Machine learning classification model for cancer detection

Abstract

In implementations described herein, sequence representations are identified having at least at threshold likelihood of corresponding to a nucleic acid molecule produced by a subject in which a tumor-related biological condition is present. The identified sequence representations can be provided to a machine learning classification model to determine an indication of the tumor-related biological condition being present in subjects.

Description

MACHINE LEARNING CLASSIFICATION MODEL FOR CANCER DETECTION CROSS-REFERENCE TO RELATED APPLICATIONS

[0001] The present application incorporates by reference in its entirety for all purposes, the patent application number 63 / 652,359 filed May 28, 2024. BACKGROUND

[0002] Cancer is a major cause of disease worldwide. Each year, tens of millions of people are diagnosed with cancer around the world, and more than half eventually die from it. In many countries, cancer ranks the second most common cause of death following cardiovascular diseases. Early detection is associated with improved outcomes for many cancers.

[0003] Cancer can be caused by the accumulation of genetic variations within an individual's normal cells, at least some of which result in improperly regulated cell division. Such variations commonly include copy number variations (CNVs), single nucleotide variations (SNVs), gene fusions, insertions and / or deletions (indels), epigenetic variations including 5-methylation of cytosine (5-methylcytosine), and association of DNA with chromatin and transcription factors.

[0004] Cancers are often detected by biopsies of tumors followed by analysis of cell markers or DNA extracted from cells. But more recently it has been proposed that cancers can also be detected from cell-free nucleic acids in body fluids, such as blood or urine. Such tests have the advantage that they are noninvasive and can be performed without identifying suspected cancer cells in biopsy. However, such tests are complicated by the fact that the amount of nucleic acids in body fluids is very low and that the nucleic acids that are present are heterogeneous in form (e.g., RNA and DNA, single-stranded and double-stranded, and various states of post-replication modification and association with proteins, such as histones).

[0005] Thus, there is a need for improved systems and methods for improved cancer detection using liquid biopsy assays. Therefore, it is an object of the disclosure to provide computer- implemented systems and methods and other processes that have improved capability to classify a sample as containing tumor-derived DNA. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate certain implementations, and together with the written description, serve to explain certain principles of the methods, computer readable media, and systems disclosed herein. The description provided herein is better understood when read in conjunction with the accompanying drawings which are included by way of example and not by way of limitation. Itwill be understood that like reference numerals identify like components throughout the drawings, unless the context indicates otherwise. It will also be understood that some or all of the figures may be schematic representations for purposes of illustration and do not necessarily depict the actual relative sizes or locations of the elements shown.

[0007] Figure 1 is a diagrammatic representation of an example computational architecture that determines sequence representations to be analyzed by a machine learning classification model to detect a tumor-related biological condition, according to one or more example implementations.

[0008] Figure 2 is a diagrammatic representation of an example process to apply one or more criteria to identify a group of sequence representations to be analyzed by a machine learning classification model to detect a tumor-related biological condition, according to one or more example implementations.

[0009] Figure 3 is a diagrammatic representation of an example computational environment to determine one or more criteria that can be applied to sequence representations to identify a group of sequence representations corresponding to nucleic acid molecules derived from subjects in which a tumor-related biological condition is present, according to one or more example implementations.

[0010] Figure 4 is a flow diagram of an example process to determine sequence representations to be analyzed by a machine learning classification model to detect a tumor-related biological condition, according to one or more implementations.

[0011] Figure 5 is a block diagram illustrating components of a machine, in the form of a computer system, that may read and execute instructions from one or more machine-readable media to perform any one or more methodologies described herein, in accordance with one or more example implementations.

[0012] Figure 6 is block diagram illustrating a representative software architecture that may be used in conjunction with one or more hardware architectures described herein, in accordance with one or more example implementations. SUMMARY

[0013] In some aspects, the techniques described herein relate to a method including: obtaining first training data from first samples derived from first subjects in which a tumor is detected, the first training data including first nucleotide sequence representations corresponding to first nucleic acid molecules included in the first samples; obtaining second training data from second samples derived from second subjects in which a tumor is not detected, the second training data includingsecond nucleotide sequence representations corresponding to second nucleic acid molecules included in the second samples; performing a training process for one or more machine learning models, wherein the training process includes: determining first quantitative measures based on first values for features of the first nucleotide sequence representations; determining second quantitative measures based on second values for the features of the second nucleotide sequence representations; and determining, based on the first quantitative measures and the second quantitative measures, one or more threshold quantitative measures to identify nucleotide sequence representations to provide to a machine learning classification model to determine an indication of a tumor-related biological condition; obtaining additional nucleotide sequence representations derived from an additional subject and corresponding to additional nucleic acid molecules included in an additional sample obtained from the additional subject; determining additional quantitative measures based on additional values for the features of the additional nucleotide sequence representations; determining a subset of the additional nucleotide sequence representations for which the additional quantitative measures correspond to the one or more threshold quantitative measures; and providing the subset of the additional nucleotide sequence representations to the machine learning classification model to determine the indication of the tumor-related biological condition for the additional subject.

[0014] In some aspects, the techniques described herein relate to a method, including: for individual genomic regions, determining, based on differences between the first values and the second values, probabilities that specified sets of values of the features for nucleotide sequence representations correspond to samples obtained from subjects in which the tumor-related biological condition is present.

[0015] In some aspects, the techniques described herein relate to a method, including: determining, based on the probabilities, that nucleic acid molecules aligned with one or more genomic regions have less than a threshold probability of identifying samples obtained from subjects in which the tumor-related biological condition is present; and producing a diagnostic test that does not enrich the one or more genomic regions.

[0016] In some aspects, the techniques described herein relate to a method, including: generating a probability map that indicates the probabilities for the individual genomic regions.

[0017] In some aspects, the techniques described herein relate to a method, wherein: the first training data includes first methylation data that indicates a first number of methylated cytosine- guanine dinucleotides (CpGs) present in the first nucleic acid molecules; and the second training data includes second methylation data that indicates a second number of methylated CpGs present in the second nucleic acid molecules; and the method includes: determining the firstvalues for the features with respect to the first nucleotide sequence representations based on first characteristics of the first nucleotide sequence representations and the first methylation data; and determining the second values for the features with respect to the second nucleotide sequence representations based on second characteristics of the second nucleotide sequence representations and the second methylation data.

[0018] In some aspects, the techniques described herein relate to a method, wherein determining the first values for the features includes determining a first number of CpGs within one or more regions of the first nucleotide sequence representations and determining the second values for the features includes determining a second number of CpGs within the one or more regions of the second nucleotide sequence representations.

[0019] In some aspects, the techniques described herein relate to a method, wherein determining the first values for the features includes determining a first amount of methylated CpGs within the one or more regions of the first nucleotide sequence representations and determining the second values for the features includes determining a second amount of methylated CpGs within the one or more regions of the second nucleotide sequence representations.

[0020] In some aspects, the techniques described herein relate to a method, wherein the first amount of methylated CpGs within the one or more regions includes a first count of individual methylated CpGs for the first nucleotide sequence representations within the one or more regions and the second amount of methylated CpGs within the one or more regions includes a second count of individual methylated CpGs for the second nucleotide sequence representations within the one or more regions.

[0021] In some aspects, the techniques described herein relate to a method, wherein the first amount of methylated CpGs within the one or more regions for the first nucleotide sequence representations corresponds to a first range of methylated CpGs and the second amount of methylated CpGs within the one or more regions for the second nucleotide sequence representations corresponds to a second range of methylated CpGs.

[0022] In some aspects, the techniques described herein relate to a method, wherein determining the first values for the features includes determining a first number of CpGs within individual first nucleic acid molecules and determining the second values for the features includes determining a second number of CpGs within individual second nucleic acid molecules.

[0023] In some aspects, the techniques described herein relate to a method, wherein determining the first values for the features includes determining a first amount of methylated CpGs within individual first nucleic acid molecules and determining the second values for the features includesdetermining a second amount of methylated CpGs within individual second nucleic acid molecules.

[0024] In some aspects, the techniques described herein relate to a method, wherein the first amount of methylated CpGs includes a first count of individual methylated CpGs within the first nucleic acid molecules and the second amount of methylated CpGs includes a second count of individual methylated CpGs within the second nucleic acid molecules.

[0025] In some aspects, the techniques described herein relate to a method, wherein the first amount of methylated CpGs corresponds to a first range of methylated CpGs within the first nucleic acid molecules and the second amount of methylated CpGs corresponds to a second range of methylated CpGs within the second nucleic acid molecules.

[0026] In some aspects, the techniques described herein relate to a method, wherein determining the first values for the features and the second values for the features includes determining, a first number of first sequencing reads that correspond to individual first nucleic acid molecules included in the first samples and determining a number of second sequencing reads that correspond to individual second nucleic acid molecules included in the second samples.

[0027] In some aspects, the techniques described herein relate to a method, wherein determining the first values for the features includes determining a first number of restriction enzyme cut sites for the first nucleotide sequence representations and determining the second values for the features includes determining a second number of restriction enzyme cut sites for the second nucleotide sequence representations.

[0028] In some aspects, the techniques described herein relate to a method, wherein determining the first values for the features includes determining first lengths for individual first nucleotide sequence representations and determining the second values for the features includes determining second lengths for individual second nucleotide sequence representations.

[0029] In some aspects, the techniques described herein relate to a method, wherein determining the first values for the features and the second values for the features includes determining, for individual first nucleotide sequence representations and individual second nucleotide sequence representations, an offset of a start position with respect to a genomic region in which the individual first nucleotide sequence representations and individual second nucleotide sequence representations are located.

[0030] In some aspects, the techniques described herein relate to a method, wherein determining the first values for the features and the second values for the features includes determining, for a given genomic region, a number of the first nucleic acid molecules and a number of the second nucleic acid molecules aligned with the given genomic region in relation to a number of the firstnucleic acid molecules and a number of the second nucleic acid molecules aligned with one or more control genomic regions.

[0031] In some aspects, the techniques described herein relate to a method, wherein the machine learning classification model includes a random forests model.

[0032] In some aspects, the techniques described herein relate to a method, wherein the first quantitative measures and the second quantitative measures are determined using a logistic regression model.

[0033] In some aspects, the techniques described herein relate to a method, wherein the first values for the features with respect to the first nucleotide sequence representations, the second values for the features with respect to the second nucleotide sequence representations, and the additional values for the features with respect to the additional nucleotide sequence representations are represented as vectors that are provided to the machine learning classification model.

[0034] In some aspects, the techniques described herein relate to a method, wherein the indication of the tumor-related biological condition indicates that a tumor is present or a tumor is absent in a subject.

[0035] In some aspects, the techniques described herein relate to a method, wherein the indication of the tumor-related biological condition corresponds to tumor fraction for a given sample.

[0036] In some aspects, the techniques described herein relate to a method, wherein the indication of the tumor-related biological condition corresponds to reoccurrence of cancer within a subject.

[0037] In some aspects, the techniques described herein relate to a method, wherein the indication of the tumor-related biological condition indicates a responsiveness of a subject to one or more treatments for the tumor-related biological condition.

[0038] In some aspects, the techniques described herein relate to a method, wherein the first quantitative measures and the second quantitative measures indicate differences between nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present and nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is not present.

[0039] In some aspects, the techniques described herein relate to a method, wherein the first quantitative measures and the second quantitative measures indicate probabilities of nucleotidesequence representations corresponding to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present.

[0040] In some aspects, the techniques described herein relate to a method, wherein the one or more threshold quantitative measures comprise molecule selection data indicative of nucleic acid molecules derived from subjects in which a tumor-related biological condition is present; and the method includes: obtaining a plurality of further nucleotide sequence representations derived from one or more further samples of a further subject, the plurality of further nucleotide sequence representations corresponding to further nucleic acid molecules included in the one or more further samples; determining further quantitative measures based on further values for one or more features of the plurality of further nucleotide sequence representations; determining a subset of the further plurality of nucleotide sequence representations for which the further quantitative measures correspond to the one or more threshold quantitative measures; and executing a computational model to determine to determine an indication of a somatic variant being present in the one or more further samples, wherein weightings of the subset of the further plurality of nucleotide sequence representations within the computational model are modified to determine the indication of the somatic variant being present.

[0041] In some aspects, the techniques described herein relate to a method wherein the one or more threshold quantitative measures include a threshold quantitative measure indicating at least a threshold probability of a tumor-being present in a subject.

[0042] In some aspects, the techniques described herein relate to a method, wherein the subset of the further quantitative measures are greater than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the further plurality of nucleotide sequence representations are upweighted with respect to initial weightings of the further plurality of nucleotide sequence representations.

[0043] In some aspects, the techniques described herein relate to a method, wherein the subset of the further quantitative measures are less than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the further plurality of nucleotide sequence representations are downweighted with respect to initial weightings of the further plurality of nucleotide sequence representations

[0044] In some aspects, the techniques described herein relate to a method, wherein the somatic variant includes a copy number deletion.

[0045] In some aspects, the techniques described herein relate to a method, wherein the somatic variant includes a copy number duplication.

[0046] In some aspects, the techniques described herein relate to a method, wherein the somatic variant includes a single nucleotide variant.

[0047] In some aspects, the techniques described herein relate to a method including: obtaining molecule selection data generated from a training process of one or more machine learning models, the molecule selection data including one or more threshold quantitative measures indicative of nucleic acid molecules derived from subjects in which a tumor-related biological condition is present; obtaining a plurality of nucleotide sequence representations derived from one or more samples of a subject, the plurality of nucleotide sequence representations corresponding to nucleic acid molecules included in the one or more samples; determining quantitative measures based on values for features of the plurality of nucleotide sequence representations; determining a subset of the plurality of nucleotide sequence representations for which the quantitative measures correspond to the one or more threshold quantitative measures; providing the subset of the plurality of nucleotide sequence representations to a machine learning classification model; and determining, by the machine learning classification model, an indication of the tumor-related biological condition for the subject.

[0048] In some aspects, the techniques described herein relate to a method, including obtaining methylation data that indicates a number of methylated cytosine-guanine dinucleotides (CpGs) present in the nucleic acid molecules; and the method includes: determining the values for the features with respect to the plurality of nucleotide sequence representations based on characteristics of the plurality of nucleotide sequence representations and the methylation data.

[0049] In some aspects, the techniques described herein relate to a method, wherein determining the values for the features includes determining a number of CpGs within one or more regions of the plurality of nucleotide sequence representations.

[0050] In some aspects, the techniques described herein relate to a method, wherein determining the values for the features includes determining an amount of methylated CpGs within the one or more regions of the plurality of nucleotide sequence representations.

[0051] In some aspects, the techniques described herein relate to a method, wherein the amount of methylated CpGs within the one or more regions includes a count of individual methylated CpGs for the plurality of nucleotide sequence representations within the one or more regions.

[0052] In some aspects, the techniques described herein relate to a method, wherein the amount of methylated CpGs within the one or more regions for the plurality of nucleotide sequence representations corresponds to a range of methylated CpGs.

[0053] In some aspects, the techniques described herein relate to a method, wherein determining the values for the features includes determining a number of CpGs within individual nucleic acid molecules.

[0054] In some aspects, the techniques described herein relate to a method, wherein determining the values for the features includes determining an amount of methylated CpGs within individual nucleic acid molecules.

[0055] In some aspects, the techniques described herein relate to a method, wherein the amount of methylated CpGs includes a count of individual methylated CpGs within the nucleic acid molecules.

[0056] In some aspects, the techniques described herein relate to a method, wherein the amount of methylated CpGs corresponds to a range of methylated CpGs within the nucleic acid molecules.

[0057] In some aspects, the techniques described herein relate to a method, wherein determining the values for the features includes determining a number of sequencing reads that correspond to individual nucleic acid molecules included in the one or more samples.

[0058] In some aspects, the techniques described herein relate to a method, wherein determining the values for the features includes determining a number of restriction enzyme cut sites for the plurality of nucleotide sequence representations.

[0059] In some aspects, the techniques described herein relate to a method, wherein determining the values for the features includes determining lengths for individual nucleotide sequence representations.

[0060] In some aspects, the techniques described herein relate to a method, wherein determining the values for the features includes determining, for individual nucleotide sequence representations an offset of a start position with respect to a genomic region in which the individual nucleotide sequence representations are located.

[0061] In some aspects, the techniques described herein relate to a method, wherein determining the values for the features includes determining, for a given genomic region, a number of the nucleic acid molecules aligned with the given genomic region in relation to a number of the nucleic acid molecules.

[0062] In some aspects, the techniques described herein relate to a method, wherein the machine learning classification model includes a random forests model.

[0063] In some aspects, the techniques described herein relate to a method, wherein the quantitative measures are determined using a logistic regression model.

[0064] In some aspects, the techniques described herein relate to a method, wherein the values for the features with respect to the plurality of nucleotide sequence representations are represented as vectors that are provided to the machine learning classification model.

[0065] In some aspects, the techniques described herein relate to a method, wherein the indication of the tumor-related biological condition indicates that a tumor is present or a tumor is absent in the subject.

[0066] In some aspects, the techniques described herein relate to a method, wherein the indication of the tumor-related biological condition corresponds to tumor fraction for the one or more samples.

[0067] In some aspects, the techniques described herein relate to a method, wherein the indication of the tumor-related biological condition corresponds to reoccurrence of cancer within the subject.

[0068] In some aspects, the techniques described herein relate to a method, wherein the indication of the tumor-related biological condition indicates a responsiveness of the subject to one or more treatments for the tumor-related biological condition.

[0069] In some aspects, the techniques described herein relate to a method, wherein the quantitative measures indicate differences between nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present and nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is not present.

[0070] In some aspects, the techniques described herein relate to a method, wherein the quantitative measures indicate probabilities of nucleotide sequence representations corresponding to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present.

[0071] In some aspects, the techniques described herein relate to a method including obtaining first training data from first samples derived from first subjects in which a tumor is detected, the first training data including first nucleotide sequence representations corresponding to first nucleic acid molecules included in the first samples; obtaining second training data from second samples derived from second subjects in which a tumor is not detected, the second training data including second nucleotide sequence representations corresponding to second nucleic acid molecules included in the second samples; performing a training process for one or more machine learning models, wherein the training process includes: determining first quantitative measures based on first values for features of the first nucleotide sequence representations; determining second quantitative measures based on second values for the features of the second nucleotidesequence representations; and determining, based on the first quantitative measures and the second quantitative measures, one or more threshold quantitative measures to identify nucleotide sequence representations to provide to a machine learning classification model to determine an indication of a tumor-related biological condition.

[0072] In some aspects, the techniques described herein relate to a method including obtaining a plurality of additional nucleotide sequence representations derived from one or more additional samples of an additional subject, the plurality of additional nucleotide sequence representations corresponding to additional nucleic acid molecules included in the one or more additional samples; determining additional quantitative measures based on additional values for one or more features of the plurality of additional nucleotide sequence representations; determining a subset of the additional plurality of nucleotide sequence representations for which the additional quantitative measures correspond to the one or more threshold quantitative measures; and executing a computational model to determine to determine an indication of a somatic variant being present in the one or more additional samples, wherein weightings of the subset of the additional plurality of nucleotide sequence representations within the computational model are modified to determine the indication of the somatic variant being present.

[0073] In some aspects, the techniques described herein relate to a method, wherein the one or more threshold quantitative measures include a threshold quantitative measure indicating at least a threshold probability of a tumor-being present in a subject.

[0074] In some aspects, the techniques described herein relate to a method, wherein the subset of the additional quantitative measures are greater than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the additional plurality of nucleotide sequence representations are upweighted with respect to initial weightings of the additional plurality of nucleotide sequence representations.

