least about 123%, at least about 124%, at least about 125%, at least about 126%, at least about 127%, at least about 128%, at least about 129%, at least about 130%, at least about 131%, at least about 132%, at least about 133%, at least about It may be configured to identify the absence of a disease, disorder, or abnormal condition with a clinical specificity of 1%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, at least about 99%, at least about 99.1%, at least about 99.2%, at least about 99.3%, at least about 99.4%, at least about 99.5%, at least about 99.6%, at least about 99.7%, at least about 99.8%, at least about 99.9%, at least about 99.99%, at least about 99.999%, or more. As used herein, specificity means "the probability that the test will be negative among those without the disease". It is equal to the number of those without the disease who test negative divided by the total number of individuals without the disease. Clinical specificity for identifying the absence of a disease, disorder, or abnormal condition using a trained classifier can be calculated as the percentage of independent test samples associated with the absence of a disease, disorder, or abnormal condition (e.g., subjects with a negative clinical test result for the disease, disorder, or abnormal condition) that are correctly identified or classified as not having the disease, disorder, or abnormal condition. In some examples, the model, classifier, or predictive test has a specificity of at least about 40%, at least about 45%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, or at least about 99%.

[0176] The trained classifier may be at least about 5%, at least about 10%, at least about 15%, at least about 20%, at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 81%, at least about 82%, at least about 83%, at least about 84%, at least about 85%, at least about 86%, at least about 87%, at least about 88%, at least about 89%, at least about 90%, at least about The present invention may be configured to identify the presence of a disease, disorder, or abnormal condition with a clinical sensitivity of 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, at least about 99%, at least about 99.1%, at least about 99.2%, at least about 99.3%, at least about 99.4%, at least about 99.5%, at least about 99.6%, at least about 99.7%, at least about 99.8%, at least about 99.9%, at least about 99.99%, at least about 99.999%, or more. As used herein, sensitivity refers to the "probability of a test being positive among individuals with a disease." It is equal to the number of individuals with a disease who test negative divided by the total number of individuals with the disease.

[0177] In some examples, the model, classifier, or predictive test has a sensitivity of at least about 40%, at least about 45%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 85%, at least about 90%, at least about 95%, or at least about 99%. The clinical sensitivity of identifying the presence of a disease, disorder, or abnormal condition using a trained classifier can be calculated as the proportion of independent test samples associated with the presence of a disease, disorder, or abnormal condition (e.g., subjects known to have a disease, disorder, or abnormal condition) that are correctly identified or classified as having the disease, disorder, or abnormal condition.

[0178] The trained classifier can be configured to discriminate the presence or absence of a disease, disorder, or abnormal condition with an area under the receiver operating characteristic (AUROC) of at least about 0.50, at least about 0.55, at least about 0.60, at least about 0.65, at least about 0.70, at least about 0.75, at least about 0.80, at least about 0.81, at least about 0.82, at least about 0.83, at least about 0.84, at least about 0.85, at least about 0.86, at least about 0.87, at least about 0.88, at least about 0.89, at least about 0.90, at least about 0.91, at least about 0.92, at least about 0.93, at least about 0.94, at least about 0.95, at least about 0.96, at least about 0.97, at least about 0.98, at least about 0.99, or more. The AUROC may be calculated as the integral of a receiver operating characteristic (ROC) curve (e.g., area under the ROC curve, AUC) associated with a trained classifier in classifying samples as having or not having a disease, disorder, or abnormal condition.

[0179] The trained classifier may be adjusted or tuned to improve one or more of the performance, accuracy, PPV, NPV, clinical sensitivity, clinical specificity, or AUC of identifying a disease, disorder, or abnormal condition. The trained classifier may be adjusted or tuned by adjusting the parameters of the trained classifier (e.g., the set of cutoff values ​​used to classify samples as described elsewhere herein, or the weights of a neural network). The trained classifier may be adjusted or tuned continuously during the training process or after the training process is completed.

[0180] After the trained classifier is initially trained, a subset of inputs may be identified as the most influential or most important to be included to make a high-quality classification. For example, a subset of the multiple input variables may be identified as the most influential or most important to be included to make a high-quality classification or identification of a disease, disorder, or abnormal condition assessment. The multiple input variables or subsets thereof may be ranked based on a classification metric indicating the influence or importance of each input variable to making a high-quality classification or identification of a disease, disorder, or abnormal condition assessment. Such a metric may be used to reduce, in some cases, significantly the number of input variables (e.g., predictor variables) that may be used to train the trained classifier to a desired performance level (e.g., based on a desired minimum accuracy, PPV, NPV, clinical sensitivity, clinical specificity, AUC, or a combination thereof). For example, if training a classifier trained with a plurality of input variables including tens or hundreds of input variables in the classifier results in classification accuracy of greater than 99%, instead training a classifier trained with only a selected subset of about 5 or less, about 10 or less, about 15 or less, about 20 or less, about 25 or less, about 30 or less, about 35 or less, about 40 or less, about 45 or less, about 50 or less, or about 100 or less such most influential or most important input variables in the plurality of input variables may result in a reduced but acceptable classification accuracy (e.g., at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 81%, at least about 82%, at least about 83%, at least about 84%, at least about 85%, at least about 86%, at least about 87%, at least about 88%, at least about 89%, at least about 90%, at least about 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, or at least about 99%.The subset may be selected by ranking across the multiple input variables and selecting a predetermined number of input variables (e.g., about 5 or less, about 10 or less, about 15 or less, about 20 or less, about 25 or less, about 30 or less, about 35 or less, about 40 or less, about 45 or less, about 50 or less, or about 100 or less) with the best classification metric.

[0181] D. Digital Processing Device In some examples, the subject matter described herein may include a digital processing device or use thereof. In some examples, the digital processing device may include one or more hardware central processing units (CPUs), graphic processing units (GPUs), or tensor processing units (TPUs) that perform the functions of the device. In some examples, the digital processing device may include an operating system configured to execute executable instructions. In some examples, the digital processing device may be connected to a computer network. In some examples, the digital processing device may be connected to the Internet. In some examples, the digital processing device may be connected to a cloud computing infrastructure. In some examples, the digital processing device may be connected to an intranet. In some examples, the digital processing device may be connected to a data storage device.

[0182] Non-limiting examples of suitable digital processing devices include server computers, desktop computers, laptop computers, notebook computers, sub-notebook computers, netbook computers, netpad computers, set-top computers, handheld computers, Internet appliances, mobile smart phones, and tablet computers. Suitable tablet computers may include, for example, booklet, slate, and convertible configurations.

[0183] In some examples, the digital processing device may include an operating system configured to execute executable instructions. For example, an operating system may include software, including programs and data, that manages the device's hardware and provides services for the execution of applications. Non-limiting examples of operating systems include Ubuntu, FreeBSD, OpenBSD, NetBSD, Linux, Apple Mac OS X Server, Oracle Solaris, Windows Server, and Novell NetWare. Non-limiting examples of suitable personal computer operating systems include Microsoft Windows, Apple Mac OS X, UNIX, and UNIX-like operating systems, e.g., GNU / Linux. In some examples, the operating system may be provided by cloud computing, and cloud computing resources may be provided by one or more service providers.

[0184] In some examples, the device may include a storage device and / or memory device. The storage device and / or memory device may be one or more physical devices used to store data or programs on a temporary or permanent basis. In some examples, the device may be a volatile memory, requiring power to maintain stored information. In some examples, the device may be a non-volatile memory, retaining stored information when power is not provided to the digital processing device. In some embodiments, the non-volatile memory may include flash memory. In some examples, the non-volatile memory may include dynamic random access memory (DRAM). In some examples, the non-volatile memory may include ferroelectric random access memory (FRAM®). In some examples, the non-volatile memory may include phase change random access memory (PRAM). In some examples, the device may be a storage device, including, for example, CD-ROM, DVD, flash memory device, magnetic disk drive, magnetic tape drive, optical disk drive, and cloud computing based storage device. In some examples, the storage device and / or memory device may be a combination of devices such as those disclosed herein. In some examples, the digital processing device may include a display for transmitting visual information to a user. In some examples, the display may be a cathode ray tube (CRT). In some examples, the display may be a liquid crystal display (LCD). In some examples, the display may be a thin film transistor liquid crystal display (TFT-LCD). In some examples, the display may be an organic light emitting diode (OLED) display. In some examples, the OLED display may be a passive-matrix OLED (PMOLED) or active-matrix OLED (AMOLED) display. In some examples, the display may be a plasma display. In some examples, the display may be a video projector. In some examples, the display may be a combination of devices as disclosed herein.

[0185] In some examples, the digital processing device may include an input device for receiving information from a user. In some examples, the input device may be a keyboard. In some examples, the input device may be a pointing device including, for example, a mouse, a trackball, a trackpad, a joystick, a game controller, or a stylus. In some examples, the input device may be a touch screen or a multi-touch screen. In some examples, the input device may be a microphone that captures voice or other audio input. In some examples, the input device may be a video camera that captures motion or visual input. In some examples, the input device may be a combination of devices as disclosed herein.

[0186] E. Non-Transitory Computer-Readable Storage Medium In some examples, the subject matter disclosed herein may include one or more non-transitory computer-readable storage media encoded with a program including instructions executable by an operating system. The operating system may be part of a networked digital processing device. In some examples, the computer-readable storage medium may be a tangible component of the digital processing device. In some examples, the computer-readable storage medium may be removable from the digital processing device. In some examples, the computer-readable storage medium may include, for example, CD-ROMs, DVDs, flash memory devices, solid-state memory, magnetic disk drives, magnetic tape drives, optical disk drives, cloud computing systems and services, and the like. In some examples, the programs and instructions may be encoded on the medium permanently, nearly permanently, semi-permanently, or non-transitoryly.

[0187] F. Computer Systems The present disclosure provides a computer system programmed to perform the methods described herein. Figure 1 shows a computer system (101) programmed or otherwise configured to perform the methods of the present disclosure, such as storing, processing, identifying, or interpreting subject (e.g., patient) data, biological data, biological sequences, reference sequences, TFBS data, or TFBS features such as Z-scores or TFBS accessibility scores. The computer system (101) can process various aspects of the subject (e.g., patient) data, biological data, biological sequences, or reference sequences of the present disclosure. The computer system (101) is an electronic device of a user or a computer system, which user or computer system is located remotely relative to the electronic device. The electronic device may be a mobile electronic device.

[0188] The computer system (101) includes a central processing unit (CPU, also referred to herein as "processor" and "computer processor") (105), which may be a single-core or multi-core processor, or multiple processors for parallel processing. The computer system (101) also includes memory or storage locations (110) (e.g., random access memory, read-only memory, flash memory), electronic storage (115) (e.g., hard disk), communication interfaces (120) (e.g., network adapters) for communicating with one or more other systems, and peripherals (125), such as cache, other memory, data storage devices, and / or electronic display adapters. The memory (110), storage (115), interface (120), and peripherals (125) communicate with the CPU (105) via a communication bus (solid line), such as a motherboard. The storage (115) may be a data storage device (or data repository) for storing data. The computer system (101) may be operatively connected to a computer network ("network") (130) with the aid of the communication interface (120). The network (130) may be the Internet and / or an extranet, or an intranet and / or an extranet in communication with the Internet. The network (130) may, in some examples, be a telecommunications and / or data network. The network (130) may include one or more computer servers, which may enable distributed computing, such as cloud computing. The network (130) may, in some examples, implement a peer-to-peer network with the aid of the computer system (101), which may enable devices coupled to the computer system (101) to act as clients or servers.

[0189] The CPU (105) can execute a series of machine-readable instructions, which may be embodied in a program or software. The instructions may be stored in a memory location, such as the memory (110). The instructions may be directed to the CPU (105), which may then program or otherwise configure the CPU (105) to perform the methods of the present disclosure. Examples of operations performed by the CPU (105) include fetch, decode, execute, and writeback.

[0190] The CPU 105 may be part of a circuit, such as an integrated circuit. One or more other components of the system 101 may be included in the circuit. In some examples, the circuit is an application specific integrated circuit (ASIC).

[0191] The storage device (115) can store files such as drivers, libraries, and saved programs. The storage device (115) can store user data, such as user preferences and user programs. The computer system (101) may include one or more additional data storage devices that are external to the computer system (101), such as located on a remote server in communication with the computer system (101) over an intranet or the Internet, in some examples.

[0192] The computer system (101) can communicate with one or more remote computer systems via the network (130). For example, the computer (101) can communicate with a user's remote computer system. Examples of remote computer systems include a personal computer (e.g., a portable PC), a slate or tablet PC (e.g., Apple® iPad®, Samsung® Galaxy Tab), a phone, a smartphone (e.g., Apple® iPhone®, Android-enabled device, Blackberry®), or a personal digital assistant. A user can access the computer system (101) via the network (130).

[0193] The methods described herein can be performed by machine (e.g., computer processor) executable code stored in an electronic storage location of the computer system (101), such as, for example, on memory (110) or electronic storage (115). The machine executable or machine readable code can be provided in the form of software. In use, the code can be executed by the processor (105). In some examples, the code can be retrieved from the electronic storage (115) and stored on the memory (110) for easy access by the processor (105). In some examples, the electronic storage (115) can be omitted and the machine executable instructions are stored in the memory (110).

[0194] The code may be pre-compiled and configured for use with a machine having a suitable processor to execute the code, or may be interpreted or compiled during run-time. The code may be provided in a programming language that may be selected to render the code executable in a pre-compiled, interpreted, or as-compiled fashion.

[0195] Aspects of the systems and methods provided herein, such as the computer system (101), may be embodied in programming. Various aspects of the technology may be thought of as an "article of manufacture" or "article of manufacture," typically in the form of machine (or processor) executable code and / or associated data executed or embodied on a type of machine-readable medium. The machine-executable code may be stored in electronic storage, such as memory (e.g., read-only memory, random access memory, flash memory) or a hard disk. A "storage" type medium may include any or all of the tangible memory of a computer or processor, or its associated modules, such as various semiconductor memories, tape drives, disk drives, etc., which may provide non-transitory recording media at any time for programming the software. All or portions of the software may be communicated from time to time over the Internet or various other telecommunications networks. Such communication may enable, for example, loading of the software from one computer or processor to another, for example, from a management server or host computer to an application server computer platform. Thus, other types of media that may bear software elements include light waves, radio waves, and electromagnetic waves, such as those used across physical interfaces between local devices, over wired and optical landline networks, and on various air-links. The physical elements that carry such waves, such as wired or wireless links, optical links, etc., may also be considered media bearing the software. As used herein, unless limited to non-transitory, tangible "storage" media, terms such as computer or machine "readable medium" refer to media that participate in providing instructions to a processor for execution.