[0075] In some aspects, the techniques described herein relate to a method, wherein the subset of the additional quantitative measures are less than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the additional plurality of nucleotide sequence representations are downweighted with respect to initial weightings of the additional plurality of nucleotide sequence representations

[0076] In some aspects, the techniques described herein relate to a method, wherein the somatic variant includes a copy number deletion.

[0077] In some aspects, the techniques described herein relate to a method, wherein the somatic variant includes a copy number duplication.

[0078] In some aspects, the techniques described herein relate to a method, wherein the somatic variant includes a single nucleotide variant.

[0079] In some aspects, the techniques described herein relate to one or more computing apparatuses, including: one or more hardware processors; and memory storing computer- readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations comprising: obtaining first training data from first samples derived from first subjects in which a tumor is detected, the first training data including first nucleotide sequence representations corresponding to first nucleic acid molecules included in the first samples; obtaining second training data from second samples derived from second subjects in which a tumor is not detected, the second training data including second nucleotide sequence representations corresponding to second nucleic acid molecules included in the second samples; performing a training process for one or more machine learning models, wherein the training process includes: determining first quantitative measures based on first values for features of the first nucleotide sequence representations; determining second quantitative measures based on second values for the features of the second nucleotide sequence representations; and determining, based on the first quantitative measures and the second quantitative measures, one or more threshold quantitative measures to identify nucleotide sequence representations to provide to a machine learning classification model to determine an indication of a tumor-related biological condition; obtaining additional nucleotide sequence representations derived from an additional subject and corresponding to additional nucleic acid molecules included in an additional sample obtained from the additional subject; determining additional quantitative measures based on additional values for the features of the additional nucleotide sequence representations; determining a subset of the additional nucleotide sequence representations for which the additional quantitative measures correspond to the one or more threshold quantitative measures; and providing the subset of the additional nucleotide sequence representations to the machine learning classification model to determine the indication of the tumor-related biological condition for the additional subject.

[0080] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the memory stores additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: for individual genomic regions, determining, based on differences between the first values and the second values, probabilities that specified sets of values of the features for nucleotide sequence representations correspond to samples obtained from subjects in which the tumor-related biological condition is present.

[0081] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the memory stores additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: determining, based on the probabilities, that nucleic acid molecules aligned with one or more genomic regions have less than a threshold probability of identifying samples obtained from subjects in which the tumor-related biological condition is present; and producing a diagnostic test that does not enrich the one or more genomic regions.

[0082] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the memory stores additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: generating a probability map that indicates the probabilities for the individual genomic regions.

[0083] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein: the first training data includes first methylation data that indicates a first number of methylated cytosine-guanine dinucleotides (CpGs) present in the first nucleic acid molecules; and the second training data includes second methylation data that indicates a second number of methylated CpGs present in the second nucleic acid molecules; and the method includes: determining the first values for the features with respect to the first nucleotide sequence representations based on first characteristics of the first nucleotide sequence representations and the first methylation data; and determining the second values for the features with respect to the second nucleotide sequence representations based on second characteristics of the second nucleotide sequence representations and the second methylation data.

[0084] In some aspects, the techniques described herein relate to one or more computer apparatuses wherein determining the first values for the features includes determining a first number of CpGs within one or more regions of the first nucleotide sequence representations and determining the second values for the features includes determining a second number of CpGs within the one or more regions of the second nucleotide sequence representations.

[0085] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the first values for the features includes determining a first amount of methylated CpGs within the one or more regions of the first nucleotide sequence representations and determining the second values for the features includes determining a second amount of methylated CpGs within the one or more regions of the second nucleotide sequence representations.

[0086] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the first amount of methylated CpGs within the one or more regions includes a first count of individual methylated CpGs for the first nucleotide sequence representations within the one or more regions and the second amount of methylated CpGs within the one or more regions includes a second count of individual methylated CpGs for the second nucleotide sequence representations within the one or more regions.

[0087] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the first amount of methylated CpGs within the one or more regions for the first nucleotide sequence representations corresponds to a first range of methylated CpGs and the second amount of methylated CpGs within the one or more regions for the second nucleotide sequence representations corresponds to a second range of methylated CpGs.

[0088] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the first values for the features includes determining a first number of CpGs within individual first nucleic acid molecules and determining the second values for the features includes determining a second number of CpGs within individual second nucleic acid molecules.

[0089] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the first values for the features includes determining a first amount of methylated CpGs within individual first nucleic acid molecules and determining the second values for the features includes determining a second amount of methylated CpGs within individual second nucleic acid molecules.

[0090] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the first amount of methylated CpGs includes a first count of individual methylated CpGs within the first nucleic acid molecules and the second amount of methylated CpGs includes a second count of individual methylated CpGs within the second nucleic acid molecules.

[0091] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the first amount of methylated CpGs corresponds to a first range of methylated CpGs within the first nucleic acid molecules and the second amount of methylated CpGs corresponds to a second range of methylated CpGs within the second nucleic acid molecules.

[0092] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the first values for the features and the second values for the features includes determining, a first number of first sequencing reads that correspond toindividual first nucleic acid molecules included in the first samples and determining a number of second sequencing reads that correspond to individual second nucleic acid molecules included in the second samples.

[0093] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the first values for the features includes determining a first number of restriction enzyme cut sites for the first nucleotide sequence representations and determining the second values for the features includes determining a second number of restriction enzyme cut sites for the second nucleotide sequence representations.

[0094] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the first values for the features includes determining first lengths for individual first nucleotide sequence representations and determining the second values for the features includes determining second lengths for individual second nucleotide sequence representations.

[0095] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the first values for the features and the second values for the features includes determining, for individual first nucleotide sequence representations and individual second nucleotide sequence representations, an offset of a start position with respect to a genomic region in which the individual first nucleotide sequence representations and individual second nucleotide sequence representations are located.

[0096] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the first values for the features and the second values for the features includes determining, for a given genomic region, a number of the first nucleic acid molecules and a number of the second nucleic acid molecules aligned with the given genomic region in relation to a number of the first nucleic acid molecules and a number of the second nucleic acid molecules aligned with one or more control genomic regions.

[0097] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the machine learning classification model includes a random forests model.

[0098] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the first quantitative measures and the second quantitative measures are determined using a logistic regression model.

[0099] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the first values for the features with respect to the first nucleotide sequence representations, the second values for the features with respect to the second nucleotide sequence representations, and the additional values for the features with respect to the additionalnucleotide sequence representations are represented as vectors that are provided to the machine learning classification model.

[0100] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the indication of the tumor-related biological condition indicates that a tumor is present or a tumor is absent in a subject.

[0101] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the indication of the tumor-related biological condition corresponds to tumor fraction for a given sample.

[0102] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the indication of the tumor-related biological condition corresponds to reoccurrence of cancer within a subject.

[0103] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the indication of the tumor-related biological condition indicates a responsiveness of a subject to one or more treatments for the tumor-related biological condition.

[0104] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the first quantitative measures and the second quantitative measures indicate differences between nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present and nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is not present.

[0105] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the first quantitative measures and the second quantitative measures indicate probabilities of nucleotide sequence representations corresponding to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present.

[0106] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the one or more threshold quantitative measures comprise molecule selection data indicative of nucleic acid molecules derived from subjects in which a tumor-related biological condition is present; and the memory stores additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: obtaining a plurality of further nucleotide sequence representations derived from one or more further samples of a further subject, the plurality of further nucleotide sequence representations corresponding to further nucleic acid molecules included in the one or more further samples; determining further quantitative measures based on further values for one or more features of the plurality of further nucleotide sequencerepresentations; determining a subset of the further plurality of nucleotide sequence representations for which the further quantitative measures correspond to the one or more threshold quantitative measures; and executing a computational model to determine to determine an indication of a somatic variant being present in the one or more further samples, wherein weightings of the subset of the further plurality of nucleotide sequence representations within the computational model are modified to determine the indication of the somatic variant being present.

[0107] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the one or more threshold quantitative measures include a threshold quantitative measure indicating at least a threshold probability of a tumor-being present in a subject.

[0108] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the subset of the further quantitative measures are greater than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the further plurality of nucleotide sequence representations are upweighted with respect to initial weightings of the further plurality of nucleotide sequence representations.

[0109] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the subset of the further quantitative measures are less than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the further plurality of nucleotide sequence representations are downweighted with respect to initial weightings of the further plurality of nucleotide sequence representations

[0110] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the somatic variant includes a copy number deletion.

[0111] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the somatic variant includes a copy number duplication.

[0112] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the somatic variant includes a single nucleotide variant.

[0113] In some aspects, the techniques described herein relate to one or more computer apparatuses comprising: one or more hardware processors; and memory storing computer- readable instructions that, when executed by the one or more hardware processors, causes the one or more hardware processors to perform operations comprising: obtaining molecule selection data generated from a training process of one or more machine learning models, the molecule selection data including one or more threshold quantitative measures indicative of nucleic acid molecules derived from subjects in which a tumor-related biological condition is present; obtaining a plurality of nucleotide sequence representations derived from one or more samples of a subject,the plurality of nucleotide sequence representations corresponding to nucleic acid molecules included in the one or more samples; determining quantitative measures based on values for features of the plurality of nucleotide sequence representations; determining a subset of the plurality of nucleotide sequence representations for which the quantitative measures correspond to the one or more threshold quantitative measures; providing the subset of the plurality of nucleotide sequence representations to a machine learning classification model; and determining, by the machine learning classification model, an indication of the tumor-related biological condition for the subject.

[0114] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the memory stores additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: obtaining methylation data that indicates a number of methylated cytosine-guanine dinucleotides (CpGs) present in the nucleic acid molecules; and the method includes: determining the values for the features with respect to the plurality of nucleotide sequence representations based on characteristics of the plurality of nucleotide sequence representations and the methylation data.

[0115] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the values for the features includes determining a number of CpGs within one or more regions of the plurality of nucleotide sequence representations.

[0116] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the values for the features includes determining an amount of methylated CpGs within the one or more regions of the plurality of nucleotide sequence representations.

[0117] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the amount of methylated CpGs within the one or more regions includes a count of individual methylated CpGs for the plurality of nucleotide sequence representations within the one or more regions.

[0118] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the amount of methylated CpGs within the one or more regions for the plurality of nucleotide sequence representations corresponds to a range of methylated CpGs.

[0119] In some aspects, the techniques described herein relate one or more computer apparatuses, wherein determining the values for the features includes determining a number of CpGs within individual nucleic acid molecules.

[0120] In some aspects, the techniques described herein relate one or more computer apparatuses, wherein determining the values for the features includes determining an amount of methylated CpGs within individual nucleic acid molecules.

[0121] In some aspects, the techniques described herein relate one or more computer apparatuses, wherein the amount of methylated CpGs includes a count of individual methylated CpGs within the nucleic acid molecules.

[0122] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the amount of methylated CpGs corresponds to a range of methylated CpGs within the nucleic acid molecules.

[0123] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the values for the features includes determining a number of sequencing reads that correspond to individual nucleic acid molecules included in the one or more samples.

[0124] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the values for the features includes determining a number of restriction enzyme cut sites for the plurality of nucleotide sequence representations.

[0125] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the values for the features includes determining lengths for individual nucleotide sequence representations.

[0126] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the values for the features includes determining, for individual nucleotide sequence representations an offset of a start position with respect to a genomic region in which the individual nucleotide sequence representations are located.

[0127] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein determining the values for the features includes determining, for a given genomic region, a number of the nucleic acid molecules aligned with the given genomic region in relation to a number of the nucleic acid molecules.

[0128] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the machine learning classification model includes a random forests model.

[0129] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the quantitative measures are determined using a logistic regression model.

[0130] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the values for the features with respect to the plurality of nucleotide sequence representations are represented as vectors that are provided to the machine learning classification model.

[0131] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the indication of the tumor-related biological condition indicates that a tumor is present or a tumor is absent in the subject.

[0132] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the indication of the tumor-related biological condition corresponds to tumor fraction for the one or more samples.

[0133] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the indication of the tumor-related biological condition corresponds to reoccurrence of cancer within the subject.

[0134] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the indication of the tumor-related biological condition indicates a responsiveness of the subject to one or more treatments for the tumor-related biological condition.

[0135] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the quantitative measures indicate differences between nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present and nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is not present.

[0136] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the quantitative measures indicate probabilities of nucleotide sequence representations corresponding to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present.

[0137] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the memory stores additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: obtaining first training data from first training samples derived from first training subjects in which a tumor is detected, the first training data including first training nucleotide sequence representations corresponding to first training nucleic acid molecules included in the first training samples; obtaining second training data from second training samples derived from second training subjects in which a tumor is not detected, thesecond training data including second training nucleotide sequence representations corresponding to second training nucleic acid molecules included in the second training samples; performing a training process for one or more machine learning models, wherein the training process includes: determining first training quantitative measures based on first values for features of the first training nucleotide sequence representations; determining second training quantitative measures based on second values for the features of the second training nucleotide sequence representations; and determining, based on the first training quantitative measures and the second training quantitative measures, the one or more threshold quantitative measures to identify one or more nucleotide sequence representations to provide to the machine learning classification model to determine indications of the tumor-related biological condition.

[0138] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the memory stores additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: obtaining a plurality of additional nucleotide sequence representations derived from one or more additional samples of an additional subject, the plurality of additional nucleotide sequence representations corresponding to additional nucleic acid molecules included in the one or more additional samples; determining additional quantitative measures based on additional values for one or more features of the plurality of additional nucleotide sequence representations; determining a subset of the additional plurality of nucleotide sequence representations for which the additional quantitative measures correspond to the one or more threshold quantitative measures; and executing a computational model to determine to determine an indication of a somatic variant being present in the one or more additional samples, wherein weightings of the subset of the additional plurality of nucleotide sequence representations within the computational model are modified to determine the indication of the somatic variant being present.

[0139] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the one or more threshold quantitative measures include a threshold quantitative measure indicating at least a threshold probability of a tumor-being present in a subject.

[0140] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the subset of the additional quantitative measures are greater than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the additional plurality of nucleotide sequence representations areupweighted with respect to initial weightings of the additional plurality of nucleotide sequence representations.

[0141] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the subset of the additional quantitative measures are less than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the additional plurality of nucleotide sequence representations are downweighted with respect to initial weightings of the additional plurality of nucleotide sequence representations

[0142] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the somatic variant includes a copy number deletion.

[0143] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the somatic variant includes a copy number duplication.

[0144] In some aspects, the techniques described herein relate to one or more computing apparatuses, wherein the somatic variant includes a single nucleotide variant.

[0145] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media storing computer-readable instructions that, when executed by one or more hardware processors, cause the one or more hardware processors to perform operations comprising: obtaining first training data from first samples derived from first subjects in which a tumor is detected, the first training data including first nucleotide sequence representations corresponding to first nucleic acid molecules included in the first samples; obtaining second training data from second samples derived from second subjects in which a tumor is not detected, the second training data including second nucleotide sequence representations corresponding to second nucleic acid molecules included in the second samples; performing a training process for one or more machine learning models, wherein the training process includes: determining first quantitative measures based on first values for features of the first nucleotide sequence representations; determining second quantitative measures based on second values for the features of the second nucleotide sequence representations; and determining, based on the first quantitative measures and the second quantitative measures, one or more threshold quantitative measures to identify nucleotide sequence representations to provide to a machine learning classification model to determine an indication of a tumor-related biological condition; obtaining additional nucleotide sequence representations derived from an additional subject and corresponding to additional nucleic acid molecules included in an additional sample obtained from the additional subject; determining additional quantitative measures based on additional values for the features of the additional nucleotide sequence representations; determining a subset of the additional nucleotide sequence representations for which theadditional quantitative measures correspond to the one or more threshold quantitative measures; and providing the subset of the additional nucleotide sequence representations to the machine learning classification model to determine the indication of the tumor-related biological condition for the additional subject.

[0146] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media storing additional computer-readable instructions that, when executed by one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: for individual genomic regions, determining, based on differences between the first values and the second values, probabilities that specified sets of values of the features for nucleotide sequence representations correspond to samples obtained from subjects in which the tumor-related biological condition is present.

[0147] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media storing additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: determining, based on the probabilities, that nucleic acid molecules aligned with one or more genomic regions have less than a threshold probability of identifying samples obtained from subjects in which the tumor-related biological condition is present; and producing a diagnostic test that does not enrich the one or more genomic regions.

[0148] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media storing additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: generating a probability map that indicates the probabilities for the individual genomic regions.

[0149] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein: the first training data includes first methylation data that indicates a first number of methylated cytosine-guanine dinucleotides (CpGs) present in the first nucleic acid molecules; and the second training data includes second methylation data that indicates a second number of methylated CpGs present in the second nucleic acid molecules; and the method includes: determining the first values for the features with respect to the first nucleotide sequence representations based on first characteristics of the first nucleotide sequence representations and the first methylation data; and determining the second values for the features with respect to the second nucleotide sequence representations based on second characteristics of the second nucleotide sequence representations and the second methylation data.

[0150] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media wherein determining the first values for the features includes determining a first number of CpGs within one or more regions of the first nucleotide sequence representations and determining the second values for the features includes determining a second number of CpGs within the one or more regions of the second nucleotide sequence representations.

[0151] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the first values for the features includes determining a first amount of methylated CpGs within the one or more regions of the first nucleotide sequence representations and determining the second values for the features includes determining a second amount of methylated CpGs within the one or more regions of the second nucleotide sequence representations.

[0152] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the first amount of methylated CpGs within the one or more regions includes a first count of individual methylated CpGs for the first nucleotide sequence representations within the one or more regions and the second amount of methylated CpGs within the one or more regions includes a second count of individual methylated CpGs for the second nucleotide sequence representations within the one or more regions.

[0153] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the first amount of methylated CpGs within the one or more regions for the first nucleotide sequence representations corresponds to a first range of methylated CpGs and the second amount of methylated CpGs within the one or more regions for the second nucleotide sequence representations corresponds to a second range of methylated CpGs.

[0154] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the first values for the features includes determining a first number of CpGs within individual first nucleic acid molecules and determining the second values for the features includes determining a second number of CpGs within individual second nucleic acid molecules.

[0155] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the first values for the features includes determining a first amount of methylated CpGs within individual first nucleic acid molecules and determining the second values for the features includes determining a second amount of methylated CpGs within individual second nucleic acid molecules.

[0156] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the first amount of methylated CpGs includes a first count of individual methylated CpGs within the first nucleic acid molecules and the second amount of methylated CpGs includes a second count of individual methylated CpGs within the second nucleic acid molecules.

[0157] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the first amount of methylated CpGs corresponds to a first range of methylated CpGs within the first nucleic acid molecules and the second amount of methylated CpGs corresponds to a second range of methylated CpGs within the second nucleic acid molecules.

[0158] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the first values for the features and the second values for the features includes determining, a first number of first sequencing reads that correspond to individual first nucleic acid molecules included in the first samples and determining a number of second sequencing reads that correspond to individual second nucleic acid molecules included in the second samples.

[0159] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the first values for the features includes determining a first number of restriction enzyme cut sites for the first nucleotide sequence representations and determining the second values for the features includes determining a second number of restriction enzyme cut sites for the second nucleotide sequence representations.

[0160] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the first values for the features includes determining first lengths for individual first nucleotide sequence representations and determining the second values for the features includes determining second lengths for individual second nucleotide sequence representations.

[0161] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the first values for the features and the second values for the features includes determining, for individual first nucleotide sequence representations and individual second nucleotide sequence representations, an offset of a start position with respect to a genomic region in which the individual first nucleotide sequence representations and individual second nucleotide sequence representations are located.

[0162] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the first values for the features and the second values for the features includes determining, for a given genomic region, a number of the first nucleic acid molecules and a number of the second nucleic acid molecules aligned with the given genomic region in relation to a number of the first nucleic acid molecules and a number of the second nucleic acid molecules aligned with one or more control genomic regions.

[0163] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the machine learning classification model includes a random forests model.

[0164] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the first quantitative measures and the second quantitative measures are determined using a logistic regression model.

[0165] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the first values for the features with respect to the first nucleotide sequence representations, the second values for the features with respect to the second nucleotide sequence representations, and the additional values for the features with respect to the additional nucleotide sequence representations are represented as vectors that are provided to the machine learning classification model.

[0166] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the indication of the tumor-related biological condition indicates that a tumor is present or a tumor is absent in a subject.

[0167] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the indication of the tumor-related biological condition corresponds to tumor fraction for a given sample.

[0168] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the indication of the tumor-related biological condition corresponds to reoccurrence of cancer within a subject.

[0169] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the indication of the tumor-related biological condition indicates a responsiveness of a subject to one or more treatments for the tumor-related biological condition.

[0170] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the first quantitative measures and the second quantitative measures indicate differences between nucleotide sequence representations thatcorrespond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present and nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is not present.

[0171] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the first quantitative measures and the second quantitative measures indicate probabilities of nucleotide sequence representations corresponding to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present.

[0172] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the one or more threshold quantitative measures comprise molecule selection data indicative of nucleic acid molecules derived from subjects in which a tumor-related biological condition is present; and the non-transitory computer-readable media store additional computer-readable instructions that, when executed by one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: obtaining a plurality of further nucleotide sequence representations derived from one or more further samples of a further subject, the plurality of further nucleotide sequence representations corresponding to further nucleic acid molecules included in the one or more further samples; determining further quantitative measures based on further values for one or more features of the plurality of further nucleotide sequence representations; determining a subset of the further plurality of nucleotide sequence representations for which the further quantitative measures correspond to the one or more threshold quantitative measures; and executing a computational model to determine to determine an indication of a somatic variant being present in the one or more further samples, wherein weightings of the subset of the further plurality of nucleotide sequence representations within the computational model are modified to determine the indication of the somatic variant being present.

[0173] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the one or more threshold quantitative measures include a threshold quantitative measure indicating at least a threshold probability of a tumor- being present in a subject.

[0174] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the subset of the further quantitative measures are greater than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the further plurality of nucleotide sequence representationsare upweighted with respect to initial weightings of the further plurality of nucleotide sequence representations.

[0175] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the subset of the further quantitative measures are less than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the further plurality of nucleotide sequence representations are downweighted with respect to initial weightings of the further plurality of nucleotide sequence representations

[0176] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the somatic variant includes a copy number deletion.

[0177] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the somatic variant includes a copy number duplication.

[0178] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the somatic variant includes a single nucleotide variant.

[0179] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, comprising computer-readable instructions that, when executed by one or more hardware processors, causes the one or more hardware processors to perform operations comprising: obtaining molecule selection data generated from a training process of one or more machine learning models, the molecule selection data including one or more threshold quantitative measures indicative of nucleic acid molecules derived from subjects in which a tumor-related biological condition is present; obtaining a plurality of nucleotide sequence representations derived from one or more samples of a subject, the plurality of nucleotide sequence representations corresponding to nucleic acid molecules included in the one or more samples; determining quantitative measures based on values for features of the plurality of nucleotide sequence representations; determining a subset of the plurality of nucleotide sequence representations for which the quantitative measures correspond to the one or more threshold quantitative measures; providing the subset of the plurality of nucleotide sequence representations to a machine learning classification model; and determining, by the machine learning classification model, an indication of the tumor-related biological condition for the subject.

[0180] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media storing additional computer-readable instructions that, whenexecuted by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: obtaining methylation data that indicates a number of methylated cytosine-guanine dinucleotides (CpGs) present in the nucleic acid molecules; and the method includes: determining the values for the features with respect to the plurality of nucleotide sequence representations based on characteristics of the plurality of nucleotide sequence representations and the methylation data.

[0181] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the values for the features includes determining a number of CpGs within one or more regions of the plurality of nucleotide sequence representations.

[0182] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the values for the features includes determining an amount of methylated CpGs within the one or more regions of the plurality of nucleotide sequence representations.

[0183] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the amount of methylated CpGs within the one or more regions includes a count of individual methylated CpGs for the plurality of nucleotide sequence representations within the one or more regions.

[0184] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the amount of methylated CpGs within the one or more regions for the plurality of nucleotide sequence representations corresponds to a range of methylated CpGs.

[0185] In some aspects, the techniques described herein relate one or more non- transitory computer-readable media, wherein determining the values for the features includes determining a number of CpGs within individual nucleic acid molecules.

[0186] In some aspects, the techniques described herein relate one or more non- transitory computer-readable media, wherein determining the values for the features includes determining an amount of methylated CpGs within individual nucleic acid molecules.

[0187] In some aspects, the techniques described herein relate one or more non- transitory computer-readable media, wherein the amount of methylated CpGs includes a count of individual methylated CpGs within the nucleic acid molecules.

[0188] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the amount of methylated CpGs corresponds to a range of methylated CpGs within the nucleic acid molecules.

[0189] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the values for the features includes determining a number of sequencing reads that correspond to individual nucleic acid molecules included in the one or more samples.

[0190] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the values for the features includes determining a number of restriction enzyme cut sites for the plurality of nucleotide sequence representations.

[0191] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the values for the features includes determining lengths for individual nucleotide sequence representations.

[0192] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the values for the features includes determining, for individual nucleotide sequence representations an offset of a start position with respect to a genomic region in which the individual nucleotide sequence representations are located.

[0193] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein determining the values for the features includes determining, for a given genomic region, a number of the nucleic acid molecules aligned with the given genomic region in relation to a number of the nucleic acid molecules.

[0194] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the machine learning classification model includes a random forests model.

[0195] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the quantitative measures are determined using a logistic regression model.

[0196] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the values for the features with respect to the plurality of nucleotide sequence representations are represented as vectors that are provided to the machine learning classification model.

[0197] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the indication of the tumor-related biological condition indicates that a tumor is present or a tumor is absent in the subject.

[0198] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the indication of the tumor-related biological condition corresponds to tumor fraction for the one or more samples.

[0199] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the indication of the tumor-related biological condition corresponds to reoccurrence of cancer within the subject.

[0200] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the indication of the tumor-related biological condition indicates a responsiveness of the subject to one or more treatments for the tumor- related biological condition.

[0201] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the quantitative measures indicate differences between nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present and nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is not present.

[0202] In some aspects, the techniques described herein relate to one or more computer apparatuses, wherein the quantitative measures indicate probabilities of nucleotide sequence representations corresponding to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present.

[0203] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the one or more non-transitory computer-readable media store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: obtaining first training data from first training samples derived from first training subjects in which a tumor is detected, the first training data including first training nucleotide sequence representations corresponding to first training nucleic acid molecules included in the first training samples; obtaining second training data from second training samples derived from second training subjects in which a tumor is not detected, the second training data including second training nucleotide sequence representations corresponding to second training nucleic acid molecules included in the second training samples; performing a training process for one or more machine learning models, wherein the training process includes: determining first training quantitative measures based on first values for features of the first training nucleotide sequence representations; determining second training quantitative measures based on secondvalues for the features of the second training nucleotide sequence representations; and determining, based on the first training quantitative measures and the second training quantitative measures, the one or more threshold quantitative measures to identify one or more nucleotide sequence representations to provide to the machine learning classification model to determine indications of the tumor-related biological condition.

[0204] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the one or more non-transitory computer-readable media store additional computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform additional operations comprising: obtaining a plurality of additional nucleotide sequence representations derived from one or more additional samples of an additional subject, the plurality of additional nucleotide sequence representations corresponding to additional nucleic acid molecules included in the one or more additional samples; determining additional quantitative measures based on additional values for one or more features of the plurality of additional nucleotide sequence representations; determining a subset of the additional plurality of nucleotide sequence representations for which the additional quantitative measures correspond to the one or more threshold quantitative measures; and executing a computational model to determine to determine an indication of a somatic variant being present in the one or more additional samples, wherein weightings of the subset of the additional plurality of nucleotide sequence representations within the computational model are modified to determine the indication of the somatic variant being present.

[0205] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the one or more threshold quantitative measures include a threshold quantitative measure indicating at least a threshold probability of a tumor- being present in a subject.

[0206] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the subset of the additional quantitative measures are greater than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the additional plurality of nucleotide sequence representations are upweighted with respect to initial weightings of the additional plurality of nucleotide sequence representations.

[0207] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the subset of the additional quantitative measures are less than a threshold quantitative measure of the one or more threshold quantitativemeasures, and the weightings of the subset of the additional plurality of nucleotide sequence representations are downweighted with respect to initial weightings of the additional plurality of nucleotide sequence representations

[0208] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the somatic variant includes a copy number deletion.

[0209] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the somatic variant includes a copy number duplication.

[0210] In some aspects, the techniques described herein relate to one or more non- transitory computer-readable media, wherein the somatic variant includes a single nucleotide variant.

[0211] Other technical features may be readily apparent to one skilled in the art from the following figures, descriptions, and claims. DEFINITIONS

[0212] In order for the present disclosure to be more readily understood, certain terms are first defined below. Additional definitions for the following terms and other terms may be set forth through the specification. If a definition of a term set forth below is inconsistent with a definition in an application or patent that is incorporated by reference, the definition set forth in this application should be used to understand the meaning of the term.

[0213] As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise. Thus, for example, a reference to “a method” includes one or more methods, and / or steps of the type described herein and / or which will become apparent to those persons of ordinary skill in the art upon reading this disclosure and so forth.

[0214] It is also to be understood that the terminology used herein is for the purpose of describing particular implementations only and is not intended to be limiting. Further, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. In describing and claiming the methods, computer readable media, and systems, the following terminology, and grammatical variants thereof, will be used in accordance with the definitions set forth below.

[0215] About: As used herein, “about” or “approximately” as applied to one or more values or elements of interest, refers to a value or element that is similar to a stated reference value or element. In certain implementations, the term “about” or “approximately” refers to a range of values or elements that falls within 25%, 20%, 19%, 18%, 17%, 16%, 15%, 14%, 13%, 12%, 11%, 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, or less in either direction (greater than or less than) of the stated reference value or element unless otherwise stated or otherwise evident from the context (except where such number would exceed 100% of a possible value or element).

[0216] Administer: As used herein, “administer” or “administering” a therapeutic agent (e.g., an immunological therapeutic agent) to a subject means to give, apply or bring the composition into contact with the subject. Administration can be accomplished by any of a number of routes, including, for example, topical, oral, subcutaneous, intramuscular, intraperitoneal, intravenous, intrathecal and intradermal.

[0217] Adapter: As used herein, “adapter” refers to a short nucleic acid (e.g., less than about 500 nucleotides, less than about 100 nucleotides, or less than about 50 nucleotides in length) that can be at least partially double-stranded and used to link to either or both ends of a given sample nucleic acid molecule. Adapters can include sequences of nucleic acid primer binding sites to permit amplification of a nucleic acid molecule flanked by adapters at both ends, and / or a sequencing primer binding site, including primer binding sites for sequencing applications, such as various next-generation sequencing (NGS) applications. Adapters can also include binding sites for capture probes, such as an oligonucleotide attached to a flow cell support or the like. Adapters can also include a nucleic acid tag as described herein. Nucleic acid tags can be positioned relative to amplification primer and sequencing primer binding sites, such that a nucleic acid tag is included in amplicons and sequence reads of a given nucleic acid molecule. The same or different adapters can be linked to the respective ends of a nucleic acid molecule. In some implementations, the same adapter is linked to the respective ends of the nucleic acid molecule except that the nucleic acid tag differs. In some implementations, the adapter is a Y- shaped adapter in which one end is blunt ended or tailed as described herein, for joining to a nucleic acid molecule, which is also blunt ended or tailed with one or more complementary nucleotides. In still other example implementations, an adapter is a bell-shaped adapter that includes a blunt or tailed end for joining to a nucleic acid molecule to be analyzed. Other examples of adapters include T-tailed and C-tailed adapters.

[0218] Alignment: As used herein, “alignment” or “align” refers to determining whether at least two sequence representations have at least a threshold amount of homology. In one or more examples, the threshold amount of homology can be at least about 90%, at least about 91%, atleast 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 99.5%, or at least about 99.9%. In situations where two sequence representations have at least the threshold amount of homology, the two sequence representations can be referred to as being “aligned.” .” In various examples, alignment can include determining whether a sequence representation has at least a threshold amount of homology with respect to a reference sequence.

[0219] Amplify: As used herein, “amplify” or “amplification” in the context of nucleic acids refers to the production of multiple copies of a polynucleotide, or a portion of the polynucleotide, starting from a small amount of the polynucleotide (e.g., a single polynucleotide molecule), where the amplification products or amplicons are generally detectable. Amplification of polynucleotides encompasses a variety of chemical and enzymatic processes.

[0220] Barcode: As used herein, “barcode” or “molecular barcode” in the context of nucleic acids refers to a nucleic acid molecule comprising a sequence that can serve as a molecular identifier. For example, individual "barcode" sequences can be added to each DNA fragment during next-generation sequencing (NGS) library preparation so that each read can be identified and sorted before the final data analysis. In one or more examples, individual barcode sequences can be added to DNA fragments during NGS library preparation so that reads corresponding to each unique molecule included in a sample can be identified.

[0221] Cancer Type: As used herein, “cancer type” refers to a type or subtype of cancer defined, e.g., by histopathology. Cancer type can be defined by any conventional criterion, such as on the basis of occurrence in a given tissue (e.g., blood cancers, central nervous system (CNS), brain cancers, lung cancers (small cell and non-small cell), skin cancers, nose cancers, throat cancers, liver cancers, bone cancers, lymphomas, pancreatic cancers, bowel cancers, rectal cancers, thyroid cancers, bladder cancers, kidney cancers, mouth cancers, stomach cancers, breast cancers, prostate cancers, ovarian cancers, lung cancers, intestinal cancers, soft tissue cancers, neuroendocrine cancers, gastroesophageal cancers, head and neck cancers, gynecological cancers, colorectal cancers, urothelial cancers, solid state cancers, heterogeneous cancers, homogenous cancers), unknown primary origin and the like, and / or of the same cell lineage (e.g., carcinoma, sarcoma, lymphoma, cholangiocarcinoma, leukemia, mesothelioma, melanoma, or glioblastoma) and / or cancers exhibiting cancer markers, such as Her2, CA15-3, CA19-9, CA-125, CEA, AFP, PSA, HCG, EGFR, KRAS, APC, RB1, hormone receptor, estrogen receptor (ER), progesterone receptor (PR), and NMP-22. Cancers can also be classified by stage (e.g., stage 1, 2, 3, or 4), cell morphology (e.g., small vs. non-small cell lung cancer) and whether of primary or secondary origin.

[0222] Carrier Signal: As used herein, “carrier signal” refers to any intangible medium that is capable of storing, encoding, or carrying transitory or non-transitory instructions for execution by a machine, and includes digital or analog communications signals or other intangible medium to facilitate communication of such instructions. Instructions may be transmitted or received over a network using a transitory or non-transitory transmission medium via a network interface device and using any one of a number of data transfer protocols.

[0223] Cell-Free Nucleic Acid: As used herein, “cell-free nucleic acid” refers to nucleic acids not contained within or otherwise bound to a cell or, in some implementations, nucleic acids remaining in a sample following the removal of intact cells. Cell-free nucleic acids can include, for example, all non-encapsulated nucleic acids sourced from a bodily fluid (e.g., blood, plasma, serum, urine, cerebrospinal fluid (CSF), etc.) from a subject. Cell-free nucleic acids include DNA (cfDNA), RNA (cfRNA), and hybrids thereof, including genomic DNA, mitochondrial DNA, circulating DNA, siRNA, miRNA, circulating RNA (cRNA), tRNA, rRNA, small nucleolar RNA (snoRNA), Piwi-interacting RNA (piRNA), long non-coding RNA (long ncRNA), and / or fragments of any of these. Cell-free nucleic acids can be double-stranded, single-stranded, or a hybrid thereof. A cell-free nucleic acid can be released into bodily fluid through secretion or cell death processes, e.g., cellular necrosis, apoptosis, or the like. Some cell-free nucleic acids are released into bodily fluid from cancer cells, e.g., circulating tumor DNA (ctDNA). Others are released from healthy cells. CtDNA can be non-encapsulated tumor-derived fragmented DNA. A cell-free nucleic acid can have one or more epigenetic modifications, for example, a cell-free nucleic acid can be acetylated, 5-methylated, ubiquitylated, phosphorylated, sumoylated, ribosylated, and / or citrullinated.

[0224] Cellular Nucleic Acids: As used herein, “cellular nucleic acids” means nucleic acids that are disposed within one or more cells at least at the point a sample is taken or collected from a subject, even if those nucleic acids are subsequently removed as part of a given analytical process.

[0225] Classification Region: As used herein, “classification region” refers to a genomic region that may show sequence-independent changes in neoplastic cells (e.g., tumor cells and cancer cells) or that may show sequence-independent changes in cfDNA from subjects having cancer relative to cfDNA from subjects in which cancer is not present. Examples of sequence- independent changes include, but are not limited to, changes in methylation rate (increases or decreases), nucleosome distribution, CTCF binding, transcription start sites, and regulatory protein binding regions. In one or more examples, sequence-independent changes in a classification region can indicate the presence of a single form of cancer in a subject. In one ormore additional examples, sequence-independent changes in a classification region can correspond to the presence of multiple forms in a subject. The classification region can be enriched by one or more probes. In addition, the classification region can be defined by a pair of primer binding sites. Further, the classification region can be defined by a predetermined beginning genomic locus and a predetermined ending genomic locus. In one or more examples, the genomic locus can correspond to at least one of one or more genomic positions or one or more genomic coordinates. In various examples, the classification region can include nucleotide sequences corresponding to one or more genomic regions that are to be amplified during one or more nucleic acid sequencing processes. The classification region can include from about 10 nucleotides to about 10,000 nucleotides, from about 50 nucleotides to about 8000 nucleotides, from about 100 nucleotides to about 5000 nucleotides, from about 50 nucleotides to about 2000 nucleotides, from about 25 nucleotides to about 250 nucleotides, from about 50 nucleotides to about 200 nucleotides, or from about 75 nucleotides to about 150 nucleotides. For instance, classification region can be a differentially methylated region. “Differentially methylated region” or “DMR” refers to a region of DNA, such as a region of a genome, having a detectably different degree of methylation in at least one cell or tissue type relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type; or having a detectably different degree of methylation in at least one cell or tissue type obtained from a subject having a disease or disorder relative to the degree of methylation in the same region of DNA in the same cell or tissue type obtained from a healthy subject. In some embodiments, a differentially methylated region has a detectably higher degree of methylation in at least one cell or tissue type relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type that contribute to cfDNA in healthy individuals, or from the same cell or tissue type from a healthy subject. In some embodiments, a differentially methylated region has a detectably lower degree of methylation in at least one cell or tissue type relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type, such as other immune cell types and / or cell types that contribute to cfDNA in healthy individuals, or from the same cell or tissue type from a healthy subject.