[0196] Thus, a machine-readable medium such as a computer executable code may take many forms, including, but not limited to, a tangible storage medium, a carrier wave medium, or a physical transmission medium. Non-volatile storage media include, for example, optical or magnetic disks, any of the storage devices in a computer, such as those that may be used to implement, for example, the databases shown in the figures. Volatile storage media include dynamic memory, such as the main memory of such a computer platform. Tangible transmission media include coaxial cables; copper wire and fiber optics, including the wires that comprise a bus in a computer system. Carrier wave transmission media may take the form of electric or electromagnetic signals, or sound or light waves, such as those generated during radio frequency (RF) and infrared (IR) data communications. Thus, common forms of computer readable media include, for example: floppy disks, flexible disks, hard disks, magnetic tape, other magnetic media, CD-ROMs, DVDs or DVD-ROMs, other optical media, punch cards, paper tape, other physical storage media with patterns of holes, RAM, ROM, PROMs and EPROMs, FLASH®-EPROMs, other memory chips or cartridges, carrier waves carrying data or instructions, cables or links which transmit such carrier waves, or other media from which a computer can read programming code and / or data. Many of these forms of computer readable media may be involved in carrying one or more sequences of one or more instructions to a processor for execution.

[0197] The computer system (101) may include or be in communication with an electronic display (135) that includes a user interface (UI) (140) for providing, for example, nucleic acid sequences, enriched nucleic acid samples, transcription factor binding profiles, accessibility scores, expression profiles, and analysis of the expression profiles. Examples of UIs include, but are not limited to, graphical user interfaces (GUIs) and web-based user interfaces.

[0198] The disclosed methods and systems can be implemented by one or more algorithms. The algorithms can be implemented by software when executed by a central processing unit (105). The algorithms can, for example, search for multiple regulatory elements, sequence the nucleic acid sample, enrich the nucleic acid sample, determine an expression profile of the nucleic acid sample, analyze the expression profile of the nucleic acid sample, and record or disseminate the results of the analysis of the expression profile.

[0199] In some examples, the subject matter disclosed herein may include at least one computer program or the use of a computer program. A computer program may be a set of instructions that can be executed by a CPU, GPU, or TPU of a digital processing device and written to perform a specific task. The computer-readable instructions may be implemented as program modules, such as functions, objects, application programming interfaces (APIs), data structures, etc., that perform a specific task or extract a specific data type. For example, a computer program may be written in a variety of languages ​​and versions.

[0200] The functionality of the computer readable instructions may be combined or distributed as desired in various environments. In some examples, a computer program may include one sequence of instructions. In some examples, a computer program may include multiple sequences of instructions. In some examples, a computer program may be provided from one location. In some examples, a computer program may be provided from multiple locations. In some examples, a computer program may include one or more software modules. In some examples, a computer program may include, in part or in whole, one or more web applications, one or more mobile applications, one or more standalone applications, one or more web browser plug-ins, extensions, add-ins, or add-ons, or combinations thereof.

[0201] In some examples, the computational methods may be methods of statistics, mathematics, biology, or any combination thereof. In some examples, the computational methods include dimensionality reduction methods including, for example, logistic regression, dimensionality reduction, principal component analysis, autoencoders, singular value decomposition, Fourier bases, singular value decomposition, wavelets, discriminant analysis, support vector machines, tree-based methods, random forests, gradient boost trees, logistic regression, matrix factorization, network clustering, and neural networks.

[0202] In some examples, the computational methods are supervised machine learning methods including, for example, regression, support vector machines, tree-based methods, and networks.

[0203] In some examples, the computational methods are unsupervised machine learning methods including, for example, clustering, networks, principal component analysis, and matrix factorization.

[0204] G. Database In some examples, the subject matter disclosed herein may include one or more databases, or the use of one or more of the above databases to store subject (e.g., patient) data, biological data, biological sequences, or reference sequences. The reference sequence may be derived from a database. For example, many databases may be suitable for storing and retrieving sequence information. In some examples, suitable databases may include, for example, relational databases, non-relational databases, object-oriented databases, object databases, entity-relationship model databases, associative databases, and XML databases. In some examples, the database may be internet-based. In some examples, the database may be web-based. In some examples, the database may be cloud computing-based. In some examples, the database may be based on one or more local computer storage devices.

[0205] 676 TFs from the Gene Transcription Regulation Database (GTRD; version 18.01) were used because it contains detailed TFBS information based on ChIP-seq data for various tissue samples. TFs were annotated using an up-to-date curated list of 1,639 known or probable human TFs (Figure 6A). Due to the potentially large number of TFBSs bound by TFs with variable frequency, three different stringency criteria were defined (Figure 6A): first, all TFBSs for all tissue samples in the GTRD; second, those peaks supported by more than 50% of the maximum number of samples (hereafter referred to as ">50%-TFBS"; these two analyses included all 676 GTRD TFs); and third, 1,000 TFBSs per TF supported by the majority of samples ("1,000-msTFBS"; 505 TFs met this criterion).

[0206] In some examples, the reference genome is selected from GrCH38, GrCH37, hg19, or hg38.

[0207] In some examples, a database of reference genomes is used in the alignment and mapping steps of the methods disclosed herein.

[0208] IV.How to use A. Methods and Systems for Diagnosis and Subject Characterization The methods and systems provided herein can use an artificial intelligence-based approach to perform predictive analysis to analyze TFBS data obtained from a subject (e.g., a patient) to generate an output of an assessment (e.g., diagnosis, prognosis, therapy selection, treatment monitoring, staging, or sub-typing) of a subject suffering from cancer (e.g., colon cancer, breast cancer, prostate cancer). The application can, for example, apply a predictive algorithm to the obtained TFBS data to generate an assessment (e.g., diagnosis, prognosis, therapy selection, treatment monitoring, staging, or sub-typing) of a subject suffering from cancer. The predictive algorithm can include an artificial intelligence-based predictor, e.g., a machine learning-based model, configured to process the obtained TFBS data to generate an assessment (e.g., diagnosis, prognosis, therapy selection, treatment monitoring, staging, or sub-typing) of a subject suffering from cancer.

[0209] The machine learning predictor can be trained using a dataset from one or more sets of cohorts of patients with cancer (e.g., a dataset generated by performing a TFBS assay on individual biological samples) as input to the machine learning predictor, and using the subjects' known diagnostic (e.g., staging and / or tumor fraction, subtype, treatment responder vs. non-responder (progressor vs. non-progressor) outcomes as output for the machine learning predictor.

[0210] A training dataset (e.g., a dataset generated by performing multi-analyte assays on an individual's biological sample) can be generated, for example, from one or more sets of subjects with common characteristics (features) and results (labels). The training dataset can include a set of features and labels corresponding to the features related to a diagnosis. The features can include features such as a range or category of cfDNA assay measurements, such as z-scores, accessibility scores, etc. For example, a set of features collected from a given subject at a given time point can collectively function as a diagnostic signature, which can be indicative of the subject's identified cancer at a given time point. The features may also include labels indicative of the subject's diagnostic outcome for one or more cancers, etc.

[0211] The label can include, for example, an outcome, such as a known diagnosis of the subject (e.g., stage, subtype, tumor fraction, or diseased vs. non-diseases). The outcome can include a characteristic associated with the subject's cancer. For example, the characteristic can indicate that the subject is afflicted with one or more cancers.

[0212] The training set (e.g., training data set) may be selected by random sampling of a set of data corresponding to one or more sets of subjects (e.g., a retrospective and / or prospective cohort of subjects (e.g., patients) with one or more cancers or without one or more cancers). Alternatively, the training set (e.g., training data set) may be selected by proportional sampling of a set of data corresponding to one or more sets of subjects (e.g., a retrospective and / or prospective cohort of subjects (e.g., patients) with one or more cancers or without one or more cancers). The training set may be balanced across multiple sets of data corresponding to one or more sets of subjects (e.g., patients from different clinical centers or clinical trials). A machine learning predictor may be trained until certain predefined conditions for accuracy or performance are met, such as having a minimum target value corresponding to a measure of diagnostic accuracy. For example, diagnostic accuracy may correspond to the prediction of the diagnosis, staging, or subtype of one or more cancers in subjects.

[0213] Examples of diagnostic accuracy include the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, and area under the curve (AUC) of a receiver operating characteristic (ROC) curve, which correspond to the diagnostic accuracy of detecting or predicting cancer (e.g., colon cancer).

[0214] In one aspect, the disclosure provides a computer-implemented method for detecting the presence or absence of disease in a subject or diagnosing disease in a subject, the method comprising: (a) providing a set of sequence reads of deoxyribonucleic acid (DNA) extracted from a subject; (b) using the set of sequence reads to generate a coverage pattern for a transcription factor; (c) processing the coverage pattern to provide a signal; and (d) processing the signal with a reference signal, wherein the signal and the reference signal have different frequencies, thereby detecting the presence or absence of disease in the subject or diagnosing disease in the subject.

[0215] In some instances, the DNA is cell-free DNA.

[0216] In some instances, the disease is cancer.

[0217] In some examples, (b) includes aligning the sequence reads to a reference sequence to provide an aligned sequence pattern, selecting a region of the aligned sequence pattern that corresponds to a binding site of a transcription factor, and normalizing the aligned sequence pattern in the region.

[0218] In some examples, (c) includes using a low pass filter. In some examples, (c) includes using a Savitzky-Golay filter.

[0219] In some examples, the subject is a human.

[0220] In some instances, the transcription factor is a cancer-specific transcription factor.

[0221] In some examples, accessibility scores for at least 2, or at least 5, or at least 10, or at least 15, or at least 20, or at least 25 transcription factor binding sites are determined and input into a machine learning model to train a classifier capable of distinguishing between healthy subjects and cancer patients.

[0222] In some examples, accessibility scores for at least 2, or at least 5, or at least 10, or at least 15, or at least 20, or at least 25 transcription factor binding sites are determined and input into a machine learning model to train a classifier capable of discriminating between disease progressors and non-progressors, disease subtypes among multiple disease subtypes, disease stages among multiple disease stages, or any combination thereof.

[0223] In some examples, accessibility scores for at least 2, or at least 5, or at least 10, or at least 15, or at least 20, or at least 25 transcription factor binding sites are determined and input into a machine learning model to train a classifier capable of discriminating between responders and non-responders to a disease treatment.

[0224] In one aspect, the methods disclosed herein allow for classification of patients by tumor type, including tumor subtype (e.g., prostate, colon, breast, lung cancer subtypes), which may have important clinical implications for patient management, including, for example, treatment regimens and responsiveness. Accordingly, the methods provided herein for mapping tumor-specific transcription factors in vivo based on patient samples (e.g., blood, plasma, or serum samples) render key portions of the non-coding genome suitable for clinical analysis.

[0225] In some instances, the methods distinguish between subtypes of a disease.

[0226] In some instances, the method distinguishes between subtypes of cancer.

[0227] In some examples, the method distinguishes between subtypes of prostate cancer, colon cancer, breast cancer, and lung cancer.

[0228] In some examples, the methods identify prostate cancer subtype patients as having prostate adenocarcinoma or small cell neuroendocrine prostate cancer.

[0229] In another aspect, the disclosure provides a computer-implemented method for monitoring disease progression or regression in a subject, the method comprising: (a) providing a first set of sequence reads of deoxyribonucleic acid (DNA) extracted from the subject at a first time and a second set of sequence reads of DNA extracted from the subject at a second time, the second time being later than the first time; (b) using the first set of sequence reads to generate a first coverage pattern for a transcription factor and using the second set of sequence reads to generate a second coverage pattern for the transcription factor; (c) providing a first set of sequence reads of deoxyribonucleic acid (DNA) extracted from the subject at a first time and a second set of sequence reads of DNA extracted from the subject at a second time, the second time being later than the first time; (d) processing the first coverage pattern to provide a first signal, and processing the second coverage pattern to provide a second signal; (d) processing the first signal with a reference signal, wherein the first signal and the reference signal have different frequencies; (e) processing a second signal with the reference signal, wherein the second signal and the reference signal have different frequencies; and (f) monitoring disease progression or regression of the subject based on processing of the first signal and the second signal with the reference signal.

[0230] In some instances, the DNA is cell-free DNA.

[0231] In some instances, the disease is cancer.

[0232] In some examples, (b) includes aligning the first set of sequence reads and the second set of sequence reads to a reference sequence to provide a first aligned sequence pattern and a second aligned sequence pattern, respectively; selecting regions of the first aligned sequence pattern and the second aligned sequence pattern that correspond to binding sites of a transcription factor; and normalizing the first aligned sequence pattern and the second aligned sequence pattern in the regions.

[0233] In some examples, (c) includes using a low pass filter. In some examples, (c) includes using a Savitzky-Golay filter.

[0234] In some examples, the subject is a human.

[0235] In some instances, the transcription factor is a cancer-specific transcription factor.

[0236] In a further aspect, the present disclosure provides a system for detecting or diagnosing a disease in a subject, the system comprising a processor configured to: (i) use deoxyribonucleic acid (DNA) sequence reads extracted from the subject to generate a coverage pattern for a transcription factor; (ii) process the coverage pattern to provide a signal, the signal having a different frequency from a reference signal; and (iii) provide detection or diagnosis of disease in the subject based on the signal.

[0237] In another aspect, the disclosure provides a system for monitoring disease progression or regression during or after a course of treatment of a subject, the system including a processor configured to: (i) use a first set of deoxyribonucleic acid (DNA) sequence reads extracted from the subject at a first time and a second set of DNA sequence reads extracted from the subject at a second time during or after treatment later than the first time to generate a first coverage pattern for transcription factors corresponding to a first set of sequence reads and a second coverage pattern for transcription factors corresponding to a second set of sequence reads; (ii) process the first coverage pattern to provide a first signal, and process the second coverage pattern to provide a second signal, wherein the first signal and the second signal have a different frequency than a reference signal; and (iii) monitor disease progression or regression during or after the course of treatment of the subject based on processing the first signal and the second signal with the reference signal.

[0238] In a further aspect, the present disclosure provides a system for detecting or diagnosing a disease in a subject, the system comprising a processor configured to: (i) use deoxyribonucleic acid (DNA) sequence reads extracted from the subject to generate a coverage pattern for a transcription factor; (ii) process the coverage pattern to provide a signal, the signal having a different frequency from a reference signal; and (iii) provide detection or diagnosis of disease in the subject based on the signal.

[0239] In some embodiments, the trained classifier is capable of determining that a subject is at least about 5%, at least about 10%, at least about 15%, at least about 20%, at least about 25%, at least about 30%, at least about 35%, at least about 40%, at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 81%, at least about 82%, at least about 83%, at least about 84%, at least about 85%, at least about 86%, at least about 87%, at least about 88%, at least about 89%, at least about 90%, at least about 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, at least about 96%, at least about 97%, at least about 98%, at least about 99% or more at risk for a disease, disorder, or abnormal condition (e.g., cancer).

[0240] The trained classifier may be at least about 50%, at least about 55%, at least about 60%, at least about 65%, at least about 70%, at least about 75%, at least about 80%, at least about 81%, at least about 82%, at least about 83%, at least about 84%, at least about 85%, at least about 86%, at least about 87%, at least about 88%, at least about 89%, at least about 90%, at least about 91%, at least about 92%, at least about 93%, at least about 94%, at least about 95%, %, at least about 96%, at least about 97%, at least about 98%, at least about 99%, at least about 99.1%, at least about 99.2%, at least about 99.3%, at least about 99.4%, at least about 99.5%, at least about 99.6%, at least about 99.7%, at least about 99.8%, at least about 99.9%, at least about 99.99%, at least about 99.999% or more accurately, a subject can be determined to be at risk for a disease, disorder, or abnormal condition.

[0241] Once a subject is identified as having a disease, disorder, or abnormal condition, a therapeutic intervention can be provided to the subject (e.g., prescribing an appropriate course of treatment to treat the subject's disease, disorder, or abnormal condition). Therapeutic intervention includes prescribing an effective amount of a drug, further testing or evaluation of the disease, disorder, or abnormal condition, further monitoring of the disease, disorder, or abnormal condition, or a combination thereof. If the subject is currently undergoing treatment for a disease, disorder, or abnormal condition, the course of treatment, then the therapeutic intervention, can include a subsequent different course of treatment (e.g., to increase the effectiveness of the treatment because the current course of treatment is not effective).

[0242] Therapeutic intervention includes recommending a subject undergo secondary clinical trials to confirm the diagnosis or other evaluation of a disease, disorder, or abnormal condition. This secondary clinical trial may include imaging tests, blood tests, computed tomography (CT), magnetic resonance imaging (MRI), ultrasound scans, chest x-rays, positron emission tomography (PET), PET-CT scans, cytology assays, or any combination thereof.

[0243] Multiple input variables (e.g., TFBS information) may be evaluated over a period of monitoring a patient (e.g., a subject having or undergoing treatment for a disease, disorder, or abnormal condition). In such cases, the input variables (e.g., TFBS information) of a patient sample may change during the course of treatment. For example, the TFBS information of a patient whose risk of a disease, disorder, or abnormal condition has been reduced due to effective treatment may change toward a profile or distribution of a healthy subject (e.g., a subject without a disease, disorder, or abnormal condition). Conversely, the TFBS information of a patient whose risk of a disease, disorder, or abnormal condition has been increased due to ineffective treatment may change toward a profile or distribution of a subject at higher risk of a disease, disorder, or abnormal condition or in a more advanced state of the disease, disorder, or abnormal condition.

[0244] The disease, disorder, or abnormal condition of a subject can be monitored by monitoring a course of treatment for treating the disease, disorder, or abnormal condition of the subject. The monitoring can include evaluating the TFBS information of the subject at two or more time points. The evaluation can be based on the TFBS information determined at each of at least two or more time points.

[0245] In some examples, the difference in TFBS information determined between two or more time points can be indicative of one or more clinical indications, such as: (i) detection or diagnosis of a disease, disorder, or abnormal condition in a subject; (ii) prognosis of a disease, disorder, or abnormal condition in a subject; (iii) an increased risk of a disease, disorder, or abnormal condition in a subject; (iv) a decreased risk of a disease, disorder, or abnormal condition in a subject; (v) the effectiveness of a course of treatment for treating a disease, disorder, or abnormal condition in a subject; (vi) the ineffectiveness of a therapeutic course of treatment for treating a disease, disorder, or abnormal condition in a subject.

[0246] In some examples, the difference in TFBS information determined between two or more time points may indicate a diagnosis of a disease, disorder, or abnormal condition in a subject. For example, if a disease, disorder, or abnormal condition is not detected in a subject at an early time point, but is detected in the subject at a later time point, the difference may indicate detection or diagnosis of the disease, disorder, or abnormal condition in the subject. A clinical action or decision may be made based on this indication of detection or diagnosis of the disease, disorder, or abnormal condition in the subject, such as, for example, prescribing a new therapeutic intervention to the subject. The clinical action or decision may include recommending a secondary clinical trial to confirm the diagnosis of the disease, disorder, or abnormal condition to the subject. This secondary clinical trial may include imaging tests, blood tests, computed tomography (CT), magnetic resonance imaging (MRI), ultrasound scans, chest x-rays, positron emission tomography (PET), PET-CT scans, cytology assays, or any combination thereof.

[0247] In some examples, differences in the TFBS information determined between two or more time points may indicate a prognosis of a disease, disorder, or abnormal condition in a subject.

[0248] In some examples, the difference in TFBS information determined between two or more time points may indicate an increased risk of disease, disorder, or abnormal condition in the subject. For example, if a disease, disorder, or abnormal condition is detected in the subject at both an earlier time point and a later time point, the difference may indicate an increased risk of disease, disorder, or abnormal condition in the subject. A clinical action or decision may be made based on this indication of an increased risk of disease, disorder, or abnormal condition in the subject, such as prescribing a new therapeutic intervention or switching the therapeutic intervention (e.g., terminating the current treatment and prescribing a new treatment). The clinical action or decision may include recommending a secondary clinical trial for the subject to confirm the increased risk of disease, disorder, or abnormal condition. The secondary clinical trial may include imaging tests, blood tests, computed tomography (CT), magnetic resonance imaging (MRI), ultrasound scans, chest x-rays, positron emission tomography (PET), PET-CT scans, cytology assays, or any combination thereof.

[0249] In some examples, the difference in TFBS information determined between two or more time points may indicate a reduced risk of the disease, disorder, or abnormal condition in the subject. For example, if a disease, disorder, or abnormal condition is detected in the subject at both an earlier time point and a later time point, the difference may indicate a reduced risk of the disease, disorder, or abnormal condition in the subject. A clinical action or decision may be made based on this indication of reduced risk of the disease, disorder, or abnormal condition, such as prescribing a new therapeutic intervention or switching the subject's therapeutic intervention (e.g., continuing or terminating current treatment). The clinical action or decision may include recommending a secondary clinical trial for the subject to confirm an increased risk of the disease, disorder, or abnormal condition. This secondary clinical trial may include imaging tests, blood tests, computed tomography (CT), magnetic resonance imaging (MRI), ultrasound scans, chest x-rays, positron emission tomography (PET), PET-CT scans, cytology assays, or any combination thereof.

[0250] In some examples, the difference in TFBS information determined between two or more time points may indicate the effectiveness of a course of treatment to treat a disease, disorder, or abnormal condition of a subject. For example, if a disease, disorder, or abnormal condition is detected in a subject at an early time point, but not in the subject at a later time point, the difference may indicate the effectiveness of a course of treatment to treat the disease, disorder, or abnormal condition of the subject. A clinical action or decision may be made based on this indication of the effectiveness of a course of treatment to treat the disease, disorder, or abnormal condition of the subject, for example, to continue or terminate the subject's current therapeutic intervention. The clinical action or decision may include recommending a secondary clinical trial for the subject to confirm the effectiveness of a course of treatment to treat the disease, disorder, or abnormal condition. This secondary clinical trial may include imaging tests, blood tests, computed tomography (CT), magnetic resonance imaging (MRI), ultrasound scans, chest x-rays, positron emission tomography (PET), PET-CT scans, cytology assays, or any combination thereof.

[0251] In some examples, the difference in TFBS information determined between two or more time points may indicate the ineffectiveness of a course of treatment for treating a disease, disorder, or abnormal condition of a subject. For example, if a disease, disorder, or abnormal condition is detected in a subject at both an early time point and a later time point, and if effective treatment is shown at the early time point, the difference may indicate the ineffectiveness of a course of treatment for treating a disease, disorder, or abnormal condition of a subject. A clinical action or decision may be made based on the ineffectiveness of a course of treatment for treating a disease, disorder, or abnormal condition of a subject, for example, terminating a current therapeutic intervention and / or switching (e.g., prescribing) a different new therapeutic intervention for the subject. The clinical action or decision may include recommending a secondary clinical trial to the subject to confirm the ineffectiveness of a course of treatment for treating a disease, disorder, or abnormal condition. This secondary clinical trial may include imaging tests, blood tests, computed tomography (CT), magnetic resonance imaging (MRI), ultrasound scans, chest x-rays, positron emission tomography (PET), PET-CT scans, cytology assays, or any combination thereof.

[0252] B. Indications Non-limiting examples of cancers that can be predicted by the disclosed methods include acute lymphoblastic leukemia (ALL), acute myeloid leukemia (AML), adrenocortical carcinoma, Kaposi's sarcoma, anal cancer, basal cell carcinoma, bile duct cancer, bladder cancer, bone cancer, osteosarcoma, malignant fibrous histiocytoma, brain stem glioma, brain cancer, craniopharyngioma, ependymoblastoma, ependymoma, medulloblastoma, medulloepithelioma, pineal parenchymal tumor, and the like. tumor), breast cancer, bronchial tumor, Burkitt's lymphoma, non-Hodgkin's lymphoma, carcinoid tumor, cervical cancer, chordoma, chronic lymphocytic leukemia (CLL), chronic myeloid leukemia (CML), colon cancer, colorectal cancer, cutaneous T-cell lymphoma, ductal carcinoma in situ, endometrial cancer, esophageal cancer, Ewing's sarcoma, eye cancer, intraocular melanoma, retinoblastoma, fibrous histiocytoma, gallbladder cancer, gastric cancer, glioma, hairy cell leukemia, head and neck cancer, heart cancer, hepatocellular (liver) cancer, Hodgkin's lymphoma, hypopharyngeal cancer, renal cancer, laryngeal cancer, lip cancer, oral cavity cancer, lung cancer, non-small cell lung cancer, melanoma, mouth cancer, myelodysplastic syndrome, multiple myeloma, medulloblastoma, nasal cavity cancer, paranasal sinus cancer, neuroblastoma, nasopharyngeal cancer, oral cancer cancer), oropharyngeal cancer, osteosarcoma, ovarian cancer, pancreatic cancer, papilloma, paraganglioma, parathyroid cancer, penile cancer, pharyngeal cancer, pituitary tumor, plasma cell neoplasm, prostate cancer, rectal cancer, renal cell carcinoma, rhabdomyosarcoma, salivary gland cancer, Sezary syndrome, skin cancer, small intestine cancer, soft tissue sarcoma, squamous cell carcinoma, testicular cancer, laryngeal cancer, thymoma, thyroid cancer, urethral cancer, uterine cancer, uterine sarcoma, vaginal cancer, vulvar cancer, Waldenstrom's macroglobulinemia, and Wilms' tumor.

[0253] In various examples, the tumor is a colon disease selected from the group consisting of colon cancer, advanced adenoma, ulcerative colitis, Crohn's disease, and irritable bowel syndrome (IBS).

[0254] In some instances, colon cancer is classified by stage, such as stage 0, stage I, stage IIA, stage IIB, stage IIC, stage IIIA, stage IIIB, stage IIIC, stage IVA, stage IVB, or stage IVC. EXAMPLES

[0255] Example 1: Prediction of cell-specific transcription factor binding from cell-free DNA enables tumor subtype prediction and early detection of cancer

[0256] In accordance with the methods and systems of the present disclosure, an analytical program was developed to determine the accessibility of transcription factor binding sites, and the program was applied to 244 cfDNA samples from patients with prostate cancer, breast cancer, or colon cancer.

[0257] Inferring TF binding from cfDNA has enormous diagnostic potential in cancer and beyond, and an improved and optimized bioinformatics pipeline was developed. This process resolves the components involved in nucleosome signatures in TFBSs, allowing for objective assessment and comparison of TFBS accessibility in various plasma samples. To validate this pipeline for clinical purposes, deep whole genome sequencing (WGS) data were obtained from 24 plasma samples from healthy donors and 15 plasma samples from patients with metastatic prostate, colon, or breast cancer, where cfDNA also contains circulating tumor DNA (ctDNA). In addition, shallow WGS data were generated for 229 plasma samples from patients with the aforementioned tumor entities (totaling more than 18.5 billion mapped plasma sequence reads). An additional 769 plasma samples from colon cancer patients (n=592) and healthy controls (n=177) were also included (providing approximately 238 billion mapped plasma sequence reads). This approach profiles individual TFs instead of using a mixture of cfDNA signals from multiple cell types and Fourier transform analysis to establish common tissue-specific patterns, as in other approaches. The methods and systems provided herein provide insight into both tissue contributions and biological processes, allowing the identification of lineage-specific TFs suitable for both tissue and tumor origin analysis. Furthermore, the plasticity of TFBS in cfDNA from cancer patients and the potential of TFs to classify prostate cancer subtypes are illustrated by two examples of relevant clinical applications. First, these TF-based cfDNA assays can differentiate between prostate adenocarcinoma and small cell neuroendocrine prostate cancer, a distinction with important therapeutic implications. Second, a large colon cancer cohort allowed for the precise establishment of separation limits of TF-based plasma analysis for the detection of early stages of cancer and the exploration of its use.

[0258] The ability to use cfDNA fragmentation patterns to detect non-hematopoietic signatures was demonstrated by analysis of a small panel of individuals (n=5) with advanced cancer. To explore the potential of TF-nucleosome interaction mapping from cfDNA in more detail, known hematopoietic TF nucleosome footprints were confirmed in plasma samples of healthy controls. A curated list of TFBSs from the Gene Transcription Regulation Database (GTRD) was annotated with a recently published list of known or likely human TFs to generate a comprehensive TFBS nucleosome occupancy map of cfDNA for 676 TFs. Using the bioinformatics pipeline provided herein, we measure nucleosome signatures at TFBSs by evaluating various stringency criteria and establish a metric called an "accessibility score" and a z-score statistic to objectively compare significant changes in TFBS accessibility across various plasma samples. For clinical purposes, a set of lineage-specific TFs was used to identify the tissue origin of cfDNA or, in cancer patients, the tumor origin. Finally, accessibility scores and z-score statistics were used to elucidate the altered TFBS accessibility in cfDNA from cancer patients.

[0259] Knowing the exact location of nucleosomes in the genome relative to TF binding sites (TFBS) is useful to understand how genes are regulated. To that end, the analysis of cell-free DNA (cfDNA) from plasma (which also contains circulating tumor DNA (ctDNA) from cancer patients) improves the opportunity to non-invasively study TFBS in humans in vivo. cfDNA is mainly released from apoptotic cells after enzymatic digestion and circulates mostly as mononucleosomal DNA. Thus, sequencing of cfDNA fragments allows the generation of nucleosome maps, in which "perfectly positioned" nucleosome dyads, e.g., sites with high nucleosome preference, result in strong peaks of reads reflecting nucleosome topology, while less preferentially positioned nucleosome dyads showed reduced peaks or no peaks at all. Thus, cfDNA represents a unique analyte generated by endogenous physiological processes, which allows the generation of in vivo maps of nucleosome occupancy by whole genome sequencing. This can be exploited to infer expressed genes by detailed analysis of nucleosome occupancy at transcription start sites (TSSs). cfDNA nucleosome occupancy can reflect the footprint of TFs. In a small panel of cancer-bearing individuals (n=5), cfDNA fragmentation patterns are matched against a reference dataset to detect non-hematopoietic signatures. However, TF-nucleosome interactions remain largely unmapped, and measurements of real-time dynamics of TFs at a genome-scale in vivo in humans are needed.

[0260] The strategy of nucleosome position mapping from cfDNA and bioinformatics pipeline is used to address the following questions: (1) whether cfDNA accurately reflects known TF-nucleosome interactions; (2) to generate the most comprehensive TF-nucleosome interaction map, including data on 676 TFs; (3) to objectively compare TFBS accessibility in consecutive analyses from the same or between different individuals to establish an improved metric, called the "accessibility score"; and (4) to define a set of lineage-specific TFs suitable for identifying the tissue origin of cfDNA or tumor origin in cancer patients. In addition, this study also explored whether TFBS tracking from cfDNA of cancer patients could elucidate altered TFBS accessibility and associated pathways. To this end, high-coverage whole genome sequencing (WGS) data were obtained from 24 plasma samples of healthy donors (12 males and 12 females) and from 16 plasma samples of patients with metastatic prostate, colon, or breast cancer. Furthermore, for validation purposes, shallow coverage sequencing data of cfDNA from 229 patients was used along with the aforementioned tumor entities to provide a broad in vivo view of a significant portion of the non-coding genome, generating over 18.3 billion mapped plasma sequence reads overall. Inferred nucleosome occupancy from cfDNA shows characteristic TF binding footprints.