[0226] Communications Network: As used herein, “communications network” refers to one or more portions of a network that may be an ad hoc network, an intranet, an extranet, a virtual private network (VPN), a local area network (LAN), a wireless LAN (WLAN), a wide area network (WAN), a wireless WAN (WWAN), a metropolitan area network (MAN), the Internet, a portion of the Internet, a portion of the Public Switched Telephone Network (PSTN), a plain old telephone service (POTS) network, a cellular telephone network, a wireless network, a Wi-Fi®network, another type of network, or a combination of two or more such networks. For example, a network or a portion of a network may include a wireless or cellular network and the coupling may be a Code Division Multiple Access (CDMA) connection, a Global System for Mobile communications (GSM) connection, or other type of cellular or wireless coupling. In this example, the coupling may implement any of a variety of types of data transfer technology, such as Single Carrier Radio Transmission Technology (1xRTT), Evolution-Data Optimized (EVDO) technology, General Packet Radio Service (GPRS) technology, Enhanced Data rates for GSM Evolution (EDGE) technology, third Generation Partnership Project (3GPP) including 3G, fourth generation wireless (4G) networks, Universal Mobile Telecommunications System (UMTS), High Speed Packet Access (HSPA), Worldwide Interoperability for Microwave Access (WiMAX), Long Term Evolution (LTE) standard, others defined by various standard setting organizations, other long range protocols, or other data transfer technology.

[0227] Confidence Interval: As used herein, “confidence interval” means a range of values so defined that there is a specified probability that the value of a given parameter lies within that range of values.

[0228] CpG: As used herein, “CpG” or “cytosine-guanine dinucleotide” refers to a cytosine-phosphate-guanine site within a nucleic acid molecule sequence such that a cytosine molecule is followed by a guanine molecule in a 5’ ^ 3’ direction of the nucleic acid molecule sequence.

[0229] Deoxyribonucleic Acid or Ribonucleic Acid: As used herein, “deoxyribonucleic acid” or “DNA” refers to a natural or modified nucleotide which has a hydrogen group at the 2′- position of the sugar moiety. DNA can include a chain of nucleotides comprising four types of nucleotide bases: adenine (A), thymine (T), cytosine (C), and guanine (G). As used herein, “ribonucleic acid” or “RNA” refers to a natural or modified nucleotide which has a hydroxyl group at the 2′-position of the sugar moiety. RNA can include a chain of nucleotides comprising four types of nucleotides: A, uracil (U), G, and C. As used herein, the term “nucleotide” refers to a natural nucleotide or a modified nucleotide. Certain pairs of nucleotides specifically bind to one another in a complementary fashion (called complementary base pairing). In DNA, adenine (A) pairs with thymine (T) and cytosine (C) pairs with guanine (G). In RNA, adenine (A) pairs with uracil (U) and cytosine (C) pairs with guanine (G). When a first nucleic acid strand binds to a second nucleic acid strand made up of nucleotides that are complementary to those in the first strand, the two strands bind to form a double strand. As used herein, “nucleic acid sequencing data”, “nucleic acid sequencing information”, “sequence information”, “sequence representation”, “nucleic acid sequence”, “nucleotide sequence”, “genomic sequence”, “genetic sequence”,“fragment sequence”, “sequencing read”, or “nucleic acid sequencing read” denotes any information or data that is indicative of the order and identity of the nucleotide bases (e.g., adenine, guanine, cytosine, and thymine or uracil) in a molecule (e.g., a whole genome, whole transcriptome, exome, oligonucleotide, polynucleotide, or fragment) of a nucleic acid such as DNA or RNA. It should be understood that the present teachings contemplate sequence information obtained using all available varieties of techniques, platforms or technologies, including, but not limited to capillary electrophoresis, microarrays, ligation-based systems, polymerase-based systems, hybridization-based systems, direct or indirect nucleotide identification systems, pyrosequencing, ion- or pH-based detection systems, and electronic signature-based systems. Differentially Methylated Region: As used herein, differentially methylated region” refers to a region of DNA, such as a region of a genome, having a detectably different degree of methylation in at least one cell or tissue type relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type; or having a detectably different degree of methylation in at least one cell or tissue type obtained from a subject having a disease or disorder relative to the degree of methylation in the same region of DNA in the same cell or tissue type obtained from a healthy subject. In some embodiments, a differentially methylated region has a detectably higher degree of methylation in at least one cell or tissue type, such as at least one immune cell type, relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type, such as other immune cell types and / or cell types that contribute to cfDNA in healthy individuals, or from the same cell or tissue type from a healthy subject. In one or more additional examples, a differentially methylated region can include a genomic region, such as a genomic region corresponding to immune system function, that has a greater number of methylated nucleic acid molecules in a given sample due to a higher than expected turnover of cells related to the genomic region in an organ caused by the presence of a tumor in the organ. In some embodiments, a differentially methylated region has a detectably lower degree of methylation in at least one cell or tissue type, such as at least one immune cell type, relative to the degree of methylation in the same region of DNA from at least one other cell or tissue type, such as other immune cell types and / or cell types that contribute to cfDNA in healthy individuals, or from the same cell or tissue type from a healthy subject. In some embodiments, a differentially methylated region show no or negligible methylation signal when analyzing cell-free DNA (cfDNA) from healthy individuals (e.g. in blood) but exhibit detectable methylation when analyzing cfDNA from individuals with cancer. Such regions are characterized by low background methylation levels in healthy populations, thereby providing an enhanced contrast that facilitates sensitive detection of tumor-derived DNA. In some embodiments, a differentially methylated region showno or negligible methylation signal for a particular cell or tissue type (e.g., lung tissue) when analyzing cell-free DNA (cfDNA) from healthy individuals (e.g. in blood) but exhibit detectable methylation when analyzing cfDNA from individuals with disease associated with that particular cell or tissue type (e.g., if the tissue type is lung, then the disease can be lung cancer or pulmonary disorder).

[0230] Driver Mutation: As used herein, “driver mutation” means a mutation that drives cancer progression.

[0231] Epigenetic Target Regions: As used herein, “epigenetic target regions” refers to target regions that may show sequence-independent differences in different cell or tissue types (e.g., different types of immune cells) or in neoplastic cells (e.g., tumor cells and cancer cells) relative to normal cells; or that may show sequence- independent differences (i.e., in which there is no change to the nucleotide sequence, e.g., differences in methylation, nucleosome distribution, or other epigenetic features) in DNA, such as cfDNA, from different cell types or from subjects having cancer relative to DNA, such as cfDNA, from healthy subjects, or in cfDNA originating from different cell or tissue types that ordinarily do not substantially contribute to cfDNA (e.g., immune, lung, colon, etc.) relative to background cfDNA (e.g., cfDNA that originated from hematopoietic cells). Examples of sequence-independent changes include, but are not limited to, changes in methylation (increases or decreases), nucleosome distribution, cfDNA fragmentation patterns, CCCTC-binding factor (“CTCF”) binding, transcription start sites (e.g., with respect to any one of more of binding of RNA polymerase components, binding of regulatory proteins, fragmentation characteristics, and nucleosomal distribution), and regulatory protein binding regions. Epigenetic target region sets thus include, but are not limited to, hypermethylation variable target region sets, hypomethylation variable target region sets, and fragmentation variable target region sets, such as CTCF binding sites and transcription start sites. For present purposes, loci susceptible to neoplasia-, tumor-, or cancer-associated focal amplifications and / or gene fusions may also be included in an epigenetic target region set because detection of a change in copy number by sequencing or a fused sequence that maps to more than one locus in a reference genome tends to be more similar to detection of exemplary epigenetic changes discussed above than detection of nucleotide substitutions, insertions, or deletions, e.g., in that the focal amplifications and / or gene fusions can be detected at a relatively shallow depth of sequencing because their detection does not depend on the accuracy of base calls at one or a few individual positions. An epigenetic target region set is a set of two or more epigenetic target regions.

[0232] Immunotherapy: As used herein, “immunotherapy” refers to treatment with one or more agents that act to stimulate the immune system so as to kill or at least to inhibit growth ofcancer cells, and preferably to reduce further growth of the cancer, reduce the size of the cancer and / or eliminate the cancer. Some such agents bind to a target present on cancer cells; some bind to a target present on immune cells and not on cancer cells; some bind to a target present on both cancer cells and immune cells. Such agents include, but are not limited to, checkpoint inhibitors, genetically engineered immune cells and / or antibodies, including natural antibodies and genetically engineered antibodies. Checkpoint inhibitors are inhibitors of pathways of the immune system that maintain self-tolerance and modulate the duration and amplitude of physiological immune responses in peripheral tissues to minimize collateral tissue damage (see, e.g., Pardoll, Nature Reviews Cancer 12, 252–264 (2012)). Example agents include antibodies against any of PD-1, PD-2, PD-L1, PD-L2, CTLA-40, OX40, B7.1, B7He, LAG3, CD137, KIR, CCR5, CD27, or CD40. Other example agents include proinflammatory cytokines, such as IL-1β, IL-6, and TNF-α. Other example agents are T-cells designed to be activated against a tumor, such as T-cells activated by expressing a chimeric antigen receptor (CAR) targeting a tumor antigen and / or cell-surface protein engineered to be recognized by the T-cell.

[0233] Indel: As used herein, “indel” refers to a mutation that involves the insertion or deletion of nucleotides in the genome of a subject.

[0234] Limit of Detection (LoD): As used herein, “limit of detection” means the smallest amount of a substance (e.g., a nucleic acid) in a sample that can be measured by a given assay or analytical approach.

[0235] Machine-Readable Medium: As used herein, “machine-readable medium” refers to a component, device, or other tangible media able to store instructions and data temporarily or permanently and may include, but is not limited to, random-access memory (RAM), read-only memory (ROM), buffer memory, flash memory, optical media, magnetic media, cache memory, other types of storage (e.g., erasable programmable read-only memory (EEPROM)) and / or any suitable combination thereof. The term "machine-readable medium" may be taken to include a single medium or multiple media (e.g., a centralized or distributed database, or associated caches and servers) able to store instructions. The term "machine-readable medium" shall also be taken to include any medium, or combination of multiple media, that is capable of storing instructions (e.g., code) for execution by a machine and that when executed by one or more processors of the machine, cause the machine to perform any one or more of the methodologies described herein. Accordingly, a "machine-readable medium" refers to a single storage apparatus or device, as well as "cloud-based" storage systems or storage networks that include multiple storage apparatus or devices. The term "machine-readable medium" excludes signals per se.

[0236] Maximum MAF: As used herein, “maximum MAF” or “max MAF” refers to the maximum MAF (mutant allele fraction) of all somatic tumor variants in a sample.

[0237] Methylation: As used herein, “methylation” or “DNA methylation” refers to addition of a methyl group to a nucleotide base in a nucleic acid molecule. In some embodiments, methylation refers to addition of a methyl group to a cytosine at a CpG site. In some embodiments, DNA methylation refers to addition of a methyl group to adenine, such as in N6-methyladenine. In some embodiments, DNA methylation is 5-methylation (modification of the 5th carbon of the 6- carbon ring of cytosine). In some embodiments, 5-methylation refers to addition of a methyl group to the 5C position of the cytosine to create 5-methylcytosine (5mC). In some embodiments, methylation comprises a derivative of 5mC. Derivatives of 5mC include, but are not limited to, 5- hydroxymethylcytosine (5-hmC), 5-formylcytosine (5-fC), and 5-caryboxylcytosine (5-caC). In some embodiments, DNA methylation is 3C methylation (modification of the 3rd carbon of the 6- carbon ring of cytosine). In some embodiments, 3C methylation comprises addition of a methyl group to the 3C position of the cytosine to generate 3-methylcytosine (3mC). Methylation can also occur at non CpG sites, for example, methylation can occur at a CpA, CpT, or CpC site. DNA methylation can change the activity of methylated DNA region. For example, when DNA in a promoter region is methylated, transcription of the gene may be repressed. DNA methylation is critical for normal development and abnormality in methylation may disrupt epigenetic regulation. The disruption, e.g., repression, in epigenetic regulation may cause diseases, such as cancer. Promoter methylation in DNA may be indicative of cancer.

[0238] Methylation rate: As used herein, “methylation rate” refers to the probability, likelihood, or percentage that a given base (for example: cytosine residue in a CpG) is methylated on a DNA molecule at a particular genomic region analyzed in the sample. In some embodiments, the methylation rate may be applied to a defined region that comprises one or more potentially methylated bases. In some embodiments, the methylation rate refers to the percentage of CpG residues methylated in a DNA molecule. In some embodiments, the methylation rate refers to the percentage of CpG residues methylated in molecules aligned to particular genomic position or genomic region. Methylation rate can be measured by a variety of methods including, but not limited to, either using bisulfite sequencing (any single base resolution like TAPS, EM-SEQ, etc.) or using partitioning (DNA molecule resolution) , such as DNA methylation partitioning using methyl-binding antibodies or proteins. Methylation rate can be measured in different ways. One estimation can be by counting how many DNA fragments end up in each methylation dependent partition or by counting the number of converted CpGs per fragment in the case of bisulfite sequencing or any other base-level resolution sequencing methods, such as qPCR-basedmethods that can use either converted DNA or methyl-precipitated DNA. In addition, in the case of methylation dependent partitioning, the rate calculation can be normalized using a set of predefined genomic control regions with known methylation state (i.e., positive control regions and / or negative control regions) or spiked- in synthetic DNA with known methylation state, deriving rate-parametrized partition distributions and estimating the rate using a maximum likelihood approach. In various examples, methylation rate can be calculated by dividing or “normalizing” the count of methylated molecules corresponding to one or more genomic regions by the number of molecules present within the genomic control regions. In one or more examples, the methylation rate can be determined by measuring an abundance of sequencing reads that correspond to a portion of a genomic region. The portion of the genomic region can include a number of genomic locations of the genomic region for which at least a threshold number of sequencing reads overlap.

[0239] Methylation Status: As used herein, “methylation status” or “methylation state” can refer to the presence or absence of methyl group on a DNA base (e.g., cytosine) at a particular genomic position in a nucleic acid molecule. It can also refer to the degree of methylation in a nucleic acid sequence (e.g., highly methylated, low methylated, intermediately methylated or unmethylated nucleic acid molecules). The methylation status can also refer to the number of nucleotides methylated in a particular nucleic acid molecule.

[0240] Mutant Allele Fraction: As used herein, “mutant allele fraction”, “mutation dose,” or “MAF” refers to the fraction of nucleic acid molecules harboring an allelic alteration or mutation at a given genomic position in a given sample. MAF is generally expressed as a fraction or a percentage. For example, an MAF can be less than about 0.5, 0.1, 0.05, or 0.01 (i.e., less than about 50%, 10%, 5%, or 1%) of all somatic variants or alleles present at a given locus.

[0241] Mutation: As used herein, “mutation” refers to a variation from a known reference sequence and includes mutations such as, for example, single nucleotide variants (SNVs), copy number variants or variations (CNVs) / aberrations, insertions or deletions (indels), gene fusions, transversions, translocations, frame shifts, duplications, repeat expansions, and epigenetic variants. A mutation can be a germline or somatic mutation. In some examples, a reference sequence for purposes of comparison is a wildtype genomic sequence of the species of the subject providing a test sample, typically the human genome.

[0242] Mutation Count: As used herein, “mutation count” or “mutational count” refers to the number of somatic mutations in a whole genome or exome or targeted regions of a nucleic acid sample.

[0243] Neoplasm: As used herein, the terms “neoplasm” and “tumor” are used interchangeably. They refer to abnormal growth of cells in a subject. A neoplasm or tumor can be benign, potentially malignant, or malignant. A malignant tumor is referred to as a cancer or a cancerous tumor.

[0244] Next Generation Sequencing: As used herein, “next generation sequencing” or “NGS” refers to sequencing technologies having increased throughput as compared to traditional Sanger- and capillary electrophoresis-based approaches, for example, with the ability to generate hundreds of thousands of relatively small sequencing reads at a time. Some examples of next generation sequencing techniques include, but are not limited to, sequencing by synthesis, sequencing by ligation, and sequencing by hybridization.

[0245] Nucleic Acid Tag: As used herein, “nucleic acid tag” refers to a short nucleic acid (e.g., less than about 500 nucleotides, about 100 nucleotides, about 50 nucleotides, or about 10 nucleotides in length), used to distinguish nucleic acids from different samples (e.g., representing a sample index), or different nucleic acid molecules in the same sample (e.g., representing a molecular barcode), of different types, or which have undergone different processing. The nucleic acid tag comprises a predetermined, fixed, non-random, random or semi-random oligonucleotide sequence. Such nucleic acid tags may be used to label different nucleic acid molecules or different nucleic acid samples or sub-samples. Nucleic acid tags can be single-stranded, double-stranded, or at least partially double-stranded. Nucleic acid tags optionally have the same length or varied lengths. Nucleic acid tags can also include double-stranded molecules having one or more blunt- ends, include 5’ or 3’ single-stranded regions (e.g., an overhang), and / or include one or more other single-stranded regions at other locations within a given molecule. Nucleic acid tags can be attached to one end or to both ends of the other nucleic acids (e.g., sample nucleic acids to be amplified and / or sequenced). Nucleic acid tags can be decoded to reveal information such as the sample of origin, form, or processing of a given nucleic acid. For example, nucleic acid tags can also be used to enable pooling and / or parallel processing of multiple samples comprising nucleic acids bearing different molecular barcodes and / or sample indexes in which the nucleic acids are subsequently being deconvolved by detecting (e.g., reading) the nucleic acid tags. Nucleic acid tags can also be referred to as identifiers (e.g., molecular identifier, sample identifier). Additionally, or alternatively, nucleic acid tags can be used as molecular identifiers (e.g., to distinguish between different molecules or amplicons of different parent molecules in the same sample or sub-sample). This includes, for example, uniquely tagging different nucleic acid molecules in a given sample, or non-uniquely tagging such molecules. In the case of non-unique tagging applications, a limited number of tags (i.e., molecular barcodes) may be used to tag each nucleic acid molecule suchthat different molecules can be distinguished based on their endogenous sequence information (for example, start and / or stop positions where they map to a selected reference sequence, a sub-sequence of one or both ends of a sequence, and / or length of a sequence) in combination with at least one molecular barcode. A sufficient number of different molecular barcodes are used such that there is a low probability (e.g., less than about a 10%, less than about a 5%, less than about a 1%, or less than about a 0.1% chance) that any two molecules may have the same endogenous sequence information (e.g., start and / or stop positions, subsequences of one or both ends of a sequence, and / or lengths) and also have the same molecular barcode.

[0246] Polynucleotide: As used herein, “polynucleotide”, “nucleic acid”, “nucleic acid molecule”, “polynucleotide molecule”, or “oligonucleotide” refers to a linear polymer of nucleosides (including deoxyribonucleosides, ribonucleosides, or analogs thereof) joined by internucleosidic linkages. A polynucleotide can comprise at least three nucleosides. Oligonucleotides often range in size from a few monomeric units, e.g., 3-4, to hundreds of monomeric units. Whenever a polynucleotide is represented by a sequence of letters, such as “AGCTG,” it will be understood that the nucleotides are in 5’ ^ 3’ order from left to right and that in the case of DNA, “A” denotes deoxyadenosine, “C” denotes deoxycytidine, “G” denotes deoxyguanosine, and “T” denotes deoxythymidine, unless otherwise noted. The letters A, C, G, and T may be used to refer to the bases themselves, to nucleosides, or to nucleotides comprising the bases, as is standard in the art.

[0247] Probe: As used herein, “probe” refers to a polynucleotide comprising a functionality. The functionality can be a detectable label (fluorescent), a binding moiety (biotin), or a solid support (a magnetically attractable particle or a chip). Probes can include single- stranded DNA / RNA polynucleotides or double stranded DNA polynucleotides that hybridize to target nucleic acid sequences (e.g., SureSelect®probes, Agilent Technologies). Sequence capture using probes generally depends, in part, on the number of consecutive nucleotides in at least a portion of the target nucleic acid sequence that is complementary (or nearly complementary) to the sequence of the probe. In some examples, probes can correspond to driver mutations.

[0248] Processing: As used herein, the terms “processing”, “calculating”, and “comparing” can be used interchangeably. In certain applications, the terms refer to determining a difference, e.g., a difference in number or sequence. For example, gene expression, copy number variation (CNV), indel, and / or single nucleotide variant (SNV) values or sequences can be processed.

[0249] Processor: As used herein, “processor” refers to any circuit or virtual circuit (a physical circuit emulated by logic executing on an actual processor) that manipulates data values according to control signals (e.g., "commands," "op codes," "machine code," etc.) and which produces corresponding output signals that are applied to operate a machine. A processor may, for example, be a CPU, a RISC processor, a CISC processor, a GPU, a DSP, an ASIC, a RFIC or any combination thereof. A processor may further be a multi-core processor having two or more independent processors (sometimes referred to as "cores") that may execute instructions contemporaneously.

[0250] Promoter Region As used herein, “promoter region” refers to a DNA sequence recognized by the synthetic machinery of the cell, or introduced synthetic machinery, required to initiate the specific transcription of a gene.

[0251] Quantitative Measures: As used herein, “quantitative measures” refers to an absolute or relative measure. A quantitative measure can be, without limitation, a number, a statistical measurement (e.g., frequency, mean, median, standard deviation, or quantile), or a degree or a relative quantity (e.g., high, medium, and low). A quantitative measure can be a ratio of two quantitative measures. A quantitative measure can be a linear combination of quantitative measures. A quantitative measure may be a normalized measure.

[0252] Reference Sequence: As used herein, “reference sequence” refers to a known sequence used for purposes of comparison with experimentally determined sequences. For example, a known sequence can be an entire genome, a chromosome, or any segment thereof. A reference sequence can include at least about 20, at least about 50, at least about 100, at least about 200, at least about 250, at least about 300, at least about 350, at least about 400, at least about 450, at least about 500, at least about 1000, or more nucleotides. A reference sequence can align with a single contiguous sequence of a genome or chromosome or can include non- contiguous segments that align with different regions of a genome or chromosome. Example reference sequences, include, for example, human genome reference sequences, such as, hG19 and hG38.

[0253] Sample: As used herein, “sample” means anything capable of being analyzed by the methods and / or systems disclosed herein.

[0254] Sensitivity: As used herein, “sensitivity” means the probability of detecting the presence of a single nucleotide variant, an insertion, and a deletion at a given MAF and coverage and the probability of detecting the presence of a copy number variant at a given tumor fraction and coverage.

[0255] Sequencing: As used herein, “sequencing” refers to any of a number of technologies used to determine the sequence (e.g., the identity and order of monomer units) of a biomolecule, e.g., a nucleic acid such as DNA or RNA. Example sequencing methods include, but are not limited to, targeted sequencing, single molecule real-time sequencing, exon or exome sequencing, intron sequencing, electron microscopy-based sequencing, panel sequencing, transistor-mediated sequencing, direct sequencing, random shotgun sequencing, Sanger dideoxy termination sequencing, whole-genome sequencing, sequencing by hybridization, pyrosequencing, capillary electrophoresis, duplex sequencing, cycle sequencing, single-base extension sequencing, solid-phase sequencing, high-throughput sequencing, massively parallel signature sequencing, emulsion PCR, co-amplification at lower denaturation temperature-PCR (COLD-PCR), multiplex PCR, sequencing by reversible dye terminator, paired-end sequencing, near-term sequencing, exonuclease sequencing, sequencing by ligation, short-read sequencing, single-molecule sequencing, sequencing-by-synthesis, real-time sequencing, reverse-terminator sequencing, nanopore sequencing, 454 sequencing, Solexa Genome Analyzer sequencing, SOLiD™ sequencing, MS-PET sequencing, and a combination thereof. In some implementations, sequencing can be performer by a gene analyzer such as, for example, gene analyzers commercially available from Illumina, Inc., Pacific Biosciences, Inc., or Applied Biosystems / Thermo Fisher Scientific, among many others.

[0256] Single Nucleotide Variant: As used herein, “single nucleotide variant” or “SNV” means a mutation or variation in a single nucleotide that occurs at a specific position in the genome.

[0257] Somatic Mutation: As used herein, “somatic mutation” means a mutation in the genome that occurs after conception. Somatic mutations can occur in any cell of the body except germ cells and accordingly, are not passed on to progeny.

[0258] Specifically binds: As used herein, “specifically binds” in the context of a probe or other oligonucleotide and a target sequence means that under appropriate hybridization conditions, the oligonucleotide or probe hybridizes to its target sequence, or replicates thereof, to form a stable probe:target hybrid, while at the same time formation of stable probe:non-target hybrids is minimized. Thus, a probe hybridizes to a target sequence or replicate thereof to a sufficiently greater extent than to a non-target sequence, to enable capture or detection of the target sequence. Appropriate hybridization conditions are well-known in the art, may be predicted based on sequence composition, or can be determined by using routine testing methods (see, e.g., Sambrook et al., Molecular Cloning, A Laboratory Manual, 2nd ed. (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, 1989) at §§ 1.90-1.91, 7.37-7.57, 9.47-9.51 and 11.47-11.57, particularly §§ 9.50-9.51, 11.12-11.13, 11.45-11.47 and 11.55-11.57, incorporated by reference herein).

[0259] Subject: As used herein, “subject” refers to an animal, such as a mammalian species (e.g., human) or avian (e.g., bird) species, or other organism, such as a plant. More specifically, a subject can be a vertebrate, e.g., a mammal such as a mouse, a primate, a simian or a human. Animals include farm animals (e.g., production cattle, dairy cattle, poultry, horses, pigs, and the like), sport animals, and companion animals (e.g., pets or support animals). A subject can be a healthy individual, an individual that has or is suspected of having a disease or a predisposition to the disease, or an individual that is in need of therapy or suspected of needing therapy. The terms “individual” or “patient” are intended to be interchangeable with “subject.”

[0260] For example, a subject can be an individual who has been diagnosed with having a cancer, is going to receive a cancer therapy, and / or has received at least one cancer therapy. The subject can be in remission of a cancer. As another example, the subject can be an individual who is diagnosed of having an autoimmune disease. As another example, the subject can be a female individual who is pregnant or who is planning on getting pregnant, who may have been diagnosed of or suspected of having a disease, e.g., a cancer, an auto-immune disease.

[0261] Target Region: As used herein, “target region” refers to a genomic locus targeted for identification and / or capture, for example, by using probes (e.g., through sequence complementarity). A “target region set” or “set of target regions” refers to a plurality of genomic loci targeted for identification and / or capture, for example, by using a set of probes (e.g., through sequence complementarity).

[0262] Threshold: As used herein, “threshold” refers to a predetermined value used to characterize experimentally determined values of the same parameter for different samples depending on their relation to the threshold.

[0263] Tumor Fraction: As used herein, “tumor fraction” refers to the estimate of the fraction of nucleic acid molecules derived from a tumor in a given sample. For example, the tumor fraction of a sample can be a measure derived from the max MAF of the sample or pattern of sequencing coverage of the sample or length of the cfDNA fragments in the sample or any other selected feature of the sample. In some instances, the tumor fraction of a sample is equal to the max MAF of the sample.

[0264] Variant: As used herein, a “variant” can be referred to as an allele. A variant is usually presented at a frequency of 50% (0.5) or 100% (1), depending on whether the allele is heterozygous or homozygous. For example, germline variants are inherited and usually have a frequency of 0.5 or 1. Somatic variants; however, are acquired variants and usually have afrequency of < 0.5. Major and minor alleles of a genetic locus refer to nucleic acids harboring the locus in which the locus is occupied by a nucleotide of a reference sequence, and a variant nucleotide different than the reference sequence respectively. Measurements at a locus can take the form of allelic fractions (AFs), which measure the frequency with which an allele is observed in a sample. DETAILED DESCRIPTION

[0265] Cancer is usually caused by the accumulation of mutations within genes of an individual's cells, at least some of which result in improperly regulated cell division. Such mutations can include single nucleotide variations (SNVs), gene fusions, insertions, transversions, translocations, and inversions. These mutations can also include copy number variations that correspond to an increase or a decrease in the number of copies of a gene within a tumor genome relative to an individual’s noncancerous cells. An extent of mutations present in cell-free nucleic acids and an amount of mutated cell-free nucleic acids of a sample can be used as biomarkers to determine tumor progression, predict patient outcome, and refine treatment choices. In various examples, the extent of mutations present in cell-free nucleic acids can be indicated by tumor cells copy number and tumor fraction for a given sample.

[0266] Additionally, cancer can be indicated by non-sequence modifications, such as methylation. Examples of methylation changes in cancer include local gains of DNA methylation in the CpG islands at the TSS of genes involved in normal growth control, DNA repair, cell cycle regulation, and / or cell differentiation. This increased amount of methylation can be associated with an aberrant loss of transcriptional capacity of involved genes and occurs at least as frequently as point mutations and deletions as a cause of altered gene expression.

[0267] Thus, DNA methylation profiling can be used to detect aberrant methylation in DNA of a sample. The DNA can correspond to certain genomic regions (“differentially methylated regions” or “DMRs”) that are normally hypermethylated or hypomethylated in a given sample type (e.g., cfDNA from the bloodstream) but which may show an abnormal degree of methylation that correlates to a neoplasm or cancer, e.g., because of unusually increased contributions of tissues to the type of sample (e.g., due to increased shedding of DNA in or around the neoplasm or cancer) and / or from extents of methylation of the genome that are altered during development or that are perturbed by disease, for example, cancer or any cancer-associated disease.

[0268] The methods and systems described herein are directed to identifying input data for a machine learning classification model, where the machine learning classification model determines indications of the presence or absence with respect to one or more tumor-relatedbiological conditions in subjects. In one or more examples, the input data for the machine learning classification model can be derived from sequence representations corresponding to nucleic acid molecules included in samples obtained from subjects. In various examples, some sequence representations can be more predictive of the tumor-related biological condition than other sequence representations. Implementations described herein are directed to generating a framework for identifying the sequence representations that have at least a threshold probability of being related to samples obtained from individuals in which the tumor-related biological condition is present. The framework for identifying the sequence representations having at least the threshold probability can be determined by a computational analysis of features of training sequence representations. The features can include genomic position, sequence representation length, number of cytosine-guanine dinucleotides (CpGs) present in the sequence representations, methylation information related to the sequence representations, sequencing related information associated with generating the sequence representations from a sample, one or more combinations thereof, and the like.

[0269] In various examples, at least one of the training process for the machine learning classification model or the performance of the machine learning classification model can be influenced by the quality of the training data and the input data provided to the machine learning classification model. By determining the framework for identifying sequence representations that are more likely to be predictive of a tumor-related biological condition, the amount of computational resources used during the training of the machine learning classification model can be decreased. Additionally, by applying a framework to determine sequence representations that are more likely to be predictive of a tumor-related biological condition and that are used to determine input data for the machine learning classification model, the performance of the machine learning classification model can be improved by improving the accuracy of the classifications determined by the machine learning classification model for samples.

[0270] Figure 1 is a diagrammatic representation of an example computational architecture 100 that determines sequence representations to be analyzed by a machine learning classification model to detect a tumor-related biological condition, according to one or more example implementations. The tumor-related biological condition can correspond to one or more diseases. In one or more examples, the disease under consideration is a type of cancer. Non- limiting examples of such cancers include biliary tract cancer, bladder cancer, transitional cell carcinoma, urothelial carcinoma, brain cancer, gliomas, astrocytomas, breast carcinoma, metaplastic carcinoma, cervical cancer, cervical squamous cell carcinoma, rectal cancer, colorectal carcinoma, colon cancer, hereditary nonpolyposis colorectal cancer, colorectaladenocarcinomas, gastrointestinal stromal tumors (GISTs), endometrial carcinoma, endometrial stromal sarcomas, esophageal cancer, esophageal squamous cell carcinoma, esophageal adenocarcinoma, ocular melanoma, uveal melanoma, gallbladder carcinomas, gallbladder adenocarcinoma, renal cell carcinoma, clear cell renal cell carcinoma, transitional cell carcinoma, urothelial carcinomas, Wilms tumor, leukemia, acute lymphocytic leukemia (ALL), acute myeloid leukemia (AML), chronic lymphocytic (CLL), chronic myeloid (CML), chronic myelomonocytic (CMML), liver cancer, liver carcinoma, hepatoma, hepatocellular carcinoma, cholangiocarcinoma, hepatoblastoma, lung cancer, non-small cell lung cancer (NSCLC), mesothelioma, B-cell lymphomas, non-Hodgkin lymphoma, diffuse large B-cell lymphoma, Mantle cell lymphoma, T- cell lymphomas, non-Hodgkin lymphoma, precursor T-lymphoblastic lymphoma / leukemia, peripheral T-cell lymphomas, multiple myeloma, nasopharyngeal carcinoma (NPC), neuroblastoma, oropharyngeal cancer, oral cavity squamous cell carcinomas, osteosarcoma, ovarian carcinoma, pancreatic cancer, pancreatic ductal adenocarcinoma, pseudopapillary neoplasms, acinar cell carcinomas, prostate cancer, prostate adenocarcinoma, skin cancer, melanoma, malignant melanoma, cutaneous melanoma, small intestine carcinomas, stomach cancer, gastric carcinoma, gastrointestinal stromal tumor (GIST), uterine cancer, or uterine sarcoma. In one or more additional examples, the tumor-related biological condition can include minimal residual disease (MRD). In still other examples, the tumor-related biological condition can correspond to one or more immunological disorders. In one or more further examples, the computational architecture 100 can be implemented to monitor subjects with regard to tumor formation or tumor remission over a period of time, such as one or more months or one or more years. The computational architecture 100 can also be implemented to monitor subjects with regard to tumor formation or tumor remission after receiving one or more treatments for the tumor- related biological condition.

[0271] The example computational architecture 100 can include a computational system 102. The computational system 102 can include one or more computing devices 104. The one or more computing devices 104 can include at least one of one or more desktop computing devices, one or more mobile computing devices, or one or more server computing device. In various examples, at least a portion of the one or more computing devices 104 can be included in a remote computing environment, such as a cloud computing environment. In one or more examples, the operations executed by the computational system 102 can be performed by, controlled by, and / or maintained by a single organization. In one or more additional examples, the operations executed by the computational system 102 can be performed by, controlled by, and / or maintained by multiple organizations.

[0272] The computational architecture 100 can obtain sequencing data 106 that is derived from one or more samples 108 obtained from one or more subjects 110. In one or more examples, the sequencing data 106 can include a number of sequence representations 112. The sequence representations 112 can include alphanumeric representations of nucleic acid molecules derived from the one or more samples 108. For example, the sequence representations 112 can include, for individual nucleic acids, data that corresponds to a string of letters that represents the respective chains of nucleotides that correspond to the individual nucleic acid molecules derived from the one or more samples 108. The sequencing data 106 can be stored in one or more data files. To illustrate, the sequencing data 106 be stored in a FASTQ file that comprises a text-based sequencing data file format storing raw sequence data and quality scores. In one or more additional examples, the sequencing data 106 can be stored in a data file according to a binary base call (BCL) sequence file format. In one or more further examples, the sequencing data 106 can be stored in a BAM file. In one or more examples, the sequencing data 106 can comprise at least about one gigabyte (GB), at least about 2 GB, at least about 3GB, at least about 4 GB, at least about 5 GB, at least about 8 GB, or at least about 10 GB.

[0273] In one or more examples, the sequence representations 112 can be generated as part of one or more sequencing processes performed with respect to nucleic acids obtained from one or more samples. In one or more illustrative examples, the sequence representations 112 can correspond to one or mores sequencing reads produced by one or more sequencing processes that are applied to the one or more samples 108. In one or more additional illustrative examples, the sequence representations 112 can correspond to nucleic acid molecules present in the one or more samples 108. In various examples, individual nucleic acid molecules derived from the one or more samples 108 can correspond to multiple sequence representations 112 as a result of the amplification of the individual nucleic acid molecules derived from the one or more samples 108. In situations where amplification is not performed with respect to nucleic acid molecules derived from the one or more samples 108, individual nucleic acid molecules derived can correspond to a single sequence representation 112 as a result of the absence of amplification of the individual nucleic acid molecules.

[0274] When multiple sequence representations 112 correspond to a single nucleic acid molecule derived from the one or more samples 108, a number of groups can be generated from the sequence representations 112 with each group corresponding to a single nucleic acid molecule derived from the one or more samples 108. In various examples, the groups of the sequence representations 112 that correspond to a single nucleic acid molecule can be referred to herein as “families.” In at least some examples, families of sequence representations 112 cancorrespond to a unique cell-free nucleic acid molecule present in a sample 108. In at least some examples, start and stop positions with respect to a reference sequence having a common molecular barcode can be used to determine groups of the sequence representations 112 that correspond to individual nucleic acid molecules. In one or more illustrative examples, an individual sequence representation 112 that represents a family of sequence representations and that corresponds to a single nucleic acid molecule derived from the one or more samples 108 can be referred to herein as a “consensus sequence representation.”

[0275] The one or more samples 108 that are used to derive the sequence representations 112 can include one or more biological fluid samples obtained from one or more subjects. In one or more examples, the one or more samples 108 can include one or more blood- based samples obtained from one or more subjects. In one or more additional examples, the one or more samples 108 can include one or more plasma samples obtained from the one or more subjects. In still other examples, the one or more samples 108 can include one or more tissue samples obtained from one or more subjects. In various examples, the one or more samples 108 can include cell-free nucleic acids. To illustrate, the one or more samples 108 can include cell- free DNA.

[0276] The one or more subjects 110 providing the one or more samples 108 used to generate the sequence representations 112 can include one or more mammals. In one or more additional examples, the one or more subjects 110 can include one or more humans. In one or more further illustrative examples, the one or more subjects 110 can include one or more non- human mammals.

[0277] The sequencing data 106 can include position data 114 that corresponds to the sequence representations 112. The position data 114 can indicate start positions and stop positions for individual sequence representations 112 in relation to positions of a reference sequence. In one or more additional examples, the position data 114 can indicate an offset of at least one of a start position or a stop position of the sequence representations 112 in relation to at least one of a start position or a stop position of a genomic region. In one or more further examples, the position data 114 can indicate a chromosome that is corresponding to the sequence representations 112. In at least some examples, the position data 114 can indicate that sequence representations 112 are aligned with one or more genomic regions. In various examples, the sequence representations 112 can be aligned with a genomic region such that start positions and stop positions of the individual sequence representations 112 include or are within the start position and the stop position of the genomic region. The one or more genomic regions can include a classification region that can include one or more mutations in subjects in which atumor-related biological condition is present. The one or more genomic regions can also include a target region that comprises a number of genomic locations that are enriched using probes of one or more diagnostic tests. The one or more diagnostic tests can include one or more assays performed to detect subjects in which one or more tumor-related biological conditions are present.

[0278] In various examples, the position data 114 can be determined by aligning the sequence representations 112 with a reference sequence. In one or more illustrative examples, individual sequence representations 112 can be aligned with a reference sequence by determining an amount of homology between individual nucleotides of the sequence representations 112 and positions of the reference sequence. The amount of homology between a given sequence representation 112 and a portion of a reference sequence can be determined using BLAST programs (basic local alignment search tools) and PowerBLAST programs (Altschul et al., J. Mol. Biol., 1990, 215, 403-410; Zhang and Madden, Genome Res., 1997, 7, 649-656).The amount of homology between a sequence representation 112 and a portion of the reference sequence can also be determined using a Burrows-Wheeler aligner (Li, H., & Durbin, R. (2009). Fast and accurate short read alignment with Burrows–Wheeler transform. Bioinformatics, 25(14), 1754–1760). The Burrows-Wheeler aligner can implement a Smith-Waterman algorithm during a full alignment operation (T.F. Smith, M.S. Waterman, Identification of common molecular subsequences, Journal of Molecular Biology, Volume 147, Issue 1, 1981, Pages 195-197.