[0261] Nucleosome occupancy maps at TFBS were generated to test for similarities and differences between healthy individuals and cancer patients. To do so, high-coverage cfDNA samples were obtained from 24 healthy controls (12 men and 12 women), where the majority (>90%) of cfDNA was derived from apoptotic leukocytes with minimal contribution from other tissues, and 11 plasma samples were derived from seven patients with three common tumor entities (e.g., four prostate cancer cases (P40, P147, P148, and P190), one colorectal cancer case (CRC; C2), and two breast cancer cases (B7 and B13)), with ctDNA percentages ranging from 18-78% (Figure 5).

[0262] We used 676 TFs from the Gene Transcription Regulation Database (GTRD; version 18.01) because it contains detailed TFBS information based on ChIP-seq data of various tissue samples. We annotated the TFs using an up-to-date curated list of 1,639 known or probable human TFs (Figure 6A). Because of the potentially large number of TFBSs bound by TFs with variable frequency, we defined three different stringency criteria (Figure 6A): first, all TFBSs for all tissue samples in the GTRD; second, those peaks supported by more than 50% of the maximum number of samples (hereafter referred to as ">50%-TFBS"; these two analyses included all 676 GTRD TFs); and third, 1,000 TFBSs per TF supported by the majority of samples ("1,000-msTFBS"; 505 TFs met this criterion).

[0263] Establishing TF-nucleosome interactions Figure 2A-Figure 2E show the establishment of TF-nucleosome interactions from cell-free deoxyribonucleic acid (cfDNA). Figure 2A shows that regions with highly organized, e.g., phased, nucleosomes provide a fluctuating read depth pattern, with peaks of reads indicating the location of the dyad, e.g., the midpoint of a canonical nucleosome. Less well-defined nucleosome locations result in a rather flat coverage profile. Figure 2B shows that TFBS data for 676 TFs were retrieved from the GTRD and aligned to a curated list of known or likely human TFs. Three different calculations, each with increasing stringency, were performed. Figure 2C shows that the coverage patterns of CCCCTC-binding factors (CTCFs) are similar across all cfDNAs analyzed, which is in agreement with the DNase hypersensitivity data showing nearly equal accessibility in blood (GM12878) and epithelial tissues, e.g., prostate (LNCaP) and colon (HCT116). In this and each subsequent panel, profiles calculated from healthy controls are shown in grey, while patient-derived profiles are displayed in the indicated colours. Figure 2D shows that the hematopoietic lineage specificity of TFs (PU.1, LYL1, SPIB) was confirmed by DNA hypersensitivity assays, and their amplitudes were decreased in plasma from cancer patients compared to healthy controls. In contrast, the amplitude of the epithelial TF GRHL2 was increased in cfDNA from cancer patients. Figure 2E shows accessibility plots and DNase hypersensitivity for the TF FOXA1, illustrating preferential amplitude changes in patients with hormone-dependent cancers, e.g., prostate and breast cancer.

[0264] Using 24 cfDNA samples from healthy controls, we obtained an average of 435,135,450 (range: 352,904,231–556,303,420) sequencing reads per sample. TF binding sites were often flanked by tightly arranged arrays of nucleosomes, visible as periodic oscillating patterns (Figure 6A-E). In contrast, using negative control normal high molecular weight DNA, we observed uniform coverage across TFBSs (Figure 8A-E). CTCF binding sites surrounded by tightly arranged arrays of nucleosomes resulted in oscillating coverage patterns that remained similar across all analyzed samples, regardless of whether cfDNA was derived from healthy controls or cancer patients (Figure 2C). These results were consistent with DNase hypersensitivity assays from the Encyclopedia of DNA Elements (ENCODE) database for the cell lines GM12878 (a B cell line from a female donor of European descent), LNCaP (an androgen-sensitive human prostate adenocarcinoma cell line), and HCT116 (a human colon carcinoma cell line) ( Fig. 2C ).

[0265] ctDNA in plasma of cancer patients altered the equilibrium between DNA of hematopoietic versus epithelial cells compared to healthy controls, resulting, for example, in decreased amplitudes of the lineage-restricted hematopoietic TFs purine-rich box 1 (PU.1), LYL1 (lymphocytic leukemia 1), and lymphocyte lineage-restricted transcription factor SPIB, as well as increased amplitudes of the TF GRHL2, a pioneer TF for epithelial cells, in cancer-derived samples (Figure 2D). We also confirmed the lineage specificity of these TFs using publicly available DNase hypersensitivity assay data (Figure 2D). As another example of a well-established TF, we analyzed FOXA1, which cooperates with nuclear hormone receptors in endocrine-driven tumors of the breast and prostate. In agreement with the DNase hypersensitivity assay, we observed preferentially increased accessibility for FOXA1 in plasma samples of prostate and breast cancer patients. A comparison was also performed with ENCODE data, in which mononucleosome-bound DNA fragments were generated by micrococcal nuclease (MNase) digestion (Figure 6A-E and Figure 7). A coverage-independent analysis was performed (Fig. 6B) and the spatial density of cfDNA fragments associated with a single recognition sequence was calculated (Fig. 6C). Sequence-specific TFs may have canonical motifs and significant secondary motifs, which may correspond to motifs of other TFs. A catalog of TFBSs was also generated, which may be affected by the simultaneous binding of more than one TF for all 676 TFs and 505 TTFs from the 1,000-msTFBS (Fig. 6D). Furthermore, using purified high molecular weight DNA as a negative control, we observed uniform coverage across TFBSs (Fig. 6E). Accordingly, these results showed that the corresponding TFBS coverage profiles were very similar to each other, thereby indicating a high accuracy of the approach, and that the patterns obtained for a given TF were reproducible across all samples.

[0266] Since sequence-specific TFs may have canonical motifs and significant secondary motifs that correspond to motifs in other TFs, overlap was calculated between various TFBSs (Figure 6D). A list of TFBSs that could be affected by the simultaneous binding of more than one TF was generated (Figure 27). Examples of such overlap effects are the TFs SP1, SP2, NF-YA, and NF-YB, with overlaps ranging between 10 and 36%, where the TFBS nucleosome profiles were indeed similar (Figure 13A). The predominant origin of cfDNA from blood was notably reflected in well-positioned nucleosomes adjacent to the binding sites of lineage-restricted hematopoietic TFs, such as purine-rich box 1 (PU.1), LYL1 (lymphocytic leukemia 1), and lymphocyte lineage-restricted transcription factor SPI-B (Figure 13A). In contrast, the TFBS profile of GRHL2, a pioneer TF in epithelial cells, showed a substantial decrease in amplitude (Figure 13A).

[0267] Applying the three different stringency criteria described above, we assessed CTCF binding sites surrounded by arrays of tightly spaced nucleosomes and preferentially observed the expected oscillating patterns for >50%-TFBSs and 1,000-ms-TFBSs (Figure 2C-Figure 2E). Furthermore, we assessed individual binding sites separately using CTCF (Figure S17) and, as a further confirmation, performed a coverage-independent analysis (Figure S17) to calculate the spatial density of cfDNA fragments associated with a single recognition sequence. The resulting heatmap showed that the nucleosome phase of most analyzed sites was uniform, which is in agreement with the coverage profile.

[0268] Figure 7 shows TF-nucleosome interaction maps for various TFs, with additional comparison of cfDNA and MNase-seq coverage profiles around transcription factor binding sites.

[0269] CTCF as an unusual example of a TF with multiple distinct binding sites To explore different TFBSs of the same TF, we used CCCTC-binding factor (CTCF). CTCF is present at 55,000-65,000 binding sites in mammalian genomes. Of these sites, about 5,000 are ultraconserved, about 50% are in intergenic regions, about 15% are near promoters, and about 40% are intragenic. Furthermore, chromosomes are divided into multiple evolutionarily conserved higher-order chromosomal structures, termed topologically associated domains (TADs), whose boundaries are enriched for binding sites of CTCF and cohesin. In mammals, 15% of the genome's CTCF binding sites are present at TAD boundaries, while the other 85% are within TADs.

[0270] CTCF sites that overlap or are outside the TAD boundaries were analyzed separately, as were ultraconserved sites proximal (e.g., within about 2 kbp) or distal (>2 kbp) to the TSS. Analysis was performed with all tissue types in the GTRD, yielding a variety of CTCF coverage patterns, with ultraconserved CTCF sites having the greatest amplitude (Figure 14A, left panel). When analysis was limited to binding sites called in more than 50% of all samples in the GTRD, the resulting profiles became more similar to each other (Figure 14A, right panel). As a coverage-independent confirmation of the TFBS signal, the length of each cfDNA fragment was plotted as a function of the distance of the fragment midpoint to the CTCF binding site. The resulting heatmap confirmed a periodicity of the signal consistent with the coverage-based oscillating pattern (Figure 6B). In addition, to more closely analyze the landscape of fragments associated with a single recognition sequence, the spatial density of cfDNA fragments was calculated within a 2 kb region centered on the TFBS, and the sites were ranked according to their coverage of the central 40 bp. The resulting heatmap showed that the nucleosome topology of most of the analyzed sites was uniform, again consistent with the coverage profile (Figure 6C).

[0271] The "accessibility score" allows accurate inference of TF binding from cfDNA. TF binding mediates nucleosome positioning, ensuring that the respective DNA is accessible to proteins and the transcription and replication machinery. Some TFs show evenly spaced nucleosomal peaks with their binding sites (e.g., PU.1 and GRHL2 in Figure S1B), whereas other TFs have broader troughs at their binding sites (e.g., Figure S1A) that resemble the troughs of TSSs. For the latter, we measured substantial binding site width differences for TFs (Figure S1D). This measurement identified 55 TFBSs with TFBSs >300 bp, 26 of which had binding sites close to the size of dinucleosomes (312–352 bp) (Figure S28A–28B). To test whether these patterns are a side effect of binding to CpG island promoters, we generated plots of CpG density (box plots) and colocalization with CpG islands (bar / pie charts) for the 55 broad TFs vs. TFs with narrowly defined binding sites.

[0272] Certain lineage-specific TFs are relevant for determining the tissue origin of plasma DNA. However, determining which TFs are useful for such applications requires assessing the accessibility of TFs, for example, at their binding sites in cfDNA. Traditional methods may not assess TF accessibility at their binding sites in cfDNA as a surrogate for their activity. To implement such an approach, we examined TF-specific nucleosome coverage profiles. Calculations were performed separately for TFBSs inside or outside the transcription start site (TSS) (Figure 8A), and for all GTRD tissues vs. >50%-TFBS (Figure 8B). These analyses show that the average TFBS pattern contains two signals that correspond to more evenly spaced peak signals, namely, TSS-proximal (within about 2 kb of the TSS resulting in a "low frequency pattern") and TSS-distal (more than 2 kb away from the TSS peak resulting in a "high frequency pattern"). To suppress effects on coverage not contributed by preferential nucleosome positioning and to remove local biases from the nucleosome data, detrending was performed using a Savitzky-Golay filter (Figure 3A). The resulting low-frequency signals were then used to normalize the high-frequency signals, and the data range of the high-frequency signals (maximum data value minus minimum, corresponding to amplitude) was then recorded. As the range of the high-frequency signals depends on the number of TFBSs (except for the 1,000-ms TFBS), these range values ​​were corrected by LOESS smoothing, and then a rank was calculated as an index of the accessibility of each TFBS. Figure 3C shows wavelet analysis of GRHL2: a heatmap of the period along the region surrounding the TFBS of GRHL2 (left panel). The color code represents the quantile of the signal power distribution. Average power of the period of the transcription factor GRHL2 (right panel). Figure 3D shows the detrended original (black) and reconstructed (red) nucleosome coverage profiles of the transcription factor GRHL2 obtained as a result of wavelet analysis.

[0273] To test potential alternatives for TF accessibility assessment, unbiased detrended signals for the period between 135 and 235 bp were reconstructed by wavelet analysis and the power of the signals was summed over 2,000 bp of adjacent TFBSs (Figure 3E-F). To evaluate the performance of Savitzky-Golay filtering and wavelet analysis, we used cfRNA data and observed a significantly reduced accessibility of unexpressed TFs (e.g., FPKM [Fragments Per Kilobase Million] of <0.01) compared to the accessibility of expressed TFs (e.g., FPKM >10) (>50%-TFBS; Savitzky-Golay filtering: p=1.75x10 -13 ; Power sum (wavelet analysis): p = 0.0004049; 1,000-ms TFBS; Savitzky-Golay filtering: p = 1.254x10 -11 ; Mann-Whitney U test, respectively) (Figure ​(Figure14B). Furthermore, when the adjusted coverage was compared to the average DNase coverage, these differences were significant (>50%-TFBS; Savitzky-Golay filtering: p<2.2x10 -16 ; Power sum (wavelet analysis): p<2.2x10 -16 ;1,000-msTFBS;Savitzky-Golay filtering: p<2.2x10 -16 ; Mann-Whitney U test, respectively). This approach was preferred because Savitzky-Golay filtering performed slightly better, and we subsequently defined the detection threshold for TFBS accessibility deviating from normal samples as ±3 standard deviations of the mean (as a z-score of 3). For evaluation based on all or >50%-TFBSs, the detection thresholds for normalized accessibility scores were ±253 and ±88 of the 1,000-ms TFBSs, respectively, with fewer analyzable TFs (Figure 8C).

[0274] In addition, a comprehensive TF-nucleosome interaction map for 676 GTRD TFs from cfDNA was generated (FIG. 14B, FIG. 17). TF-nucleosome interactions may be mapped using, for example, the ChIP-seq dataset of the ENCODE consortium, and chromatin structures around 119 human TFs were characterized. These efforts resulted in Factorbook, a TF-centric web repository that encompasses data on 167 TFs. However, these data are based on ex vivo tissue samples, while in vivo accessibility generated by endogenous processes is discussed herein.

[0275] These results demonstrate a robust approach for assessing TFBS accessibility with particular utility in clinical diagnostics using cfDNA.

[0276] TFBS accessibility of cfDNA across cell types To demonstrate clinical application, we used plasma samples from three common tumor entities. In this study, we first analyzed 11 plasma samples from seven patients, including four prostate cancer cases (P40, P147, P148, and P190), one colorectal cancer case (CRC; C2), and two breast cancer cases (B7 and B13) (Figure 4A-F). cfDNA from C2, P40, P147, and P148 was sequenced on an Illumina NovaSeq platform with an average of 688,482,254 sequencing reads (range: 541,216,395–870,285,698), while B7 (328,515,075 reads) and B13 (379,733,061 reads) were sequenced on an Illumina NextSeq platform.