[0279] In at least some examples, the sequencing data 106 can include methylation data 116. The methylation data 116 can be determined by one or more nucleobase methylation state detection processes. In one or more examples, the methylation data 116 can indicate modified nucleotides that include one or more methyl groups that are not present in unmodified forms of the nucleotides. In one or more illustrative examples, the methylation data 116 can indicate modified cytosines. That is, in various examples, the methylation data 116 can indicate positions of nucleic acid molecules derived from one or more samples where at least one of a 5- methylcytosine and / or a 5-hydroxymethylcytosine is located. For example, the methylation data 116 can indicate discrete, individual positions of individual nucleic acid molecules derived from one or more samples that include at least one of a 5-methylcytosine and / or a 5- hydroxymethylcytosine. In one or more additional examples, the methylation data 116 can indicate a group of positions of individual nucleic acid molecules derived from one or more samples that include at least one of a 5-methylcytosine and / or a 5-hydroxymethylcytosine.

[0280] The computational system 102, at 118, can generate quantitative measures for the sample 118 based on the sequencing data 106. For example, the quantitative measures can indicate a probability of a sequence representation 112 being derived from a subject in which atumor-related biological condition is present. In one or more examples, the computational system 102 can generate quantitative measures for individual sequence representations 112 that include logit values indicating a probabilities that the individual sequence representations are derived from a subject in which the tumor-related biological condition is present. In various examples, the computational system 102 can generate quantitative measures data 120 that includes at least one of one or more values specifying a probability of a tumor-related biological condition being present in a subject, one or more logit values indicating a probability of a tumor-related biological condition being present in a subject, or other values related to the probability of a tumor-related biological condition being present in a subject.

[0281] The computational system 102 can generate the quantitative measures based on one or more characteristics of the sequence representations 112. In one or more examples, the computational system 102 can determine lengths of the sequence representations 112 based on the sequencing data 106. For example, the computational system 102 can determine a number of nucleotides present in the individual sequence representations 112 based on start positions and stop positions of the individual sequence representations 112 indicated by the position data 114. Additionally, the computational system 102 can determine a number of cytosine-guanine dinucleotides (CpGs) present in the individual sequence representations 112 based on the sequencing data 106. In one or more further examples, the computational system 102 can determine a number of methylated CpGs present in the individual sequence representations 112 based on the methylation data 116. The computational system 102 can also determine genomic regions corresponding to individual sequence representations 112 based on the position data 114 by determining that the individual sequence representations 112 are aligned with one or more genomic regions.

[0282] In still other examples, the computational system 102 can determine a number of sequencing reads that correspond to a single nucleic acid molecule included in the sample 108. In at least some examples, the number of sequencing reads that correspond to a single nucleic acid molecule included in the sample 108 can be referred to herein as a “family size” of the nucleic acid molecule. In one or more additional examples, the computational system 102 can determine a sequencing depth in relation to the sequence representations 112. To illustrate, the computational system 102 can determine sequencing depth by determining a number of sequencing reads corresponding to a nucleic acid molecule included in the sample in relation to a number of sequencing reads corresponding to one or more control regions of a reference sequence.

[0283] Further, the computational system 102 can determine a methylation partition for the sequence representations 112. In various examples, nucleic acid molecules included in the sample 108 can be separated according to a strength of binding to a protein, methyl binding domain (MBD). The strength of binding to MBD can correspond to a number of methylated CpGs present in the individual nucleic acid molecules. The methylation partitions can correspond to different amounts of binding strength to MBD. For example, a first methylation partition can have relatively weak binding to MBD and correspond to relatively few methylated CpGs being present in the corresponding nucleic acid molecules. A second methylation partition can have moderate binding to MBD and correspond to an intermediate number of methylated CpGs being present in the corresponding nucleic acid molecules. In addition, a third methylation partition can have relatively high binding to MBD and correspond to a relatively high number of methylated CpGs being present in the corresponding nucleic acid molecules. The methylation partition for the sequence representations 112 can be included in the methylation data 116.

[0284] In at least some examples, the computational system 102 can also determine a number of restriction enzyme cut sites for the sequence representations 112. In one or more examples, one or more restriction enzymes can be applied to the nucleic acid molecules included in the sample 108 after the nucleic acid molecules have been arranged according to methylation partitions. In these scenarios, the methylation data 116 can indicate the number of restriction enzyme cut sites for a given sequence representation 112.

[0285] The computational system 102 can perform a computational analysis using the features determined from the sequencing data 106 to determine the quantitative measures. That is, the computational system 102 can perform a computational analysis to determine a set of features that can be used to determine an indication of a probability of a sequence representation being derived from a subject in which a tumor-related biological condition is present. For example, the computational system 102 can perform one or more machine learning feature selection processes and / or one or more machine learning feature extraction processes to determine characteristics of the sequence representations 112 to use to determine the quantitative measures. In one or more illustrative example, the computational system 102 can perform a computational analysis with respect to a plurality of characteristics of the sequence representations selected from two or more of genomic region, sequence length, number of CpGs, number of methylated CpGs, family size, methylation partition, sequencing depth, or number of MSRE cut sites. In response to determining a set of features to be used in generating the quantitative measures, the computational system 102 can generate the quantitative measures forthe sequence representations 112 based on the values for the set of features for the individual sequence representations.

[0286] In one or more illustrative examples, thousands, tens of thousands, up to millions of sequence representations 112 can be derived from a single sample 108 and the computational system 102 can generate a number of quantitative measures that corresponds to the number of sequence representations 112 derived from the sample 108. Thus, the computational system 102 can generate thousands, tens of thousands, hundreds of thousands, up to millions of quantitative measures based on the sequencing data 106. In at least some examples, the computational system 102 can determine the quantitative measures based on the sequencing data 106 using a logistic regression model. In one or more additional examples, the computational system 102 can determine the quantitative measures based on the sequencing data 106 using a random forests model. In one or more further examples, the computational system 102 can determine the quantitative measures based on the sequencing data 106 using a boosted tree algorithm. In still other examples, the computational system 102 can determine the quantitative measures based on the sequencing data 106 using a Naïve Bayes classifier. In various additional examples, the computational system 102 can determine the quantitative measures using a support vector machine.

[0287] The computational system 102 can determine, at 122, sequence representation data 124 to provide to a machine learning classification model. For example, the computational system 102 can perform a computational analysis with respect to the quantitative measures data 120 to determine the quantitative measures values and / or modified quantitative measures values to provide as input to a machine learning classification model. The machine learning classification model can determine one or more classifications corresponding to the sequence representations 112 in relation to the presence or absence of a tumor-related biological condition in the subject 110.

[0288] In one or more examples, the computational system 102 can implement a data identification framework 126 to determine the sequence representation data 124. In one or more illustrative examples, the data identification framework 126 can include one or more threshold values that the computational system 102 can apply to the quantitative measures data 120. For example, the data identification framework 126 can include at least one of a minimum probability value or a minimum logit value in relation to the presence of a tumor-related biological condition in a subject. In these scenarios, the computational system 102 can perform a computational analysis using probability values and / or logit values included in the quantitative measures data120 for individual sequence representations 112 in relation to the minimum threshold values of the data identification framework 126.

[0289] To illustrate, the computational system 102 can determine that first quantitative measures for first sequence representations 128 have at least a minimum threshold value and that second quantitative measures for second sequence representations 130 are less than the minimum threshold value. In these scenarios, the sequence representation data 124 can be determined based on information related to the first sequence representations 128. In one or more illustrative examples, the computational system 102 can generate the sequence representation data 124 such that the sequence representation data 124 includes a portion of the quantitative measures data 120 corresponding to the first sequence representations 128. For example, sequence representation data 124 can include at least one of probabilities or logit values included in the quantitative measures data 120 that correspond to the first sequence representations 128.

[0290] In one or more additional examples, the computational system 102 can determine one or more additional quantitative measures corresponding to the first sequence representations 128. In various examples, the computational system 102 can determine a number of the first sequence representations 128 that corresponds to one or more genomic regions. In these instances, the sequence representation data 124 can indicate genomic region quantitative measures that correspond to a number of the sequence representations that correspond to individual genomic regions. In at least some examples, the individual genomic regions can include at least one of one or more classification regions, one or more target regions, one or more differentially methylated regions, or one or more control regions that correspond to a reference sequence. In one or more further examples, the data identification framework 126 can indicate weightings to apply to quantitative measures of the sequence representations 112 based on the values of the quantitative measures. For example, the data identification framework 126 can comprise a scale indicating that sequence representations 112 having higher quantitative measure values have a higher weighting that sequence representations 112 having lower quantitative measure values. In this way, the computational system 102 can determine genomic region quantitative measures using a weighted count of the sequence representations that are aligned with individual genomic regions.

[0291] The data identification framework 126 can be produced by a training process 132. In one or more examples, the training process 132 can include a computational analysis of a first group of training sequence representations derived from samples of subjects in which one or more tumor-related biological conditions are present and a second group of training sequence representations derived from samples of subjects in which a tumor-related biological condition isnot present. In one or more examples, the training process 132 can determine one or more threshold values for the data identification framework 126, where the one or more threshold values can be applied to determine the sequence representations 112 used to generate the sequence representation data 124. The training process 132 can also determine a weightings scale for the data identification framework 126, where the computational system 102 can apply the weightings scale to determine the sequence representation data 124.

[0292] The computational system 102 can, at 134, execute a machine learning classification model based on the sequence representation data 124 to generate a tumor indication 136. In one or more examples, the computational system 102 can execute a machine learning classification model that includes a random forests model to generate the tumor indication 136. In one or more additional examples, the computational system 102 can execute a machine learning classification model that includes a multilayer perceptron model to generate the tumor indication 136. In one or more further examples, the computational system 102 can execute a machine learning classification model that includes a boosted tree model to generate the tumor indication 136. In still other examples, the computational system 102 can execute a machine learning classification model that includes an artificial neural network to generate the tumor indication 136.

[0293] The sequence representation data 124 can include input data for the machine learning classification model. In one or more examples, the sequence representation data 124 can include vector input for the machine learning classification model. In these examples, the sequence representation data 124 can include a vector that indicates quantitative measures for the sequence representations 112 that correspond to individual genomic regions. In one or more illustrative examples, the vector input can indicate counts, normalized counts, or weighted counts of sequence representations 112 that satisfy criteria of the data identification framework 126 that correspond to one or more genomic regions. In still other examples, the sequence representation data 124 can include at least one of a vector or a table indicating values of characteristics of the sequence representations 112 that satisfied one or more criteria of the data identification framework 126. The characteristics indicated by the vector and / or table input to the machine learning classification model can include at least one of genomic position, sequence representation length, number of CpGs present in the sequence representations, number of methylated CpGs present in the sequence representations, sequencing depth, family size, methylation partition, or number of restriction enzyme cut sites.

[0294] Additionally, the sequence representation data 124 can include probability maps with respect to a number of features of the sequence representations 112 that satisfy one or morecriteria of the data identification framework 126. To illustrate, the sequence representation data 124 can include probability maps indicating the probability of a tumor-related biological condition being present in the subject 110 for sequence representations having a given genomic location and a given value for a feature of the sequence representation. In one or more illustrative examples, the probability maps can indicate a probability of a sequence representation corresponding to the subject 110 having a tumor-related biological condition where the sequence representation is aligned with a specified genomic region and has a given length, a given number of CpGs, a given number of methylated CpGs, a given family size, and the like. In at least some examples, the sequence representation data 124 can include multiple probability maps with individual probability maps corresponding to different sequence representation features. In these instances, the sequence representation data 124 can include a first probability map indicating first probabilities of sequence representations aligned with a number of genomic regions and having individual values for sequence representation length corresponding to a tumor-related biological condition being present in the subject 110 and a second probability map indicating second probabilities of sequence representations aligned with the number of genomic regions and having individual values for the number of CpGs included in the sequence representations corresponding to a tumor-related biological condition being present in the subject 110.

[0295] In various examples, the machine learning classification model can be produced using a training process. In at least some examples, the training process for the machine learning classification model can include providing a labeled training data set. The labeled training data set can include first training data derived from subjects in which one or more tumor-related biological conditions are present and second training data derived from subjects in which a tumor- related biological condition is not present. In at least some examples, the training data for the machine learning classification model can include quantitative measures for a number of genomic regions. The number of genomic regions can include at least one of classification regions, differentially methylated regions, or target regions that are enriched as part of a diagnostic assay for the detection of the one or more tumor-related biological conditions. The training process for the machine learning classification model can include fitting features included in the training data to the labels of the training data and determining a number of hyperparameters of the machine learning classification model. In one or more examples, the training process for the machine learning classification model can include fitting model weights to achieve relatively high prediction accuracy for determining a tumor indication based on the training data features and determining optimal values for the hyperparameters to also achieve relatively high prediction accuracy for determining the tumor indication.

[0296] In one or more illustrative examples, the training process for the machine learning classification model can implement L1 regularization techniques or L2 regularization techniques to avoid or minimize overfitting the machine learning classification model. The training process for the machine learning classification model can also include performing an early stopping of the training process in relation to the number of iterations performed to train the machine learning classification model. In one or more additional illustrative examples, based on a computational analysis of the output of the machine learning classification model for training data in relation to validation data, the computational system 102 can determine that an accuracy of the machine learning classification model decrease after a specified number of iterations of the training process. In these scenarios, the hyperparameters of the machine learning classification model can include the hyperparameters that are determined at or before the specified number of iterations when the accuracy of the machine learning classification model decreases. In one or more examples, the optimal number of iterations to grain the machine learning classification model can correspond to an optimal value of the parameter lambda. In at least some examples, the lambda parameter can correspond to a weight of at least one of the L1 or L2 regularization penalty.

[0297] The tumor indication 136 can include an indication that one or more tumor-related biological conditions are present in the subject 110 or are absent from the subject 110. In one or more additional examples, the tumor indication 136 can include a probability of one or more tumor- related biological conditions being present in the subject 110. In one or more further examples, the tumor indication 136 can include a tumor fraction for the sample 108. In still other examples, the tumor indication 136 can include a tumor burden for the sample 108. In various examples, the tumor indication 136 can indicate reoccurrence of a tumor-related biological condition for the subject 110. Additionally, the tumor indication 136 can indicate a responsiveness of the subject 110 to one or more treatments for one or more tumor-related biological conditions. Further, the tumor indication 136 can indicate a progression of one or more tumor-related biological conditions with respect to the subject 110. Also, the tumor indication 136 can indicate a regression of one or more tumor-related biological conditions with respect to the subject 110. In one or more scenarios, the tumor indication 136 can indicate a tissue of origin for one or more tumor-related biological conditions with respect to the subject 110.

[0298] Although the illustrative example of Figure 1 includes a given number of sequence representations 112 and shows a number of sequence representations in the first sequence representations 128 and the second sequence representations 130, the number of sequence representations 112 can be greater than the illustrative examples shown in Figure 1. For example,the sequence representations 112 can include hundreds, thousands, up to millions of sequence representations that undergo a computational analysis by the computational system 102.

[0299] In various examples, the one or more nucleobase methylation state detection processes that are used to generate at least one of the position data 114 or the methylation data 116 can include one or more chemical processes and / or biochemical processes that impact a first type of nucleotide differently than a second type of nucleotide. For example, the one or more nucleobase methylation state detection processes implemented to generate at least one of the position data 114 or the methylation data 116 can include one or more reactions that cause at least one atomic and / or molecular moiety of the first type of nucleotide to be modified in a manner that is different from the manner in which the one or more reactions affect the second type of nucleotide. In one or more examples, the impact of the one or more nucleobase methylation state detection processes on a given type of nucleotide can be based on one or more previous modifications to the given type of nucleotide in relation to an unmodified form of the given type of nucleotide. That is, in various examples, a molecule corresponding to a given type of nucleotide may have been modified before being subjected to the one or more nucleobase methylation state detection processes. To illustrate, before being subjected to the one or more nucleobase methylation state detection processes, nucleotides of nucleic acid molecules derived from one or more samples can be modified due to mutations caused by the presence of a tumor in a subject. In at least some examples, the one or more nucleobase methylation state detection processes can modify the first type of nucleotide or the second type of nucleotide such that the nucleobase pairing of the first type of nucleotide or the second type of nucleotide is altered.

[0300] In one or more illustrative examples, the one or more nucleobase methylation state detection processes implemented to generate at least one of the position data 114 or the methylation data 116 can be performed on nucleic acid molecules included in one or more samples 108 used to generate the number of sequence representations 112. The one or more nucleobase methylation state detection processes can modify a first type of nucleotide of the nucleic acid molecules in a first manner and one or more additional types of nucleotides of the nucleic acid molecules in a second manner. To illustrate, the one or more nucleobase methylation state detection processes can modify at least one of cytosines, guanines, thiamines, or adenines differently than at least one other of cytosines, guanines, thiamines, or adenines. In at least some examples, the one or more nucleobase methylation state detection processes can modify cytosines differently than guanines, thiamines, or adenines. In various examples, the one or more nucleobase methylation state detection processes can modify cytosines such that the modified cytosines no longer pair with guanines. For example, the one or more nucleobase methylationstate detection processes can convert cytosines of the nucleic acid molecules included in one or more samples to uracils. In still other examples, the one or more nucleobase methylation state detection processes may not modify cytosines that were methylated prior to being subjected to the one or more nucleobase methylation state detection processes. In one or more examples, the one or more nucleobase methylation state detection processes may not modify 5-methylcytosines and / or 5-hydroxymethylcytosines of nucleic acid molecules derived from one or more samples In this way, the one or more nucleobase state detection processes can be used to differentiate cytosines that have been previously modified to include a 5-methyl group versus previously unmodified cytosines.

[0301] In one or more examples, the one or more nucleobase methylation state detection processes can include at least one of sodium bisulfite conversion and sequencing, Tet-assisted bisulfite sequencing (TAB-Seq), differential enzymatic cleavage, one or more single molecule sequencing methods, such as nanopore DNA sequencing, oxidative bisulfite (Ox-BS) conversion, APOBEC-coupled epigenetic (ACE) conversion, or direct methylation sequencing (DM-Seq).

[0302] In one or more additional examples, the one or more nucleobase methylation state detection processes can include one or more processes that separate nucleic acid molecules based on amounts of nucleotides of the nucleic acid molecules that have been previously modified. For example, the one or more nucleobase methylation state detection processes can determine a methylation rate for one or more regions of the nucleic acid molecules derived from one or more samples. In various examples, the one or more nucleobase methylation state detection processes can separate nucleic acid molecules included in one or more samples based on amounts of methylated cytosines included in CG regions of individual nucleic acid molecules. To illustrate, the one or more nucleobase methylation state detection processes can separate the nucleic acid molecules derived from one or more samples into a plurality of groups of nucleic acid molecules with individual groups of nucleic acid molecules corresponding to respective amounts of methylated cytosines of the nucleic acid molecules. The one or more nucleobase methylation state detection processes can include at least one of partitioning of nucleic acid molecules included in the one or more samples based on a strength of binding of the individual nucleic acid molecules to methyl binding domain (MBD) and, optionally, treatment with methylation sensitive restriction enzyme (MSRE) and / or methylation dependent restriction enzyme (MDRE). In various examples, a strength of binding of nucleic acid molecules to MBD can be determined by subjecting the nucleic acids to a series of washes having different concentrations of MBD.