[0277] CTCF is a special transcription factor active in any tissue and conserved across tissues, as it regulates chromosome 3D structure. Regardless of whether cfDNA was derived from healthy controls or cancer patients, the amplitude of CTCF remained similar across all analyzed samples (Figure 2C). This is consistent with DNase hypersensitivity assays from the ENCODE database for cell lines GM12878, LNCaP (an androgen-sensitive human prostate adenocarcinoma cell line), and HCT116 (a human colon cancer cell line), showing increased accessibility of CTCF binding sites across various tissues (Figure 2C). However, the ctDNA fraction in cancer patients is increased, which alters the equilibrium between hematopoietic-derived DNA and epithelial cells in cfDNA. Thus, the amplitude of hematopoietic TFs (PU.1, Lyl-1, Spi-B) was decreased, whereas the amplitude of epithelial TF GRH-L2 was increased, which indicates a dilution of the hematopoietic contribution and an increase of the epithelial contribution (Figure 2D). These observations were again consistent with the DNase hypersensitivity assay (Fig. 2D ).

[0278] We analyzed another example of a well-established TF, FOXA1, which is widely expressed in various tissues where cell differentiation and organ function are controlled. Furthermore, FOXA1 cooperates with nuclear hormone receptors in endocrine-derived tumors of the breast and prostate, and its expression in the prostate has been linked to castration-resistant prostate cancer (CRPC). Indeed, we observed preferentially increased FOXA1 accessibility in agreement with the DNase hypersensitivity assay in plasma samples from prostate and breast cancer patients (Figure 2E).

[0279] Inference of TF binding from cfDNA supports molecular subtyping in prostate cancer In some cases, it may be important to assess the extent to which tissue-specific TFs are relevant for identifying tumor origin and molecular subtyping. For this purpose, prostate cancer is a particularly interesting tumor entity because a frequent (~20%) mechanism in the development of therapeutic resistance to novel agents targeting the AR pathway, such as abiraterone and enzalutamide, is the transdifferentiation of adenocarcinoma into treatment-emergent small cell neuroendocrine prostate cancer (t-SCNC). This transdifferentiation has enormous clinical implications due to the need for therapeutic changes, and the involvement of various TFs in such a transdifferentiation process may be studied (Figure 4A).

[0280] Various TFs were detected with increased accessibility in one tumor entity but not in the other, for example, we observed increased accessibility to c-jun and JunD (Figure 15) oncogenes in plasma samples from patient C2 with CRC, and confirmed colonic bias by DNA hypersensitivity assays (Figure 15).

[0281] Other analyses were performed on prostate cancer samples (Lambert et al., 2018). We screened the data against the expression of human TFs across donor tissues and cell types and the publicly available human protein atlas, confirming the well-established prostate lineage specificity of the TFs AR, HOXB13, and NKX3-1, which was also reflected in DNase hypersensitivity assays of the prostate cancer cell line LNCaP (Figure 4B–D).

[0282] HOXB13 is a highly lineage-specific homeobox TF gene that is important for prostate progression and maintains high expression levels in normal prostate into adulthood (Figure 4B). NKX3-1 homeobox gene is one of the earliest genes expressed during prostate epithelial maturation, is important for prostate epithelial differentiation, and is required for prostate tumor progression (Figure 4C). Both TFs showed increased accessibility at their binding sites only in cfDNA from prostate cancer patients, and the tissue specificity was further confirmed by DNase hypersensitivity assays (Figure 4C and Figure 4B-D).

[0283] Thus, we observed increased accessibility at the binding sites of these TFs only in cfDNA of prostate cancer patients. Due to the aberrant association with AR in prostate cancer, we not only used AR binding sites as defined by GTRD, but also those reported by (Pomerantz et al., 2015), to analyze 9,179 tumor AR binding sites (tumor AR binding sites, T-ARBS) of AR cistrome identification with high binding intensity in tumors, and 2,690 normal AR binding sites with high binding intensity in normal samples. Indeed, we observed increased accessibility from T-ARBS in patient plasma samples, while N-ARBS were inaccessible from both controls and patients (Figure 4D).

[0284] Confirmation of lineage-specific transcription factors in pooled samples This approach can also be applied to samples sequenced at low coverage and highly rearranged. To test this, we repeated the TF analysis after downsampling P148_1 (819,607,690 reads) and P148_3 (768,763,081 reads) to approximately 50 million reads. This comparison revealed that the same TFs were identified as either increased or decreased accessibility, demonstrating that samples with small sequencing reads are acceptable for these analyses. We subsequently analyzed four cfDNA samples (P21_2, P111_1, P111_4, P166_1), one of which had a tandem overlap phenotype (Viswanathan et al., 2018) and one of which had chromosomal thrombus on chromosome 10 (mean 52,869,911, range 41,780,819-84,049,593) (Figure 18A-B). In these cases, we again observed increased accessibility from the epithelial TFs FOXA1, GRHL1, and GRHL2, as well as the prostate lineage-specific TFs AR, HOXB13, and NKX3-1 (Figure 18A-B), indicating that outcomes can be achieved even under hindered requirements and that altered accessibility of these TFs appears to be a universal feature in prostate cancer.

[0285] To further confirm the robustness and reproducibility of lineage-specific TFs in cfDNA, we analyzed a collection of multiple cfDNA samples generated at shallow coverage (<0.2x) and observed that TFs with increased accessibility in most or all of the samples, lineage-specific TFs had elevated accessibility scores while others averaged out. To this end, cfDNA samples were collected separately for prostate cancer cases (n=69), colon cancer cases (n=100), and breast cancer cases (n=60) and the analysis was repeated. Accessibility of the epithelial TF GRHL2 continued to increase, whereas accessibility of hematopoietic TFs decreased (Figure 8A-E). Within the prostate cancer cfDNA collection, we observed increased accessibility for the lineage-specific TFs AR (340; 4.0), HOXB13 (712; 8.4), and NKX3-1 (253; 3.0), demonstrating that altered accessibility of these TFs is a universal feature in prostate cancer (Figures 8A-E), and that these features are universally present in prostate cancer and may be suitable for identifying tumor origin from cfDNA.

[0286] Figure 9 shows an analysis of pooled shallow coverage cfDNA. Accessibility of epithelial TF GRHL2 and hematopoietic TFs (PU.1, LYL1, and SPIB) is shown for pooled cfDNA samples from prostate cancer (n=69), colon cancer (n=100), and breast cancer (n=60) cases. Accessibility of lineage-specific TFs AR, HOXB13, and NKX3-1 within the prostate cancer cfDNA pool is also shown.

[0287] These analyses demonstrated that some lineage-specific TFs are commonly altered in epithelial and prostate cancers, respectively, and thus the accessibility of the binding sites of these TFs in cfDNA may also be exploited to identify tumor origin.

[0288] For tumor subclassification, we used P148 as index case, in which analysis was performed on two plasma samples (P148_1, P148_3) obtained 12 months apart, during which the prostate adenocarcinoma transdifferentiated into t-SCNC. In these two samples, we observed significant TFBS accessibility changes (Kendall's Tau: 0.7573) specifically reflected by various TFs. t-SCNC is not at this time an androgen-dependent stage of prostate cancer, and as a consequence, AR binding site accessibility was not observed any more in sample P148_3 (Figure 4E). There was a concomitant decrease in FOXA1 accessibility, due to its close collaboration with nuclear hormone receptors (Figure 4E). Furthermore, changes in cell type identity were manifested as a decrease in the accessibility of the binding sites of the prostate-specific lineage TFs HOXB13 and NKX3-1 (Figure 4E), as well as the epithelial TF GRHL2 (Figure 10A). TF changes associated with neuronal development include the expanded accessibility of GLI-similar 1 (GLIS1) (Figure ​(B)10), and expression of TFs dramatically increases under hypoxic conditions. Hypoxia has been argued to promote the progression of prostate adenocarcinoma to an androgen-dependent state and further downregulates the repressor element-1 (RE-1) silencing transcription factor (REST), which induces neuroendocrine reprogramming, and indeed we observed a significant decrease in REST accessibility (Figure4E). Furthermore, N-MYC is involved in AR signaling suppression and neuroendocrine program regulation, which was mirrored in the increased accessibility (Figure4E). These observations led us to acknowledge that TFBS may have high plasticity to affect pathways at certain cancer disease stages.

[0289] To demonstrate that prostate cancer subclassification based on cfDNA TFBS is possible, we further analyzed plasma samples from four t-SCNC cases (P170_2, P179_4, P198_5, P240_1). In these cases, we observed that this approach is also applicable to cfDNA with small coverage by downregulating P148_1 (819,607,690 reads) and P148_3 (768,763,081 reads) to approximately 50 million reads. The reduction in reads led to an increase in noise values. This depends on the number of TFBS but is insignificant for TFs with more than 1,000 TFBS (Figure 11), and as a result, the analysis of the previously mentioned highly relevant TFs was not affected. We repeated the analysis on the four aforementioned samples, each sequenced with approximately 50 million reads, and again observed decreased accessibility for the TFs AR, FOXA1, HOX-B13, and NKX3-1, or increased accessibility for N-MYC (Figure 4F). Reduced accessibility of REST was only observed in two of these four cases (P170_2 and P198_5; Figure 4F), consistent with the REST downregulation typically observed in 50% of neuroendocrine prostate cancer cases. Accessibility of GLIS1 was again increased in only these two cases (z-score: P170_2: 4.3; P198_5: 4.4), demonstrating that this hypoxia-related TF may be associated with REST downregulation.

[0290] Accessibility of TFBS may change during disease progression To address the question of whether TF accessibility remains stable over time, we collected and analyzed sequential samples from four patients (P40, P147, P148, C2). Analysis was restricted to 1,000 msTFBS and revealed no significant differences in three of the four plasma sample pairs (control: median: 0.8404 ± 0.0196 (IQR); P40: 0.8620; P147: 0.8370; C2: 0.8719; Kendall's Tau, respectively) (Figure 12).

[0291] Between P147_1 and P147_3, a novel high-amplitude amplification involving the RET gene developed, whereas in C2_7, the amplification involving KRAS observed in C2_6 was lost. As both RET in prostate cancer and KRAS in CRC may affect the PI3K / AKT / mTOR pathway, downstream targets such as the TF CREB were considered. However, their accessibility did not differ from control plasma samples and even remained unchanged. Between P40_1 and P40_2, tolerance to androgen deprivation therapy (ADT) developed. This was reflected in high-level amplification of the AR gene. However, if AR expanded its repertoire of transcriptional targets, this was not evident in the aforementioned T-ARBS and N-ARBS (Figure 18A-B). Using a conservative approach in this analysis, we observed changes only when the accessibility scores differed by ≥100 from one analysis to the next, which may explain the reduction or limitation of the differences between these samples.

[0292] A significant change occurred in TF accessibility for example P148 (Pearson correlation: 0.777291), where the tumor transdifferentiated from a prostate adenocarcinoma (P148_1) to a neuroendocrine tumor (P148_3). Neuroendocrine tumors are not now the androgen-dependent stage of prostate cancer, and as a result, AR binding site accessibility is no longer required and was therefore reflected in these analyses (Figure S16A-C). The change in cell type identity of this prostate cancer example is evident as accessibility to epithelial cell fate binding sites is determined to be lost for TFs GRHL2 and GRHL3 (Figure S16A-C). Additionally, a similar decrease in accessibility was observed for other prostate-specific lineage TFs HOXB13, NKX3-1, FOXA1, GATA2, and GATA3.

[0293] Importantly, as mentioned above, TF changes associated with neuronal development were also observed. Hypoxia is frequently observed in advanced solid tumors and may promote the progression of prostate adenocarcinoma to an androgen-independent state and induce a neuroendocrine program. Indeed, increased accessibility of GLI-similar 1 (GLIS1) was observed, and TF expression dramatically increased under hypoxic conditions. Furthermore, hypoxia downregulates suppressor element-1 (RE-1) silencing transcription factor (REST) ​​in prostate cancer epithelium and induces the expression of neuronal genes involved in neuroendocrine reprogramming. REST is a key mediator of neuroendocrine differentiation caused by androgen deprivation, and indeed decreased accessibility of REST (Figure 16A-C) was observed.

[0294] Differences related to stem cell features were also observed. The TFs SOX2 and SOX11 are upregulated during neuroendocrine transdifferentiation. An increased accessibility for these two TFs was observed. However, they preceded other changes, as they were already present in the plasma sample P148_1 (Figure 6E). This example demonstrates that other features of this real-time analysis, such as the order of events, can be established. A further stem cell-related change was the decreased accessibility of FOXA2 (Figure 16A-C).

[0295] Changes were also observed in less well-characterized TFs such as TF ZNF644 (ZNF644 is one of the core subunits in the G9a / GLP complex and mediates mono- and dimethylation of Lys9 of histone H3 at specific loci associated with transcriptional repression) or ZNF701 ( Fig. 16, A–C ), whose possible roles in transdifferentiation remain unclear.

[0296] Classification of prostate cancer based on cfDNA TFs This approach is applicable to samples sequenced at small coverage, as well as to plasma samples P148_1 (819,607,690 reads) and P148_3 (768,763,081 reads) downsampled to approximately 50 million reads. The reduction in reads resulted in an increase in noise values, which was dependent on the number of TFBSs and was insignificant for TFs with more than 1,000 TFBSs (Figure 18A-B). Thus, the accessibility analysis for the aforementioned highly relevant TFs involved in neuroendocrine tumor transdifferentiation was not affected.

[0297] [Table 1]

[0298] Consideration This study provides a methodology and bioinformatics software pipeline to infer tumor cell-specific transcription factor binding from cell-free DNA in blood, with relevance for clinical diagnosis and non-invasive tumor classification. While some studies employ a gene-centric focus when assessing somatically acquired changes, this analysis uses a significant portion of the non-coding genome and focuses on TFBS. Since many TFs bind preferentially in open chromatin and therefore must interact with nucleosomes, mostly mononucleosomal cfDNA is used, as it allows mapping of nucleosome positions. The unique feature of this approach is that it generates in vivo data on TFBS from endogenous, physiological processes, as opposed to technical variations associated with in vitro assays. Nevertheless, these data correlated robustly with DNase I hypersensitivity data on cell lines GM12878, LNCaP, or HCT116, demonstrating the reliability of the approach.

[0299] In contrast to other analyses that may use a common tissue-specific pattern using a mixture of cfDNA signals resulting from multiple cell types and Fourier transform analysis, the disclosed method and system may characterize individual TFs to establish lineage-specific TFs for clinical applications. The improved resolution of TFBS analysis may allow monitoring of TFBS accessibility from cfDNA, which may reveal plasticity during disease course such as reprogramming to different cell lineages. Furthermore, while other analyses may require more than 1.5 billion reads per sample, this test allows thorough TF analysis with approximately 50 million reads, making this approach more efficient and cost-effective for clinical applications. Importantly, this cfDNA TFBS bioinformatics pipeline enables tumor classification, thus filling a major diagnostic gap in the management of patients with prostate cancer, for example.