[0303] In at least some examples, the nucleobase methylation state detection processes can include one or more sequencing processes. For example, the nucleobase methylation statedetection processes can include whole genome bisulfite sequencing, reduced representation bisulfite sequencing, targeted bisulfite sequencing, extended-representation bisulfite sequencing, or one or more combinations thereof. In one or more illustrative examples, whole genomic bisulfite sequencing can be performed according to the techniques described in T. Gong et al., “Analysis and performance assessment of the whole genome bisulfite sequencing data workflow: currently available tools and a practical guide to advance DNA methylation studies,” Small Methods, 6:e2101251, 2022. In one or more additional illustrative examples, reduced representation bisulfite sequencing can be pexformed according to techniques described in Meissner, A., Gnirke, A., Bell, G.W., Ramsahoye, B., Lander, E.S., and Jaenisch, R. (2005). Reduced representation bisulfite sequencing for comparative high-resolution DNA methylation analysis. Nucleic acids research 33, 5868-5877. In one or more further illustrative examples, targeted bisulfite sequencing can be performed according to techniques described in D.A. Moser et al., “Targeted bisulfite sequencing: A novel tool for the assessment of DNA methylation with high sensitivity and increased coverage,” Psychoneuroendocrinology, 120:1-8, 2020 and / or E. Leitão et al., “Locus- specific DNA methylation analysis by targeted deep bisulfite sequencing,” Methods Mol Biol, 1767:351-66, 2018. In still further illustrative examples, extended-representation bisulfite sequencing can be performed according to techniques described in Shareef, S.J., Bevill, S.M., Raman, A.T. et al. Extended-representation bisulfite sequencing of gene regulatory elements in multiplexed samples and single cells. Nat Biotechnol 39, 1086–1094 (2021).

[0304] Figure 2 is a diagrammatic representation of an example process 200 to apply one or more criteria to identify a group of sequence representations to be analyzed by a machine learning classification model to detect a tumor-related biological condition, according to one or more example implementations. In various examples, the process 200 can be performed by the computational system 102 described with respect to Figure 1. At 202, the process 200 can include generating sequencing data based on performing one or more sequencing processes with respect to one or more samples 204. The sequencing data can include position data indicating the position of sequence representations derived from the sample 204 that are aligned with one or more genomic regions of a reference sequence. The sequencing data can also include methylation data that indicates methylated cytosines present in nucleic acid molecules included in the sample 204. The one or more samples 204 can be obtained from a subject 206 that is undergoing one or more diagnostic tests to determine that a tumor-related biological condition is present or absent in the subject 206.

[0305] Additionally, the process 200 can include, at 208, determining first quantitative measures based on the sequencing data. The quantitative measures can be determined forindividual sequence representations included in the sequencing data. The quantitative measures can be determined for the individual sequence representations based on at least one of sequence representation length, number of CpGs included in the individual sequence representation, number of methylated CpGs included in the individual sequence representation, methylation partition corresponding to the individual sequence representation, genomic position of the individual sequence representation, sample depth, family size for the individual sequence representation, or number of restriction enzyme cut sites for the individual sequence representation. In one or more examples, the quantitative measures for individual sequence representations can include probability values relating to the individual sequence representation being derived from a subject in which a tumor-related biological condition is present. In one or more additional examples, the quantitative measures for individual sequence representations can include logit values corresponding to a probability of the individual sequence representation being derived from a subject in which a tumor-related biological condition is present.

[0306] Further, the process 200 can include, at 210, applying a data identification framework to the first quantitative measures. In the illustrative example of Figure 2, the data identification framework can include a threshold value 212. In various examples, the threshold value 212 can indicate a minimum probability value to determine a set of sequence representations to use to identify data to be provided to a machine learning classification model. In one or more additional examples, the threshold value 212 can indicate a minimum probit value determine a set of sequence representations to use to identify data to be provided to a machine learning classification model. In various examples, the threshold value 212 can be applied to determine a first group of sequence representations 214 having quantitative measures with values at or above the threshold value 212 and a second group of sequence representations 216 with values below the threshold value 212.

[0307] Information corresponding to the first group of sequence representations 214 can be used, at 218, to determine second quantitative measures for a number of genomic regions. The number of genomic regions can include a plurality of genomic regions that are enriched as part of one or more diagnostic assays administered to detect one or more tumor-related biological conditions. In addition, the number of genomic regions can include a plurality of differentially methylated regions. In various examples, the second quantitative measures can include a number of the individual sequence representations included in the first group of sequence representations 214 that are aligned with an individual genomic region of the plurality of genomic regions. For example, the second quantitative measures can include counts of the individual sequence representations 214 that are aligned with individual genomic regions of the number of genomicregions. In one or more illustrative examples, the first group of sequence representations 214 can be used to determine genomic region quantitative measures.

[0308] The process 200 can also include, at 220, executing a machine learning classification model to determine a tumor indication 222. In one or more examples, the second quantitative measures can be input to the machine learning classification model. The second quantitative measures can be provided to the machine learning classification model as one or more vectors with individual values of the one or more vectors corresponding to individual genomic regions. The machine learning classification model can include at least one of a random forests model, a multilayer perceptron model, a boosted tree model, a support vector machine, a logistic regression model, a Naïve Bayes classification model, or an artificial neural network.

[0309] Figure 3 is a diagrammatic representation of an example computational environment 300 to determine one or more criteria that can be applied to sequence representations to identify a group of sequence representations corresponding to nucleic acid molecules derived from subjects in which a tumor-related biological condition is present, according to one or more example implementations. The computational environment 300 can include the computational system 102 described in relation to Figure 1 and Figure 2. The computational environment 300 can perform a training process to determine the one or more criteria to be applied to identify nucleic acid molecules and / or features of nucleic acid molecules that are indicative to a tumor-related biological condition being present in subjects.

[0310] The computational environment 300 can include first training data 302 that is derived from first subjects 304. In one or more examples, a tumor-related biological condition can be detected in the first subjects 304. The first training data 302 can include values of features of first sequence representations 306 that are derived from samples obtained from the first subjects 304. For example, the first sequence representations 306 can be determined through a computational analysis of sequencing data generated based on nucleic molecules included in samples obtained from the first subjects 304. In one or more illustrative examples, the first training data 302 can include values corresponding to at least one of genomic position of the first sequence representations 306, lengths of the first sequence representations 306, family sizes related to the first sequence representations 306, methylation partitions of the first sequence representations 306, number of CpGs present in the first sequence representations 306, number of methylated CpGs present in the first sequence representations 306, number of restriction enzyme cut sites for the first sequence representations 306, or depth of the samples used to produce the first sequence representations 306.

[0311] The computational environment 300 can also include second training data 308 that is derived from second subjects 310. In one or more examples, a tumor-related biological condition is not detected in the second subjects 310. The second training data 308 can include values of features of second sequence representations 312 that are derived from samples obtained from the second subjects 310. For example, the second sequence representations 312 can be determined through a computational analysis of sequencing data generated based on nucleic molecules included in samples obtained from the second subjects 310. In one or more illustrative examples, the second training data 308 can include values corresponding to at least one of genomic position of the second sequence representations 312, lengths of the second sequence representations 312, family sizes related to the second sequence representations 312, methylation partitions of the second sequence representations 312, number of CpGs present in the second sequence representations 312, number of methylated CpGs present in the second sequence representations 312, number of restriction enzyme cut sites for the second sequence representations 312, or depth of the samples used to produce second sequence representations 312.

[0312] The computational system 102 can, at 314, perform a training process to determine criteria used to identify sequence representations that can be provided as input to a machine learning classification model to predict the presence or absence of one or more tumor- related biological conditions in subjects. In various examples, the training process can be performed in relation to values of one or more combinations of features included in the first training data 302. For example, the training process can be performed based on values of genomic position and at least one of lengths of the sequence representations, family sizes related to sequence representations, methylation partitions of sequence representations, number of CpGs present in sequence representations, number of methylated CpGs present in sequence representations, number of restriction enzyme cut sites for sequence representations, or depth of the samples used to produce sequence representations. The one or more combinations of features used in the training process can be determined through one or more machine learning features selection processes and / or one or more statistical feature selection processes.

[0313] The training process can include, at 316, determining first training quantitative measures based on the first training data 302 and the one or more combinations of features applied during the training process. The first quantitative measures can correspond to first probabilities of the first sequence representations 306 being derived from samples in which one or more tumor-related biological conditions are present. The training process can also include, at 318, determining second training quantitative measures based on the second training data 308and the one or more combinations of features applied during the training process. The second quantitative measures can correspond to second probabilities of the second sequence representations 312 being derived from samples in which the one or more tumor-related biological conditions is present.

[0314] In various examples, the training process can include performing one or more computational analyses of the first training quantitative measures and the second training quantitative measures to determine one or more genomic region probability maps 320. In at least some examples, the one or more genomic region probability maps 320 can be determined using a classification model, such as a logistic regression model. In one or more examples, the one or more genomic region probability maps 320 can indicate individual probabilities that nucleotide representations aligned with individual genomic regions correspond to the one or more tumor- related biological conditions being present in subjects. In these scenarios, the one or more probability maps 320 can indicate genomic regions that are more or less predictive of one or more tumor-related biological conditions than other genomic regions. In at least some examples, individual genomic region probability maps 320 can be generated for different combinations of features of the first sequence representations 306 and the second sequence representations 312. For example, the training process can produce a first genomic region probability map based on a first set of the sequence representation features and a second genomic region probability map based on a second set of the sequence representation features. In one or more illustrative examples, the first set of the sequence representation features can include a first combination of two, three, four, five, or six sequence representation features and the second set of sequence representation features can include a second combination of two, three, four, five, or six sequence representation features where at least one sequence representation feature included in the second set of sequence representation features is different from at least one sequence representation feature included in the first set of sequence representation features. To illustrate, a first genomic region probability map can be determined based on genomic position and sequence representation length and a second genomic region probability map can be determined based on genomic position and number of methylated CpGs present in the sequence representations. In still other examples, the training process can produce a first genomic region probability map based on genomic position, sequence representation length, and family size and a second genomic region probability map based on genomic position, number of methylated CpGs present in sequence representations, and methylation partition of the sequence representations.

[0315] The training process can include performing a computational analysis of the one or more genomic region probability maps 320 to determine a data identification framework 322. In one or more examples, the training process can determine a number of genomic regions that have at least a threshold probability of corresponding to sequence representations derived from subjects in which one or more tumor-related biological conditions is present. In these scenarios, the data identification framework 322 can include locations of the number of genomic regions identified by the training process. Additionally, in these instances, the data identification framework 322 can be applied to a given sample by determining sequence representations derived from the given sample that are aligned with the number of genomic regions and supplying data derived from the aligned sequence representations to a machine learning classification model.

[0316] Additionally, the training process can determine one or more threshold quantitative measures that can be included in the data identification framework 322. For example, the training process can determine one or more threshold probabilities for sequence representations, where the one or more threshold probabilities indicate the likelihood of sequence representations being derived from a sample obtained from a subject in which the one or more tumor-related biological conditions is present. In at least some cases, the one or more threshold probabilities can correspond to individual genomic regions. In one or more further examples, the one or more threshold probabilities can be applied with respect to a number of genomic regions. In one or more illustrative examples, the data identification framework 322 can be applied to a given sample by computationally analyzing the probability values determined by the computational system 102 for individual sequence representations derived from the given sample with respect to the one or more threshold probabilities. In situations where the probability value of an individual sequence representation derived from the given sample corresponds to at least a threshold probability value, data derived from the individual sequence representation can be provided as input a machine learning classification model used to detect subjects in which the one or more tumor- related biological conditions is present. In one or more additional scenarios where the probability value of an individual sequence representation is less than a threshold probability, data derived from the individual sequence representation can excluded from being provided as input to a machine learning classification model used to detect subjects in which the one or more tumor- related biological conditions is present.

[0317] Figure 4 is a flow diagram of an example process 400 to determine sequence representations to be analyzed by a machine learning classification model to detect a tumor- related biological condition, according to one or more implementations. At 402, the process 400can include obtaining molecule selection data generated from a training process of one or more machine learning models. The molecule selection data can include one or more threshold quantitative measures indicative of nucleic acid molecules derived from subjects in which a tumor- related biological condition is present. In one or more examples, the molecule selection data can correspond to the data identification framework described in relation to Figure 1, Figure 2, and Figure 3.

[0318] In one or more examples, the training process can include obtaining first training data from first samples derived from first subjects in which a tumor is detected. The first training data can include first nucleotide sequence representations corresponding to first nucleic acid molecules included in the first samples. In addition, the training process can include obtaining second training data from second samples derived from second subjects in which a tumor is not detected. The second training data can include second nucleotide sequence representations corresponding to second nucleic acid molecules included in the second samples. The training process can also include determining first quantitative measures based on first values for features of the first nucleotide sequence representations and determining second quantitative measures based on second values for the features of the second nucleotide sequence representations. In various examples, the training process can include determining, based on the first quantitative measures and the second quantitative measures, one or more threshold quantitative measures to identify nucleotide sequence representations to provide to a machine learning classification model to determine an indication of a tumor-related biological condition.

[0319] In one or more illustrative examples, the training process for the one or more machine learning models can include for individual genomic regions, determining, based on differences between the first values and the second values, probabilities that specified sets of values of the features for nucleotide sequence representations correspond to samples obtained from subjects in which the tumor-related biological condition is present. In one or more illustrative examples, a probability map can be generated that indicates the probabilities for the individual genomic regions. The probability map can be used to determine genomic regions having at least a threshold probability of corresponding to nucleic acid molecules derived from subjects in which a tumor-related biological condition is present. In these scenarios, the genomic regions can be part of a data identification framework that is applied to determine input data for the one or more machine learning classification models. Data related to other genomic regions having less than the threshold probability of corresponding to nucleic acid molecules derived from subjects in which a tumor-related biological condition is present can be excluded from input to the one or moremachine learning classification models. In various examples, at least one of logistic regression techniques can be performed to determine the threshold probability.

[0320] In at least some examples, the training process can include determining, based on the probabilities, that nucleic acid molecules aligned with one or more genomic regions have less than a threshold probability of identifying samples obtained from subjects in which the tumor- related biological condition is present. In these scenarios, data corresponding to the nucleic acid molecules aligned with the one or more genomic regions may not be provided to the machine learning classification model to determine an indication of the tumor-related biological condition being present in subjects. Further, in these instances one or more diagnostic tests administered for the detection of the tumor-related biological condition may not enrich the one or more genomic regions. In this way, the training process can be used as a design tool for diagnostic tests for the tumor-related biological condition.

[0321] In addition, at 404, the process 400 can include obtaining a plurality of nucleotide sequence representations derived from one or more samples of a subject. The plurality of nucleotide sequence representations can correspond to nucleic acid molecules included in the one or more samples. Further, the process 400 can include, at 406, determining quantitative measures based on values for features of the plurality of nucleotide sequence representations. In at least some examples, the quantitative measures can indicate differences between nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present and nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is not present. To illustrate, the quantitative measures can indicate probabilities of nucleotide sequence representations corresponding to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present.

[0322] In one or more examples, methylation data can be obtained that indicates a number of methylated cytosine-guanine dinucleotides (CpGs) present in the nucleic acid molecules. In various examples, the values for the features with respect to the plurality of nucleotide sequence representations can be determined based on characteristics of the plurality of nucleotide sequence representations and the methylation data. In one or more illustrative examples, the sequence representations can correspond to sequencing reads derived from nucleic acid molecules included in the one or more samples and the quantitative measures can be determined with respect to individual sequencing reads. In one or more additional illustrative examples, the sequence representations can correspond to nucleic acid molecules included in the one or more samples and the quantitative measure can be determined with respect toindividual nucleic acid molecules present in the one or more samples. In still other examples, the quantitative measures can correspond to a family of sequencing reads and the quantitative measures can be determined with respect to individual families of sequencing reads derived from the one or more samples.

[0323] In at least some examples, the values for the features with respect to the plurality of nucleotide sequence representations can include determining a number of CpGs included in individual nucleotide sequence representations of the plurality of nucleotide sequence representations. In one or more additional examples, the values for the features with respect to the plurality of nucleotide sequence representations can include determining an amount of methylated CpGs within individual sequence representations of the plurality of nucleotide sequence representations. In one or more illustrative examples, the amount of methylated CpGs can include a count of individual methylated CpGs for the plurality of nucleotide sequence representations within the one or more regions. In one or more additional illustrative examples, the amount of methylated CpGs can correspond to a range of methylated CpGs included in the individual nucleotide sequence representations. In these scenarios, the amount of methylated CpGs can indicate a methylation partition that is related to a nucleobase methylation state detection process implemented with respect to the nucleic acid molecules derived from the one or more samples of the subject.

[0324] In one or more examples, values of the features of the plurality of nucleotide sequence representations can be determined based on a number of restriction enzyme cut sites corresponding to individual nucleotide sequence representations of the plurality of nucleotide sequence representations. In one or more additional examples, values of features of the plurality of nucleotide sequence representations can be determined based on lengths of individual nucleotide sequence representations. In one or more further examples, values of the features of the plurality of nucleotide sequence representations can be determined, for a given genomic region, a number of nucleic acid molecules corresponding to the plurality of sequence representations aligned with the given genomic region in relation to a total number of the nucleic acid molecules. In still other examples, values of the features of the plurality of nucleotide sequence representations can be determined based on, for individual nucleotide sequence representations, an offset of a start position with respect to a genomic region in which the individual nucleotide sequence representations are located.

[0325] The process 400 can also include, at 408, determining a subset of the plurality of nucleotide sequence representations for which the quantitative measures correspond to the one or more threshold quantitative measures of the data identification framework. In variousexamples, at 410, the process 400 can include providing the subset of the plurality of nucleotide sequence representations to a machine learning classification model. In at least some examples, the values for the features with respect to the plurality of nucleotide sequence representations can be represented as vectors that are provided as input to the machine learning classification model. In one or more illustrative examples, the machine learning classification model can include the machine learning classification model can include at least one of a random forests model, a multilayer perceptron model, a boosted trees model, or an artificial neural network.

[0326] Additionally, the process 400 can include, at 412, determining, by the machine learning classification model, an indication of the tumor-related biological condition for the subject. The indication of the tumor-related biological condition for the subject can indicates that a tumor is present or a tumor is absent in the subject. In one or more additional examples, the indication of the tumor-related biological condition for the subject can correspond to tumor fraction for the one or more samples. In one or more further examples, the indication of the tumor-related biological condition can correspond to reoccurrence of cancer within the subject. In still other examples, the indication of the tumor-related biological condition can indicate a responsiveness of the subject to one or more treatments for the tumor-related biological condition. EXEMPLARY METHODS A. Determining an indication of a tumor-related biological condition in a sample

[0327] The techniques described herein relate to a method including: obtaining first training data from first samples derived from first subjects in which a tumor is detected. The first training data includes first nucleotide sequence representations corresponding to first nucleic acid molecules included in the first samples. In addition, the method includes obtaining second training data from second samples derived from second subjects in which a tumor is not detected. The second training data includes second nucleotide sequence representations corresponding to second nucleic acid molecules included in the second samples. The method also includes performing a training process for one or more machine learning models. The training process includes determining first quantitative measures based on first values for features of the first nucleotide sequence representations; determining second quantitative measures based on second values for the features of the second nucleotide sequence representations; and determining, based on the first quantitative measures and the second quantitative measures, one or more threshold quantitative measures to identify nucleotide sequence representations to provide to a machine learning classification model to determine an indication of a tumor-related biological condition. Further, the method includes obtaining additional nucleotide sequence representations derived from an additional subject and corresponding to additional nucleic acidmolecules included in an additional sample obtained from the additional subject and determining additional quantitative measures based on additional values for the features of the additional nucleotide sequence representations. Additionally, the method includes determining a subset of the additional nucleotide sequence representations for which the additional quantitative measures correspond to the one or more threshold quantitative measures and providing the subset of the additional nucleotide sequence representations to the machine learning classification model to determine the indication of the tumor-related biological condition for the additional subject.