[0300] This work offers several substantial improvements over current techniques for TF profiling. First, using a curated list of TFBS from GTRD annotated with cfDNA, a list of high-confidence TFs, 676 TFs are amenable to analysis from cfDNA. Second, using this bioinformatics pipeline, improved metrics and accessibility scores are established to allow comparison of TFBS accessibility between various cfDNA samples. Third, the use of z-score statistics based on comparison of control samples (e.g., reference samples) to case samples (e.g., test samples from subjects) allows identification of substantial changes in TFBS accessibility. Finally, the use of lineage-specific TFs for blood lineages (PU.1, LYL1, and SPIB), epithelial lineages (GRHL2), and prostate lineages (AR, HOXB13, NKX3-1) is demonstrated for clinical use with cfDNA. This is in contrast to other methods that require a mixture of signals resulting from multiple cell types contributing to cfDNA and analysis by Fourier transformation. Current evaluation of clinical utility recognizes that these TFs are broadly applicable to identify individuals with epithelial or prostate cancer. The ability to monitor TFBS accessibility over time has been demonstrated, and dramatic changes involving the AR lineage, epithelial lineage, prostate lineage, and neural lineage, especially during the transdifferentiation of prostate adenocarcinoma to neuroendocrine tumors, can be assessed noninvasively from peripheral blood.

[0301] Maps of TF-nucleosome interactions may be heterogeneous and contain signals from all cell types that give rise to cfDNA. Plasma samples from individuals likely to have a high ctDNA load may be used, which may affect the sensitivity of the measurement. Furthermore, this approach uses whole genome sequencing with relatively high coverage (approximately 50 million reads), which is more than shallow sequencing plasma approaches for the establishment of SCNAs.

[0302] Nevertheless, advanced prostate cancer, the tumor entity analyzed herein, is a prime example of the intractability and resulting lethality that characterize metastatic cancer. Clinical biopsies of metastatic lesions are not routinely performed, and therefore detailed information of the molecular mechanisms regulating prostate cancer cell survival and progression is missing. Indeed, tumor studies lack kinetic models to investigate the transitions and interactions between pathways, especially dynamic profiling of clinical samples. As TFs have the potential to regulate gene transcription across the genome, and in their frequent and elaborate lineage-specific manner, their detailed analysis offers unique opportunities to improve clinical diagnosis. The data may further provide a rationale for detailed scrutiny of the non-coding genome through improved approaches for transcriptional regulatory profiling.

[0303] method Subject The study was approved by the Ethics Committee of the Medical University of Graz (approval numbers 21-227 ex 09 / 10 [breast cancer], 21-228 ex 09 / 10 [prostate cancer], 21-229 ex 09 / 10 [colon cancer], 29-272 ex 16 / 17 [high-resolution analysis of plasma DNA]) and was performed in accordance with the Declaration of Helsinki, and written informed consent was obtained from all patients and healthy probands, respectively. Some plasma samples, e.g. samples from patients B7 and B13, P40, P147, and P148, were analyzed in other studies.

[0304] B7 and B13: In these studies, matched primary tumors from two metastatic breast cancer cases (B7, B13) were synchronously obtained, as well as plasma DNA, analyzed by whole-genome sequencing and RNA-Seq. Plasma DNA was sequenced at high coverage (B7: ~411 million reads; ~8.2x; B13: ~455 million reads; ~9.1x) and copy number changes were calculated. We identified focal amplifications common in breast cancer, such as amplifications at 11q13.3 (15 genes including CCND1) in B7, or at 8p11 (31 genes including FGFR1) and 17q12 (46 genes including ERBB2) in B13.

[0305] P40: Initial plasma DNA analysis for patient P40 revealed multiple copy number alterations in most of the autosomes, but none were observed in the X chromosome (Figure 3B). Prior to this treatment, the patient was treated with localized radiation. As the disease progressed, treatment was switched to degarelix, a third-generation LHRH antagonist. However, despite this change in treatment, progression was still evident 10 months later, and repeat plasma analysis revealed the same alterations in the autosomes, but a focal amplification in chromosome Xq12, which harbors the AR gene.

[0306] B7 and B13: In these studies, matched primary tumors from two metastatic breast cancer cases (B7, B13) obtained in synchrony plus plasma DNA were analyzed by whole genome sequencing and RNA-Seq. Plasma DNA was sequenced at high coverage (B7: ~411 million reads; ~8.2x; B13: ~455 million reads; ~9.1x) and copy number alterations were calculated. Focal amplifications common in breast cancer, such as amplifications at 11q13.3 (15 genes including CCND1) in B7 or at 8p11 (31 genes including FGFR1) and 17q12 (46 genes including ERBB2) in B13, were identified as previously defined (Ulz et al., 2016b).

[0307] P21: Patient P21 was diagnosed with metastatic castration-resistant prostate cancer (CRPC). Four months after treatment with the LHRH antagonist degarelix, the patient showed signs of clinical progression followed by an increase in PSA levels. At the time of progression (P21_2), the cfDNA profile was observed with a tandem overlap phenotype.

[0308] P40: At the time of diagnosis, patient P40 was classified as having castration sensitive prostate cancer (CSPC). As the patient had not shown a clinical response to prior radiation therapy, treatment was switched to degarelix, an LHRH antagonist. Initially, the patient showed a good response to androgen deprivation (PSA decreased from 425.3ng / mL to 115.3ng / mL) but progressed to CRPC (PSA: 656.0ng / mL) after 10 months. The patient's cfDNA profile at the time of progression (P40_2) revealed high-level AR amplification on chromosome X.

[0309] P40: Initial plasma DNA analysis for patient P40 revealed multiple copy number alterations in most of the autosomes, but none were observed in the X chromosome (Figure 4B). Prior to this treatment, the patient was treated with localized radiation. Due to disease progression, the patient's treatment was switched to degarelix (Rick et al., 2013), a third-generation LHRH antagonist. However, despite this switch in treatment, progression was still evident 10 months later, and repeat plasma analysis revealed the same alterations in the autosomes, but a focal amplification in chromosome Xq12, which harbors the AR gene.

[0310] P111: The first sample, P111_1, was obtained at the time of diagnosis of prostate cancer. The patient already had multiple malignant lesions in his bones, lymph nodes, and kidneys. First-line treatment was the GnRH analogue goserelin, followed by radiation therapy. Between the two samples, P111_1 and P111_4, the patient responded well to treatment (CSPC, PSA dropped to 15.5 ng / mL). Two months before P111_4 sampling, clinical progression was evident and the patient progressed to CRPC. A cfDNA sample from the time of progression (P111_4) was analyzed to observe partial AR amplification. Furthermore, the patient was administered chemotherapy (docetaxel), but no further response was observed.

[0311] P143: Six years prior to collection of sample P143_3, patient P143 had been diagnosed with metastatic prostate cancer. The patient had previously been treated with various antiandrogens, including a second generation antiandrogen (abiraterone). Thus, at the time of collection of sample P143_3, the patient was multi-treated. Due to disease progression and subsequent failure of multiple treatments with various ADTs, chemotherapy was introduced (microtubule inhibitor - cabazitaxel).

[0312] P147: The first blood sample (P147_1) was obtained 5 years after diagnosis. The patient presented with multiple bone metastases and was therefore considered to have CRPC. During these 5 years, the patient had received radiation therapy and multiple antiandrogens. At the time of P147_1 sample collection, a new PSA increase was noted. Analysis of cfDNA revealed high-level amplifications at Xq12(AR) and chromosome 5q14.3. Six months after chemotherapy (docetaxel) and antiandrogen therapy (abiraterone, enzalutamide), a new cfDNA sample (P147_3) was analyzed. This analysis revealed a novel RET amplification at chromosome 10. The development of new focal events and increasing ctDNA content (P147_1 ichorCNA: 52%; 147_3 ichorCNA: 73%) correlated with clinical progression.

[0313] In patient P147, the time between prostatectomy and the first plasma sample was 56 months. 20 months after surgery, the PSA level was significantly elevated and radiation therapy was initiated. 28 months after diagnosis, the PSA level rose again. The patient was treated with the nonsteroidal antiandrogen bicalutamide for 13 months, followed by an additional 4 months of the GnRH analogue leuprorelin, and finally the monoclonal antibody denosumab, given the presence of bone metastases. Plasma samples were taken at the time of the PSA rise and novel high-level amplifications were detected at Xq12(AR) and chromosome 5q14.3. An additional high-level focal amplification occurred at chromosome 10q11.21. This chromosome occurred during the collection of the first and second plasma samples (e.g. P147_1 and P147_2). The time between these two samples was 6 months. During this period, the patient was treated with chemotherapy, e.g. docetaxel.

[0314] P148: P148 was diagnosed with prostate adenocarcinoma. Patient was placed on ADT for a period prior to first blood draw (P148_1). Clinician reported progressive disease (PSA: 694.4) with new bone and lymph node involvement. Chemotherapy (docetaxel) was introduced at the time of P148_1 sample collection. Multiple focal events were identified (MYC amplification; PTEN loss; FOXP1, RYBP, SHQ1 loss; TMPRSS2-ERG fusion) including AR amplification (patient previously considered to have CRPC).

[0315] Six months after the first sample collection, sample P147_3 was obtained. During this period, extensive progression with multiple liver and bone metastases was evident, with a PSA value of 52.0 ng / mL and an NSE value of >370 ng / mL. Interestingly, AR amplification was not detected in sample P147_3. This sample is characterized by transdifferentiation from adenocarcinoma to neuroendocrine prostate cancer (as described in Ulz ​​et al. 2016,Belic et al. 2018). After a short response to palliative treatment with carboplatin and etoposide, disease progression was evident and the patient died two months later.

[0316] Patient P148 was diagnosed with prostate adenocarcinoma. The first plasma sample was obtained 16 months after the initial diagnosis, at which point the patient presented with clearly progressive disease with increasing bone metastases and newly diagnosed lymphadenopathy. Due to progressive disease (PSA: 694.41ng / mL), the patient was treated with docetaxel for 7 months. A second plasma DNA analysis during this period confirmed the presence of high AR amplification. Five months after the last docetaxel treatment, extensive progression was evident with multiple liver and bone metastases, PSA 52.0ng / mL, NSE >370ng / mL. The patient was treated with palliative care with carboplatin, with an initial partial response lasting 3 months. The patient's disease then progressed and he died 2 months later.

[0317] P166: A blood sample from patient P166 was taken 2 years after the initial diagnosis of metastatic prostate cancer. The patient was treated with the antiandrogen bicalutamide but developed progressive disease (CRPC). AR amplification on chromosome X was identified in sample P166_1 as a sign of progression. As the patient progressed to CRPC, chemotherapy (docetaxel) was further introduced. After 3 months of chemotherapy, the patient experienced a partial response.

[0318] P190: Five years prior to the blood draw, the patient had been diagnosed with localized prostate cancer. Two years later, the disease had progressed locally, and over the next two years, the patient developed metastatic disease, mainly bone metastases. During these years, the patient had been treated with the antiandrogen bicalutamide. The patient's disease progressed with bone metastases, some of which were characterized by a neuroendocrine phenotype. The clinician therefore treated the patient with carboplatin-based chemotherapy (carboplatin / etoposide). cfDNA analysis was performed 3 months after the initiation of carboplatin / etoposide treatment. At the time of this analysis, the patient had a good response, with a PSA response ≥ 50% and normal NSE values.

[0319] Tandem overlap phenotype: P21_2: 59,849,368 leads P111_1: 58,258,680 leads P111_4: 61,085,342 leads P166_1: 52,829,575 leads Chromosipsis (chromosome 10): P143_3: 111,958,416 reads (at least PC-specific changes)

[0320] Pairwise comparison of plasma samples To address whether TF accessibility remained stable over time, we analyzed two samples each from patients P40, P147, and C2. However, the criteria were very stringent, e.g., by limiting the analysis to 1,000-ms TFBS, and no significant differences were found between these pairs of plasma samples (control: median: 0.8404 ± 0.0196 (IQR); P40: 0.8620; P147: 0.8370; C2: 0.8719; Kendall's Tau, respectively) (Figure 12).

[0321] Between samples P147_1 and P147_3 from patient P147, a novel high-amplitude amplification involving the RET gene developed, whereas in C2_7, the amplification involving KRAS observed in the previous sample C2_6 was lost. As both RET in prostate cancer and KRAS in CRC can affect the PI3K / AKT / mTOR pathway, downstream targets such as the TF CREB were considered. However, its accessibility did not differ from the control plasma samples and even remained unchanged. Between samples P40_1 and P40_2 from patient P40, tolerance to androgen suppression therapy (ADT) developed. This was reflected in a high-level amplification of the AR gene. However, if AR expanded its repertoire of transcriptional targets, this was not evident in the aforementioned T-ARBS and N-ARBS (Figure 19). A very conservative approach limiting the analysis to 1,000-ms TFBS could explain why no differences were observed between these samples.

[0322] TFBS Shape We investigated the coverage profiles of TF-specific nucleosomes because we observed evenly spaced nucleosome peaks in some TFs containing binding sites (e.g., PU.1 and GRHL2 in Fig. 2D), whereas other TFs showed broad troughs in their binding sites (e.g., CREM in Fig. 6A) similar to those observed in their TSSs. In total, 55 TFBSs were identified, among which TFBSs were >300bp, of which 26 had binding sites close to dinucleosome size (312-352bp; Fig. 8D). These patterns were correlated with the CpG islands (p=4.2×10) for TFBSs >300bp in size. -11 ; Mann-Whitney U test) and TSS (p = 8.5 × 10 -12 We confirmed a highly significant increase in overlap for both the 1444 and 1445 mice (Figure 8E).

[0323] CTCF as a special case for TFs with multiple distinct binding sites We used CCTC-binding factor (CTCF) to investigate various TFBSs of the same TF. CTCF is present in 55,000-65,000 binding sites in mammalian genomes. Approximately 5,000 of these sites are ultraconserved, with approximately 50% in intergenic regions, 15% near promoters, and 40% intragenic. Furthermore, chromosomes are divided into multiple evolutionarily conserved higher-order chromosomal structures called topologically associated domains (TADs), whose boundaries are enriched for binding sites for CTCF and cohesin. In mammals, 15% of genomic CTCF binding sites are present at TAD boundaries, while the remaining 85% are within TADs.

[0324] Blood sampling and library preparation Peripheral blood was collected from patients with metastatic prostate, breast, or colon cancer at the Department of Hematology of the Medical University of Graz, and from anonymous healthy donors without known chronic or malignant diseases at the Department of Medical Oncology. CfDNA was isolated from plasma using the QIAamp Circulating Nucleic Acids kit (QIAGEN, Hilden, Germany) according to the manufacturer's protocol. Library preparation for WGS was performed as previously described (Heitzer et al., 2013).

[0325] Sequencing Control and high-coverage tumor samples were sequenced on 2 × 150 bp Illumina NovaSeq S4 flow cells by the Biomedical Sequencing Facility at CeMM, Vienna, Austria. For control samples, we obtained an average of 435,135,450 (range: 352,904,231–556,303,420) paired-end reads. For tumor samples (P40_1, P40_2, P147_1, P147_3, P148_1, P148_3, C2_6, C2_7), we sequenced an average of 688,482,253 reads (range: 541,216,395–870,285,698). Additional samples were sequenced using the Illumina NextSeq platform (B7_1, B13_1, P190_3; average sequencing yield: 296,733,931 reads; range: 181,953,656 to 379,733,061) and HiSeq platforms (P21_2, P111_1, P111_4, P143_3, P166_1; average sequencing yield: 52,869,911 reads; range: 41,780,819 to 84,049,593), respectively.

[0326] The low-coverage tumor samples used to generate the single-entity collection were sequenced on either the Illumina Next-Seq or MiSeq platforms, resulting in 382,306,130 reads from 69 prostate cancer samples, 254,490,128 reads from 60 breast cancer samples, and 604,080,473 reads from 100 colorectal cancer samples.

[0327] Plasma sample characterization Some plasma samples, such as those from patients B7 and B13, P40, P147, and P148, were analyzed, which included information on mutations, specific SCNAs, and tumor content of the plasma samples based on the algorithm ichorDNA.

[0328] The ETS family of oncogenic transcription factors (proposed by (Sizemore et al., 2017)) Approximately 50% of localized and 40% of metastatic prostate cancers contain TMPRSS2-ETS fusions. Recurrent gene fusions of the androgen-regulated 5' untranslated region of TMPRSS2 to ERG (TMPRSS2-ERG gene fusion) found in approximately 50% of primary prostate cancers result in hijacking of ETS expression and the transcriptional program by AR.

[0329] ERG has also been shown to block prostate neuroendocrine cell differentiation.One possible mechanism for TMPRSS2-ERG-mediated maintenance of prostate stem and progenitor cells is via SOX9.

[0330] Given their role as transcription factors, it is not surprising that ETS factors mediate tumorigenesis through multiple mechanisms ranging from basic survival cues to completing epigenetic reprogramming. ETS factors further influence nucleotide, energy, and steroid metabolism.

[0331] P40: Mutation in BRCA1: NM_007294:Q975R; specific SCNA: TMPRSS2-ERG fusion; AR amplification in sample 2; chr12 amplification (containing ARID2, HDAC7); tumor content: P40_1: 30%, P40_2: 24%. Additional focal amplification on chromosome 15 (containing SNORD (small nucleolar RNA, C / D box) genes, 16 (2x), and 19 (BRD4); P40_1 ichorCNA: 30%; P40_2 ichorCNA: 24%.

[0332] P147: Mutations: BRCA2: T298fs; TP53: F338I; Specific SCNAs: RET amplification in sample 3; AR amplification; BRAF amplification (7q34); PTEN loss; Tumor content: P147_1: 52%; P147_3: 73%. Additional focal amplifications on chromosomes 5 (XRCC4) and 21 (RBM11); P147_1 ichorCNA: 52%; 147_3 ichorCNA: 73%.

[0333] P148: Mutations: TP53:R213X; specific SCNAs: MYC amplification; PTEN loss; FOXP1, RYBP, SHQ1 loss; TMPRSS2-ERG fusion; AR amplification (gone in P148_3); tumor content: P148_1: 38%; 148_3: 49%.

[0334] C2: Specific SCNA: high amplification on chromosome 12 (KRAS) in C2_6, not visible in C2_7; tumor content: C2_6: 18%; C2_7: 28%.

[0335] Defining transcription factor binding sites Data from the GTRD database were downloaded and individual BED files were extracted for each TF. Positions were calculated again by focusing on the points reported to have the highest ChIP-seq signal in the metacluster. An additional BED file was created containing only peaks supported by more than 50% of the maximum number of samples analyzed for this specific transcription factor. All BED files were then converted to hg19 (from the original hg38) using the liftOver tool provided by UCSC.

[0336] Overlapping transcription factor binding sites To check whether transcription factor binding sites overlap, the regions of binding sites from GTRD (those supported by >50% of samples) were expanded by 25bp, 50bp, and 100bp on either side using bedtools slop, respectively. The number of overlaps was then calculated using bedtools intersect via pybedtool for all transcription factors and other transcription factors.

[0337] 2. Preparation of Single-end Sequencing Data To enhance the nucleosome signal, the sequencing reads were trimmed to remove the part of the sequencing read related to the linker region. Thus, the forward sequencing reads were trimmed to include only bases 53-113 (which may correspond to the central 60 bp of the 166 bp fragment). The reads were then aligned to the human hg19 genome using the Burrows-Wheeler aligner (bwa) and PCR overlaps were removed using samtools rmdup. The average coverage was calculated by bedtools genomecov.

[0338] 2. Preparation of Paired-end Sequencing Data Paired-end sequencing reads were aligned to the human hg19 genome using bwa mem and PCR overlaps were marked with picard MarkDuplicates.

[0339] Preparation of MNase-seq data The BAM files of the GM12878 MNase-seq experiment were downloaded from the ENCODE portal. Sequencing reads in the BAM files were trimmed directly from the BAM files using pysam. Briefly, the leftmost alignment position in the BAM files was shifted by 53 bp in each direction, and the sequence length was adjusted to 60 bp. Coverage patterns were then calculated in the same way as for the trimmed cfDNA sequencing data.

[0340] Coverage patterns of transcription factor binding sites The coverage pattern was calculated for all transcription factors in the GTRD. For this purpose, coverage data were extracted for the entire region using pysam count_coverage in the region ±1000 bp around the defined binding site. Coverage data at all sites were normalized by the local copy number variation and the average coverage. Coverage was averaged for all positions around the TFBS and 95% confidence intervals were calculated. When more than 100,000 positions were defined for a transcription factor, 100,000 sites were randomly selected and analyzed.

[0341] Insert size around transcription factor binding sites To assess whether fragment sizes around transcription factor binding sites were biased, insert size data from paired-end analysis was used. All positions from -1000bp to 1000bp of the binding site were traversed and pysam was used to fetch (single-end) sequencing reads with the central 3bp around the midpoint at this position. Additionally, paired-end alignments of the same samples were fetched and insert size information was presented for each read. Then, all insert sizes at a particular position relative to the TFBS were summarized and 1000 data points were sampled and plotted for each position ranging from -1000bp to 1000bp of the TFBS.

[0342] Measurement of transcription factor binding site size To measure the size of transcription factor binding sites, each coverage pattern was smoothed using a third-order Savitzky-Golay filter (window size: 31). Peaks were identified by examining data points with more than 20 adjacent data points on either side. Peaks were removed if they were within 50 bp of the center of the hypothetical binding site. The distance between the peaks closest to the binding site peak was identified as the transcription factor binding site size.

[0343] Since evaluation of binding sites is only reasonable if nucleosome synchronization is detectable, signals were filtered by various criteria: High frequency signal amplitude exceeds 0.1. The mean normal coverage of the central 100 bp is less than 1. The number of peaks is less than 15. The median distance between peaks is greater than 150bp. -The binding site setting includes more than 500 sites.

[0344] A total of 228 binding site configurations passed these filters and were used for binding site prediction.

[0345] Measuring transcription factor accessibility using Savitzky-Golay filters Two distinct signals constituted the coverage pattern, and two signals of different frequencies were extracted into low and high range frequencies. Low range frequency data was extracted by a Savitzky-Golay filter (cubic polynomial, window size 1001). High range frequency was extracted by another Savitzky-Golay filter (cubic polynomial, window size 51). The high frequency signals were then normalized by dividing them by the result of the low frequency signals. The data range of the high frequency signals was then recorded. Because the coverage profiles of transcription factors with few described binding sites are noisy in nature, LOESS (locally weighted smoothing) was performed on the signal range and amount of described binding sites. The range values ​​were corrected by the smoothed LOESS, and the adjusted range ranks were calculated.

[0346] Measuring transcription factor accessibility using wavelet transform As an additional method to measure transcription factor accessibility, we applied a wavelet transform using the R package "WaveletComp". For all signals, peaks were recorded in the power spectrum along a period from 2 bp to 512 bp. The highest peak in the range of 135 bp to 235 bp (185 bp ± 50 bp) was used to reconstruct a denoised high-frequency nucleosome signal at a specific period. In addition, the original series was detrended to remove residual baselines. Three parameters of the reconstructed signal were analyzed: maximum signal amplitude, sum of signal dynamics (amplitude squared), and sum of absolute amplitude along 2000 bp surrounding the transcription factor binding site.

[0347] For comparison between tumor and normal samples, the mean and standard deviation for each parameter were recorded in the normal samples for all transcription factors, and a Z-score was calculated by taking each parameter in the cancer samples, subtracting the mean value of the normal samples, and dividing by the standard deviation.

[0348] Comparison of tumor and control samples To compare tumor and control samples, the rank of each transcription factor at the adjusted range value was compared. The rank difference between tumor samples and all control samples was calculated and the average rank difference was recorded. In addition, a z-score was calculated from the accessibility rank for all transcription factors by taking each rank, subtracting the average rank of the control samples, and dividing by the standard deviation of the transcription factor rank for the control samples.

[0349] DNase hypersensitivity data analysis BAM files from DNase hypersensitivity experiments were downloaded from the ENCODE database for GM12878, LNCaP, and HCT116 cell lines. Transcription factor binding site regions were augmented by 25bp on both sides using bedtools slop. Coverage at each binding site was extracted using mosdepth and normalized to 1 million mapped reads per sample.

[0350] Analysis of somatic copy number alterations (SCNAs) For the control data, paired-end alignments were subsampled using samtools view to include only 2% of the initial alignments and converted to FastQ using samtools fastq. Separate low-coverage whole-genome sequencing was performed for the cancer samples. Plasma-Seq was applied to the subsampled FastQ files and the low-coverage data of the cancer samples, respectively. Briefly, sequencing reads were aligned to the human hg19 genome and counted in pre-specified bins. The size of the bins was determined by the size of theoretically mappable positions, which is the main driver of mappability variance across the genome. The number of reads was normalized to the total amount of reads and the GC content of the bins was corrected by LOESS smoothing on the GC spectrum. The corrected number of reads was further normalized by the average number of reads of non-cancer controls per bin to control for additional position mutations.

[0351] Accessibility scores enable accurate inference of TF binding from cfDNA No TFs were found above the z-score threshold of ±5 in samples from healthy donors (Figure 29A). However, we observed a very different pattern in samples from cancer patients. For example, in prostate sample P40_1 from patient P40, TFs with accessibility above the z-score threshold of +5 included the prostate lineage-specific TFs HOXB13, AR, and NKX3-1 in addition to GRHL2 and FOXA1, which cooperate with nuclear hormone receptors in endocrine-derived tumors of prostate and breast cancer (Figure 29B). In contrast, the accessibility of hematopoietic TFs such as LYL1, SPIB, and EVI1 (transcription regulator ectopic viral insertion site 1 (Figure 29B)) was low. These results were in high agreement with the TF ranking based on ATAC-seq data. In breast cancer samples B7 and B13, we detected increased accessibility, consistent with ATAC-seq data for GRHL2, FOXA1, and ZNF121, zinc finger proteins involved in regulating cell proliferation and breast cancer progression.

[0352] In a sample from colon cancer patient C2, we unexpectedly observed that the ATAC-seq data had EVX2 ranked as the most accessible in COAD, a TF that is not strongly associated with cancer. Indeed, we ranked EVX2 with the highest accessibility in this analysis (Figure 29C), and observed a large increase in the accessibility of EVX2 in the nucleosome position map (Figure 29D). Consistent with the ATAC-seq data, we also observed increased accessibility for the TFs HNF4A, GRHL2, DLX2, HNF4G, and HNF1A (Figure 29D).

[0353] Furthermore, as predicted by ATAC-seq data evaluation, accessibility for hematopoietic drug-related TFs, such as LYL1, TAL1 (SCL / TAL1 (Stem Cell Leukemia / T-cell Acute Lymphoblastic Leukemia [T-ALL]1), EVI1, TBX21 (T-bet), and PU.1, was decreased in all tumor samples (Figure 29B-C). To further confirm the robustness and reproducibility of lineage-specific TFs in cfDNA, we observed that in a collection of multiple cfDNA samples generated at shallow coverage (<0.2x), the accessibility scores of TFs with increased accessibility in the majority of samples were elevated, whereas others may be averaged out. To this end, cfDNA samples were collected separately for prostate cancer cases (n=69), colorectal cancer cases (n=100), and breast cancer cases (n=60). When the analysis was repeated, the accessibility scores of epithelial TFs were significantly increased in the majority of samples. GRHL2 and hematopoietic TFs recapitulated gains and losses in accessibility patterns across the three epithelial lineages, respectively. Accessibility of the TFs EVX2, DLX2, HNF1A, HNF4A, and HNF4G in colon cfDNA pools, as well as the TFs AR and HOXB13 in prostate cancer cfDNA pools, was increased, whereas FOXA1 exceeded a z-score threshold of >5 in both prostate and breast, confirming that TF accessibility estimates derived from ATAC-seq data can be reliably inferred from plasma DNA nucleosome mapping.

[0354] Figure 29E provides a bar graph of the global Z-score plot of the combined breast, prostate, and colon cancer pools. The left panel displays TFs with increased accessibility in at least one tumor entity, and the right panel summarizes the accessibility of hematopoietic-related TFs.

[0355] Figure 30A and B provide graphs showing TF-based plasma separation limits and early cancer detection. Figure 30A is a graph showing a comparison of the accessibility of selected TFs in subsamples of the COAD cohort based on their tumor fraction. Figure 30B is a graph showing logistic regression with all 504 TFs for samples from stage I (left panel) and stage II (right panel) colon cancer cohorts, respectively. All results presented are cross-validated test set values.

[0356] While certain examples of methods and systems have been shown and described herein, those skilled in the art will recognize that these are merely examples and are not intended as limitations within this specification. Numerous variations, changes, and substitutions will occur to those skilled in the art without departing from the scope described herein. Furthermore, it should be understood that all aspects of the described methods and systems are not limited to the specific depictions, configurations, or relative proportions described herein, which depend on a variety of conditions and variables, and that the description is intended to include alternatives, modifications, variations, or equivalents.

Claims

1. 1. A computer-implemented method for detecting the presence or absence of a disease in a subject, the method comprising: (a) providing a set of sequence reads obtained from sequencing cell-free deoxyribonucleic acid (cfDNA) obtained or derived from the subject; (b) using the set of sequence reads to generate a read depth coverage pattern for a transcription factor, the read depth coverage pattern being at least in part determined by: i. aligning the set of sequence reads to a reference genome to provide a set of aligned sequence reads; ii. selecting, for each binding site of a plurality of binding sites of a transcription factor, genomic regions of the set of aligned sequence reads adjacent to the binding site to provide a plurality of adjacent genomic regions, wherein each binding site of the plurality of binding sites exhibits differential accessibility to a target enhancer of the transcription factor in a first population of subjects with the disease compared to a second population of subjects without the disease, and each of the plurality of binding sites is distinct; iii. using at least a portion of the set of aligned sequence reads to generate the read depth coverage pattern for each contiguous genomic region of the plurality of contiguous genomic regions, thereby providing a plurality of read depth coverage patterns of the plurality of binding sites of the transcription factor, respectively. A process for producing (c) processing the plurality of read depth coverage patterns to provide a signal, the signal characterizing binding site accessibility of the plurality of binding sites of the transcription factor; (d) splitting each of the plurality of read depth coverage patterns into a low frequency signal and a high frequency signal using a low pass filter, the high frequency signal being indicative of binding site accessibility for a given binding site; (e) determining an accessibility score for each of the plurality of binding sites for the transcription factor based at least in part on the radio frequency signal; and (f) detecting the presence or absence of the disease in the subject based at least in part on a deviation of the accessibility score from a reference accessibility score. A method comprising:

2. The method of claim 1 , wherein the disease is cancer.

3. 2. The method of claim 1, wherein the accessibility scores of at least 2, or at least 5, or at least 10, or at least 15, or at least 20, or at least 25 transcription factor binding sites are determined and input into a machine learning model to train a classifier capable of discriminating between healthy subjects and cancer patients, disease progressors and non-progressors, multiple disease subtypes, multiple disease stages, disease treatment responders and non-responders, or any combination thereof.

4. The method of any one of claims 1 to 3, wherein (d) comprises using a Savitzky-Golay filter.

5. The method of any one of claims 1 to 4, wherein the subject is a human.

6. The method according to any one of claims 1 to 5, wherein the transcription factor is an epithelial transcription factor.

7. The method according to any one of claims 1 to 6, wherein the transcription factor is a cancer-specific transcription factor.

8. 8. The method of claim 7, wherein the transcription factor is selected from the group consisting of GRH-L2, ASH-2, HOX-B13, EVX2, PU.1, Lyl-1, Spi-B, FOXA1, HNF-1A, HNF-4A, HNF-4G, and DLX-2.

9. The method of any one of claims 1 to 8, wherein (f) further comprises identifying a signal of higher accessibility of the transcription factor.

10. 10. The method of claim 1, further comprising detecting the presence or absence of the disease in the subject with an accuracy of at least about 70%.

11. 10. The method of claim 1, further comprising detecting the presence or absence of the disease in the subject with an accuracy of at least about 80%.

12. 10. The method of claim 1, further comprising detecting the presence or absence of the disease in the subject with an accuracy of at least about 90%.

13. 10. The method of claim 1, further comprising detecting the presence of the disease in the subject with a sensitivity of at least about 70%.

14. 10. The method of claim 1, further comprising detecting the presence of the disease in the subject with a sensitivity of at least about 80%.

15. 10. The method of claim 1, further comprising detecting the presence of the disease in the subject with a sensitivity of at least about 90%.

16. 10. The method of claim 1, further comprising detecting the absence of the disease in the subject with a specificity of at least about 70%.

17. 10. The method of claim 1, further comprising detecting the absence of the disease in the subject with a specificity of at least about 80%.

18. 10. The method of claim 1, further comprising detecting the absence of the disease in the subject with a specificity of at least about 90%.

19. 10. The method of claim 1, further comprising detecting the presence of the disease in the subject with a positive predictive value (PPV) of at least about 70%.

20. 10. The method of claim 1, further comprising detecting the presence of the disease in the subject with a positive predictive value (PPV) of at least about 80%.

21. 10. The method of claim 1, further comprising detecting the presence of the disease in the subject with a positive predictive value (PPV) of at least about 90%.

22. 10. The method of claim 1, further comprising detecting the absence of the disease in the subject with a negative predictive value (NPV) of at least about 70%.

23. 10. The method of claim 1, further comprising detecting the absence of the disease in the subject with a negative predictive value (NPV) of at least about 80%.

24. 10. The method of claim 1, further comprising detecting the absence of the disease in the subject with a negative predictive value (NPV) of at least about 90%.

25. 10. The method of claim 1, further comprising detecting the presence or absence of the disease in the subject with an area under the receiver operating characteristic (AUROC) of at least about 0.

70.

26. 10. The method of claim 1, further comprising detecting the presence or absence of the disease in the subject with an area under the receiver operating characteristic (AUROC) of at least about 0.

80.

27. 10. The method of claim 1, further comprising detecting the presence or absence of the disease in the subject with an area under the receiver operating characteristic (AUROC) of at least about 0.

90.

28. 10. The method of claim 1, further comprising applying a trained classifier to the signal to detect the presence or absence of the disease in the subject.

29. 2. The method of claim 1, further comprising applying a trained classifier to the accessibility scores of the binding sites of the transcription factors to detect the presence or absence of the disease in the subject.

30. 30. The method of claim 28 or 29, wherein the trained classifier comprises a trained machine learning classifier.

31. 31. The method of claim 30, wherein the trained machine learning classifier comprises a supervised machine learning algorithm.

32. 32. The method of claim 31, wherein the supervised machine learning algorithm comprises one or more of regression, support vector machines, tree-based methods, neural networks, and random forests.

33. The method of claim 1 , wherein (e) further comprises performing locally weighted smoothing on the high frequency signals.

34. The method of claim 1 , wherein the multiple binding sites for the transcription factor comprise at least 25 different binding sites for the transcription factor.

35. The method of claim 1, wherein the plurality of sequence reads comprises at least 50 million sequence reads.

36. 3. The method of claim 2, wherein the transcription factor is selected from the group consisting of GRHL2, FOXA1, and ZNF121, and the cancer comprises breast cancer.

37. 3. The method of claim 2, wherein the transcription factor is selected from the group consisting of EVX2, DLX2, HNF1A, HNF4A, GRHL2, and HNF4G, and the cancer comprises colon cancer.

38. 3. The method of claim 2, wherein the transcription factor is selected from the group consisting of LYL1 and PU.1, and the cancer comprises colon cancer.

39. 3. The method of claim 2, wherein the transcription factor is selected from the group consisting of GRHL2, FOXA1, HOXB13, AR, and NKX3-1, and the cancer comprises prostate adenocarcinoma.

40. 3. The method of claim 2, wherein the transcription factor is selected from the group consisting of REST, GRHL2, FOXA1, HOXB13, AR, and NKX3-1, and the cancer comprises small cell neuroendocrine prostate cancer.

41. 3. The method of claim 2, wherein the transcription factor is selected from the group consisting of GLIS1, SOX2, and SOX11, and the cancer comprises prostate cancer.

42. The method of claim 4 , wherein the Savitzky-Golay filter comprises a third-order Savitzky-Golay filter.

43. 34. The method of claim 33, wherein the locally weighted smoothing comprises locally estimated scatterplot smoothing (LOESS).

44. 2. The method of claim 1, further comprising obtaining or deriving the cfDNA from a biological sample of the subject.

45. 45. The method of claim 44, wherein the biological sample is a blood sample, a plasma sample, a serum sample, a sweat sample, a urine sample, a saliva sample, a cell sample, a tissue sample, or a derivative thereof.

46. 46. ​​The method of claim 45, wherein the biological sample is a plasma sample.

47. 2. The method of claim 1, wherein the reference genome is GrCH38, GrCH37, hg19, or hg38.

48. 1. A non-transitory computer readable medium comprising machine executable code that, when executed by one or more computer processors, performs a method for detecting the presence or absence of a disease in a subject, the method comprising: The method comprises: (a) providing a set of sequence reads obtained from sequencing cell-free deoxyribonucleic acid (cfDNA) obtained or derived from the subject; (b) using the set of sequence reads to generate a read depth coverage pattern for a transcription factor, the read depth coverage pattern being at least in part determined by: i. aligning the set of sequence reads to a reference genome to provide a set of aligned sequence reads; ii. selecting, for each binding site of a plurality of binding sites of a transcription factor, genomic regions of the set of aligned sequence reads adjacent to the binding site to provide a plurality of adjacent genomic regions, wherein each binding site of the plurality of binding sites exhibits differential accessibility to a target enhancer of the transcription factor in a first population of subjects with the disease compared to a second population of subjects without the disease, and each of the plurality of binding sites is distinct; iii. using at least a portion of the set of aligned sequence reads to generate the read depth coverage pattern for each contiguous genomic region of the plurality of contiguous genomic regions, thereby providing a plurality of read depth coverage patterns of the plurality of binding sites of the transcription factor, respectively. A process for producing (c) processing the plurality of read depth coverage patterns to provide a signal, the signal characterizing binding site accessibility of the plurality of binding sites of the transcription factor; (d) splitting each of the plurality of read depth coverage patterns into a low frequency signal and a high frequency signal using a low pass filter, the high frequency signal being indicative of binding site accessibility for a given binding site; (e) determining an accessibility score for each of the plurality of binding sites for the transcription factor based at least in part on the radio frequency signal; and (f) detecting the presence or absence of the disease in the subject based at least in part on a deviation of the accessibility score from a reference accessibility score. A non-transitory computer readable medium comprising:

49. 1. A system for detecting the presence or absence of a disease in a subject, the system comprising: a database comprising a set of sequence reads obtained from sequencing cell-free deoxyribonucleic acid (cfDNA) obtained or derived from the subject; and one or more computer processors operably connected to a database, wherein said one or more computer processors individually or collectively: (a) using the set of sequence reads to generate a read depth coverage pattern for a transcription factor, the read depth coverage pattern being at least in part determined by: i. aligning the set of sequence reads to a reference genome to provide a set of aligned sequence reads; ii. selecting, for each binding site of a plurality of binding sites of a transcription factor, genomic regions of the set of aligned sequence reads adjacent to the binding site to provide a plurality of adjacent genomic regions, wherein each binding site of the plurality of binding sites exhibits differential accessibility to a target enhancer of the transcription factor in a first population of subjects with the disease compared to a second population of subjects without the disease, and each of the plurality of binding sites is distinct; iii. using at least a portion of the set of aligned sequence reads to generate the read depth coverage pattern for each contiguous genomic region of the plurality of contiguous genomic regions, thereby providing a plurality of read depth coverage patterns of the plurality of binding sites of the transcription factor, respectively. generated by, (b) processing the plurality of read depth coverage patterns to provide a signal, the signal characterizing binding site accessibility of the plurality of binding sites of the transcription factor; (c) splitting each of the plurality of read depth coverage patterns into a low frequency signal and a high frequency signal using a low pass filter, the high frequency signal being indicative of binding site accessibility for a given binding site; (d) determining an accessibility score for each of the plurality of binding sites for the transcription factor based at least in part on the radio frequency signal; and (e) detecting the presence or absence of the disease in the subject based at least in part on a deviation of the accessibility score from a reference accessibility score; The system is programmed to:

50. 1. A computer-implemented method for monitoring disease progression or regression in a subject, comprising: The method comprises: (a) providing a first set of sequence reads obtained from sequencing cell-free deoxyribonucleic acid (cfDNA) obtained or derived from the subject at a first time, and a second set of sequence reads obtained from sequencing cfDNA obtained or derived from the subject at a second time, said second time being later than the first time; (b) using the first set of sequence reads to generate a first read depth coverage pattern for a transcription factor, and using the second set of sequence reads to generate a second read depth coverage pattern for the transcription factor, wherein the first read depth coverage pattern and the second read depth coverage pattern are at least partially i. aligning the first and second sets of sequence reads to a reference genome to provide a first and second set of aligned sequence reads; ii. selecting genomic regions of the first and second sets of aligned sequence reads adjacent to each binding site of a transcription factor to provide a plurality of adjacent genomic regions, wherein each binding site of the plurality of binding sites exhibits differential accessibility to a target enhancer of the transcription factor in a first population of subjects with the disease compared to a second population of subjects without the disease, and each of the plurality of binding sites is different; iii. using at least a portion of the first set and the second set of aligned sequence reads to generate the first read depth coverage pattern and the second read depth coverage pattern for each adjacent genomic region of the plurality of adjacent genomic regions, thereby providing a plurality of read depth coverage patterns of the plurality of binding sites of the transcription factor, respectively. A process for producing (c) processing the first read depth coverage pattern to provide a first signal characterizing binding site accessibility of the plurality of binding sites of the transcription factor, and processing the second read depth coverage pattern to provide a second signal characterizing binding site accessibility of the plurality of binding sites of the transcription factor; (d) splitting each of the first and second read depth coverage patterns into low and high frequency signals using a low pass filter, the high frequency signals being indicative of binding site accessibility for a given binding site; (e) determining an accessibility score for each of the first plurality of binding sites and the second plurality of binding sites of the transcription factor based at least in part on the radio frequency signal; (f) processing the first signal with a reference signal, the first signal and the reference signal having different frequencies; (g) processing the second signal with the reference signal, the second signal and the reference signal having different frequencies; and (h) monitoring the progression or regression of the disease in the subject based on processing of the first signal and the second signal with the reference signal. A method comprising:

51. 51. The method of claim 50, wherein the disease is cancer.

52. 52. The method of claim 50 or 51, wherein (d) comprises using a low pass filter.

53. 51. The method of claim 50, wherein (d) includes using a Savitzky-Golay filter.

54. The method of any one of claims 50 to 53, wherein the subject is a human.

55. The method of any one of claims 50 to 54, wherein the transcription factor is a cancer-specific transcription factor.

56. 56. The method of claim 55, wherein the transcription factor is selected from the group consisting of GRH-L2, ASH-2, HOX-B13, EVX2, PU.1, Lyl-1, Spi-B, FOXA1, HNF-1a, HNF-4a, HNF-4g, and DLX-2.

57. 57. The method of any one of claims 50-56, further comprising adjusting a treatment regimen for said disease in said subject based on (h).

58. 1. A non-transitory computer readable medium comprising machine executable code that, when executed by one or more computer processors, performs a method for monitoring disease progression or regression in a subject, comprising: The method comprises: (a) providing a first set of sequence reads obtained from sequencing cell-free deoxyribonucleic acid (cfDNA) obtained or derived from the subject at a first time, and a second set of sequence reads obtained from sequencing cfDNA obtained or derived from the subject at a second time, said second time being later than the first time; (b) using the first set of sequence reads to generate a first read depth coverage pattern for a transcription factor, and using the second set of sequence reads to generate a second read depth coverage pattern for the transcription factor, wherein the first read depth coverage pattern and the second read depth coverage pattern are at least partially i. aligning the first and second sets of sequence reads to a reference genome to provide a first and second set of aligned sequence reads; ii. selecting genomic regions of the first and second sets of aligned sequence reads adjacent to each binding site of a transcription factor to provide a plurality of adjacent genomic regions, wherein each binding site of the plurality of binding sites exhibits differential accessibility to a target enhancer of the transcription factor in a first population of subjects with the disease compared to a second population of subjects without the disease, and each of the plurality of binding sites is different; iii. using at least a portion of the first set and the second set of aligned sequence reads to generate the first read depth coverage pattern and the second read depth coverage pattern for each adjacent genomic region of the plurality of adjacent genomic regions, thereby providing a plurality of read depth coverage patterns of the plurality of binding sites of the transcription factor, respectively. A process for producing (c) processing the first read depth coverage pattern to provide a first signal, and processing the second read depth coverage pattern to provide a second signal, wherein the first signal and the second signal characterize binding site accessibility of the multiple binding sites of the transcription factor; (d) splitting each of the first and second read depth coverage patterns into low and high frequency signals using a low pass filter, the high frequency signals being indicative of binding site accessibility for a given binding site; (e) determining an accessibility score for each of the plurality of binding sites for the transcription factor based at least in part on the radio frequency signal; (f) processing the first signal with a reference signal, the first signal and the reference signal having different frequencies; (g) processing the second signal with the reference signal, the second signal and the reference signal having different frequencies; and (h) monitoring the progression or regression of the disease in the subject based on processing the first signal and the second signal with the reference signal.

Images