[0328] The techniques described herein relate to a method including obtaining molecule selection data generated from a training process of one or more machine learning models. The molecule selection data includes one or more threshold quantitative measures indicative of nucleic acid molecules derived from subjects in which a tumor-related biological condition is present. The method also includes obtaining a plurality of nucleotide sequence representations derived from one or more samples of a subject. The plurality of nucleotide sequence representations correspond to nucleic acid molecules included in the one or more samples. In addition, the method includes determining quantitative measures based on values for features of the plurality of nucleotide sequence representations. Further, the method includes determining a subset of the plurality of nucleotide sequence representations for which the quantitative measures correspond to the one or more threshold quantitative measures and providing the subset of the plurality of nucleotide sequence representations to a machine learning classification model. Additionally, the method includes determining, by the machine learning classification model, an indication of the tumor-related biological condition for the subject.

[0329] The techniques described herein relate to one or more computing apparatuses, including: one or more hardware processors; and memory storing computer-readable instructions that, when executed by the one or more hardware processors, cause the one or more hardware processors to perform operations comprising: obtaining first training data from first samples derived from first subjects in which a tumor is detected. The first training data includes first nucleotide sequence representations corresponding to first nucleic acid molecules included in the first samples. The operations also include obtaining second training data from second samples derived from second subjects in which a tumor is not detected. The second training data includes second nucleotide sequence representations corresponding to second nucleic acid molecules included in the second samples. The operations also include performing a training process for one or more machine learning models. The training process includes: determining first quantitative measures based on first values for features of the first nucleotide sequence representations; determining second quantitative measures based on second values for thefeatures of the second nucleotide sequence representations; and determining, based on the first quantitative measures and the second quantitative measures, one or more threshold quantitative measures to identify nucleotide sequence representations to provide to a machine learning classification model to determine an indication of a tumor-related biological condition. Further, the operations include obtaining additional nucleotide sequence representations derived from an additional subject and corresponding to additional nucleic acid molecules included in an additional sample obtained from the additional subject and determining additional quantitative measures based on additional values for the features of the additional nucleotide sequence representations. Additionally, the operations include determining a subset of the additional nucleotide sequence representations for which the additional quantitative measures correspond to the one or more threshold quantitative measures and providing the subset of the additional nucleotide sequence representations to the machine learning classification model to determine the indication of the tumor-related biological condition for the additional subject.

[0330] The techniques described herein relate to one or more computer apparatuses comprising: one or more hardware processors; and memory storing computer-readable instructions that, when executed by the one or more hardware processors, causes the one or more hardware processors to perform operations comprising: obtaining molecule selection data generated from a training process of one or more machine learning models. The molecule selection data includes one or more threshold quantitative measures indicative of nucleic acid molecules derived from subjects in which a tumor-related biological condition is present. The operations also include obtaining a plurality of nucleotide sequence representations derived from one or more samples of a subject. the plurality of nucleotide sequence representations corresponding to nucleic acid molecules included in the one or more samples. Additionally, the operations include determining quantitative measures based on values for features of the plurality of nucleotide sequence representations and determining a subset of the plurality of nucleotide sequence representations for which the quantitative measures correspond to the one or more threshold quantitative measures. Further, the operations include providing the subset of the plurality of nucleotide sequence representations to a machine learning classification model and determining, by the machine learning classification model, an indication of the tumor-related biological condition for the subject.

[0331] The techniques described herein relate to one or more non-transitory computer- readable media storing computer-readable instructions that, when executed by one or more hardware processors, cause the one or more hardware processors to perform operations comprising: obtaining first training data from first samples derived from first subjects in which atumor is detected. The first training data including first nucleotide sequence representations corresponding to first nucleic acid molecules included in the first samples. The operations also include obtaining second training data from second samples derived from second subjects in which a tumor is not detected. The second training data includes second nucleotide sequence representations corresponding to second nucleic acid molecules included in the second samples. Additionally, the operations include performing a training process for one or more machine learning models. The training process includes: determining first quantitative measures based on first values for features of the first nucleotide sequence representations; determining second quantitative measures based on second values for the features of the second nucleotide sequence representations; and determining, based on the first quantitative measures and the second quantitative measures, one or more threshold quantitative measures to identify nucleotide sequence representations to provide to a machine learning classification model to determine an indication of a tumor-related biological condition. Further, the operations include obtaining additional nucleotide sequence representations derived from an additional subject and corresponding to additional nucleic acid molecules included in an additional sample obtained from the additional subject and determining additional quantitative measures based on additional values for the features of the additional nucleotide sequence representations. In addition, the operations include determining a subset of the additional nucleotide sequence representations for which the additional quantitative measures correspond to the one or more threshold quantitative measures and providing the subset of the additional nucleotide sequence representations to the machine learning classification model to determine the indication of the tumor-related biological condition for the additional subject.

[0332] The techniques described herein relate to one or more non-transitory computer- readable media, comprising computer-readable instructions that, when executed by one or more hardware processors, causes the one or more hardware processors to perform operations comprising: obtaining molecule selection data generated from a training process of one or more machine learning models. The molecule selection data includes one or more threshold quantitative measures indicative of nucleic acid molecules derived from subjects in which a tumor- related biological condition is present. The operations also include obtaining a plurality of nucleotide sequence representations derived from one or more samples of a subject. The plurality of nucleotide sequence representations correspond to nucleic acid molecules included in the one or more samples. In addition, the operations include determining quantitative measures based on values for features of the plurality of nucleotide sequence representations and determining a subset of the plurality of nucleotide sequence representations for which the quantitative measurescorrespond to the one or more threshold quantitative measures. Further, the operations include providing the subset of the plurality of nucleotide sequence representations to a machine learning classification model and determining, by the machine learning classification model, an indication of the tumor-related biological condition for the subject. B. Partitioning the sample into a plurality of subsamples

[0333] In some embodiments described herein, different forms of DNA (e.g., hypermethylated and hypomethylated DNA) are physically partitioned based on one or more characteristics of the DNA. This approach can be used to determine, for example, whether certain sites or regions are hypermethylated or hypomethylated. Partitioning can be performed before attaching adapters to DNA molecules in the sample, e.g., so as to facilitate including partition tags in the adapters. Partition tags can be used to identify which partition a molecule was found in. Following partitioning (and attachment of adapters if applicable), further steps such as amplification, target capture, and sequencing may be performed.

[0334] Methylation profiling can involve determining methylation patterns across different regions of the genome. For example, after partitioning molecules based on extent of methylation (e.g., relative number of methylated nucleobases per molecule) and further steps as discussed above including sequencing, the sequences of molecules in the different partitions can be mapped to a reference genome. This can show regions of the genome that, compared with other regions, are more highly methylated or are less highly methylated. In this way, genomic regions, in contrast to individual molecules, may differ in their extent of methylation.

[0335] Partitioning nucleic acid molecules in a sample can increase a rare signal, e.g., by enriching rare nucleic acid molecules that are more prevalent in one partition of the sample. For example, a genetic variation present in hypermethylated DNA but less (or not) present in hypomethylated DNA can be more easily detected by partitioning a sample into hypermethylated and hypomethylated nucleic acid molecules. By analyzing multiple partitions of a sample, a multi- dimensional analysis of a single molecule can be performed and hence, greater sensitivity can be achieved. Partitioning may include physically partitioning nucleic acid molecules into partitions or subsamples based on the presence or absence of one or more methylated nucleobases. A sample may be partitioned into partitions or subsamples based on a characteristic that is indicative of differential gene expression or a disease state. A sample may be partitioned based on a characteristic, or combination thereof that provides a difference in signal between a normal and diseased state during analysis of nucleic acids, e.g., cell free DNA (cfDNA), non-cfDNA, tumor DNA, circulating tumor DNA (ctDNA) and cell free nucleic acids (cfNA).

[0336] In some embodiments, hypermethylation and / or hypomethylation variable epigenetic target regions are analyzed to determine whether they show differential methylation characteristic of particular immune cell types, such as rare immune cell types, tumor cells or cells of a type that does not normally contribute to the DNA sample being analyzed (such as cfDNA).

[0337] In some instances, heterogeneous DNA in a sample is partitioned into two or more partitions (e.g., at least 3, 4, 5, 6 or 7 partitions). In some embodiments, each partition is differentially tagged. Tagged partitions can then be pooled together for collective sample prep and / or sequencing. The partitioning-tagging-pooling steps can occur more than once, with each round of partitioning occurring based on a different characteristic (examples provided herein), and tagged using differential tags that are distinguished from other partitions and partitioning means. In other instances, the differentially tagged partitions are separately sequenced.

[0338] In some embodiments, sequence reads from differentially tagged and pooled DNA are obtained and analyzed in silico. Tags are used to sort reads from different partitions. Analysis to detect genetic variants can be performed on a partition-by-partition level, as well as whole nucleic acid population level. For example, analysis can include in silico analysis to determine genetic variants, such as CNV, SNV, indel, fusion in nucleic acids in each partition. In some instances, in silico analysis can include determining chromatin structure. For example, coverage of sequence reads can be used to determine nucleosome positioning in chromatin. Higher coverage can correlate with higher nucleosome occupancy in genomic region while lower coverage can correlate with lower nucleosome occupancy or nucleosome depleted region (NDR).

[0339] In some embodiments, partitioning is on the basis of one or more characteristics such as methylation. Molecules can be sorted according to other characteristics, such as sequence length, nucleosome binding, sequence mismatch, immunoprecipitation, and / or proteins that bind to DNA, using appropriate techniques as part of data analysis or partitioning as applicable. Resulting partitions can include one or more of the following nucleic acid forms: single- stranded DNA (ssDNA), double-stranded DNA (dsDNA), shorter DNA fragments and longer DNA fragments. In some embodiments, partitioning based on a cytosine modification (e.g., cytosine methylation) or methylation generally is performed and is optionally combined with at least one additional partitioning step, which may be based on any of the foregoing characteristics or forms of DNA. In some embodiments, a heterogeneous population of nucleic acids is partitioned into nucleic acids with one or more epigenetic modifications and without the one or more epigenetic modifications. Examples of epigenetic modifications include presence or absence of methylation; level of methylation; type of methylation (e.g., 5-methylcytosine versus other types of methylation, such as adenine methylation and / or cytosine hydroxymethylation); and association and level ofassociation with one or more proteins, such as histones. Alternatively, or additionally, a heterogeneous population of nucleic acids can be partitioned into nucleic acid molecules associated with nucleosomes and nucleic acid molecules devoid of nucleosomes. Alternatively, or additionally, a heterogeneous population of nucleic acids may be partitioned into single- stranded DNA (ssDNA) and double-stranded DNA (dsDNA). Alternatively, or additionally, a heterogeneous population of nucleic acids may be partitioned based on nucleic acid length (e.g., molecules of up to 160 bp and molecules having a length of greater than 160 bp).

[0340] The agents used to partition populations of nucleic acids within a sample can be affinity agents, such as antibodies with the desired specificity, natural binding partners or variants thereof (Bock et al., Nat Biotech 28: 1106-1114 (2010); Song et al., Nat Biotech 29: 68-72 (2011)), or artificial peptides selected e.g., by phage display to have specificity to a given target. In some embodiments, the agent used in the partitioning is an agent that recognizes a modified nucleobase. In some embodiments, the modified nucleobase recognized by the agent is a modified cytosine, such as a methylcytosine (e.g., 5-methylcytosine). In some embodiments, the modified nucleobase recognized by the agent is a product of a procedure that affects the first nucleobase in the DNA differently from the second nucleobase in the DNA of the sample. In some embodiments, the modified nucleobase may be a “converted nucleobase,” meaning that its base pairing specificity was changed by the procedure. For example, certain procedures convert unmethylated or unmodified cytosine to dihydrouracil, or more generally, at least one modified or unmodified form of cytosine undergoes deamination, resulting in uracil (considered a modified nucleobase in the context of DNA) or a further modified form of uracil. Examples of partitioning agents include antibodies, such as antibodies that recognize a modified nucleobase, which may be a modified cytosine, such as a methylcytosine (e.g., 5-methylcytosine). In some embodiments, the partitioning agent is an antibody that recognizes a modified cytosine other than 5- methylcytosine, such as 5-carboxylcytosine (5caC). Alternative partitioning agents include methyl binding domain (MBDs) and methyl binding proteins (MBPs) as described herein, including proteins such as MeCP2.

[0341] Additional, non-limiting examples of partitioning agents are histone binding proteins which can separate nucleic acids bound to histones from free or unbound nucleic acids. Examples of histone binding proteins that can be used in the methods disclosed herein include RBBP4, RbAp48 and SANT domain peptides.

[0342] The binding of partitioning agents to particular nucleic acids and the partitioning of the nucleic acids into subsamples may occur to a certain extent or may occur in an essentially binary manner. In some instances, nucleic acids comprising a greater proportion of a certainmodification bind to the agent at a greater extent than nucleic acids comprising a lesser proportion of the modification. Similarly, the partitioning may produce subsamples comprising greater and lesser proportions of nucleic acids comprising a certain modification. Alternatively, the partitioning may produce subsamples comprising essentially all or none of the nucleic acids comprising the modification. In all instances, various levels of modifications may be sequentially eluted from the partitioning agent.

[0343] In some embodiments, partitioning can comprise both binary partitioning and partitioning based on degree / level of modifications. For example, methylated fragments can be partitioned by methylated DNA immunoprecipitation (MeDIP), or all methylated fragments can be partitioned from unmethylated fragments using methyl binding domain proteins (e.g., MethylMinder Methylated DNA Enrichment Kit (ThermoFisher Scientific). Subsequently, additional partitioning may involve eluting fragments having different levels of methylation by adjusting the salt concentration in a solution with the methyl binding domain and bound fragments. As salt concentration increases, fragments having greater methylation levels are eluted.

[0344] In some instances, the final partitions are enriched in nucleic acids having different extents of modifications (overrepresentative or underrepresentative of modifications). Overrepresentation and underrepresentation can be defined by the number of modifications born by a nucleic acid relative to the median number of modifications per strand in a population. For example, if the median number of 5-methylcytosine residues in nucleic acid in a sample is 2, a nucleic acid including more than two 5-methylcytosine residues is overrepresented in this modification and a nucleic acid with 1 or zero 5-methylcytosine residues is underrepresented. The effect of the affinity separation is to enrich for nucleic acids overrepresented in a modification in a bound phase and for nucleic acids underrepresented in a modification in an unbound phase (i.e., in solution). The nucleic acids in the bound phase can be eluted before subsequent processing.

[0345] When using MeDIP or MethylMiner®Methylated DNA Enrichment Kit (ThermoFisher Scientific) various levels of methylation can be partitioned using sequential elutions. For example, a hypomethylated partition (no methylation) can be separated from a methylated partition by contacting the nucleic acid population with the MBD from the kit, which is attached to magnetic beads. The beads are used to separate out the methylated nucleic acids from the non- methylated nucleic acids. Subsequently, one or more elution steps are performed sequentially to elute nucleic acids having different levels of methylation. For example, a first set of methylated nucleic acids can be eluted at a salt concentration of 160 mM or higher, e.g., at least 150 mM, at least 200 mM, 300 mM, 400 mM, 500 mM, 600 mM, 700 mM, 800 mM, 900 mM,1000 mM, or 2000 mM. After such methylated nucleic acids are eluted, magnetic separation is once again used to separate higher level of methylated nucleic acids from those with lower level of methylation. The elution and magnetic separation steps can be repeated to create various partitions such as a hypomethylated partition (enriched in nucleic acids comprising no methylation), a methylated partition (enriched in nucleic acids comprising low levels of methylation), and a hyper methylated partition (enriched in nucleic acids comprising high levels of methylation).

[0346] In some methods, nucleic acids bound to an agent used for affinity separation- based partitioning are subjected to a wash step. The wash step washes off nucleic acids weakly bound to the affinity agent. Such nucleic acids can be enriched in nucleic acids having the modification to an extent close to the mean or median (i.e., intermediate between nucleic acids remaining bound to the solid phase and nucleic acids not binding to the solid phase on initial contacting of the sample with the agent).

[0347] The affinity separation results in at least two, and sometimes three or more partitions of nucleic acids with different extents of a modification. While the partitions are still separate, the nucleic acids of at least one partition, and usually two or three (or more) partitions are linked to nucleic acid tags, usually provided as components of adapters, with the nucleic acids in different partitions receiving different tags that distinguish members of one partition from another. The tags linked to nucleic acid molecules of the same partition can be the same or different from one another. But if different from one another, the tags may have part of their code in common so as to identify the molecules to which they are attached as being of a particular partition.

[0348] For further details regarding portioning nucleic acid samples based on characteristics such as methylation, see WO2018 / 119452, which is incorporated herein by reference.

[0349] In some embodiments, the nucleic acid molecules can be fractionated into different partitions based on the nucleic acid molecules that are bound to a specific protein or a fragment thereof and those that are not bound to that specific protein or fragment thereof.

[0350] Nucleic acid molecules can be fractionated based on DNA-protein binding. Protein- DNA complexes can be fractionated based on a specific property of a protein. Examples of such properties include various epitopes, modifications (e.g., histone methylation or acetylation) or enzymatic activity. Examples of proteins which may bind to DNA and serve as a basis for fractionation may include, but are not limited to, protein A and protein G. Any suitable method can be used to fractionate the nucleic acid molecules based on protein bound regions. Examples ofmethods used to fractionate nucleic acid molecules based on protein bound regions include, but are not limited to, SDS-PAGE, chromatin-immuno-precipitation (ChIP), heparin chromatography, and asymmetrical field flow fractionation (AF4).

[0351] In some embodiments, the partitioning of the sample into a plurality of subsamples is performed by contacting the nucleic acids with an antibody that recognizes a modified nucleobase in the DNA, which may be is a modified cytosine or a product of the procedure that affects the first nucleobase in the DNA differently from the second nucleobase in the DNA of the sample. In some embodiments, the modified nucleobase is 5mC. In some embodiments, the modified nucleobase is 5caC. In some embodiments, the modified nucleobase is dihydrouracil (DHU). In some embodiments, the antibody that recognizes a modified nucleobase in the DNA is used to partition single-stranded DNA.

[0352] In some embodiments, the partitioning is performed by contacting the nucleic acids with a methyl binding domain (“MBD”) of a methyl binding protein (“MBP”). In some such embodiments, the nucleic acids are contacted with an entire MBP. In some embodiments, an MBD binds to 5-methylcytosine (5mC), and an MBP comprises an MBD and is referred to interchangeably herein as a methyl binding protein or a methyl binding domain protein. In some embodiments, an MBD binds to 5mC and 5hmC. In some embodiments, MBD is coupled to paramagnetic beads, such as Dynabeads® M-280 Streptavidin via a biotin linker. Partitioning into fractions with different extents of methylation can be performed by eluting fractions by increasing the NaCl concentration.

[0353] In some embodiments, bound DNA is eluted by contacting the antibody or MBD with a protease, such as proteinase K. This may be performed instead of or in addition to elution steps using NaCl as discussed above.

[0354] Examples of agents that recognize a modified nucleobase contemplated herein include, but are not limited to: (a) MeCP2 is a protein that preferentially binds to 5-methyl-cytosine over unmodified cytosine. (b) RPL26, PRP8 and the DNA mismatch repair protein MHS6 preferentially bind to 5- hydroxymethyl-cytosine over unmodified cytosine. (c) FOXK1, FOXK2, FOXP1, FOXP4 and FOXI3 preferably bind to 5-formyl cytosine over unmodified cytosine (Iurlaro et al., Genome Biol.14: R119 (2013)). (d) Antibodies specific to one or more methylated or modified nucleobases or conversion products thereof, such as 5mC, 5caC, or DHU.

[0355] In general, elution is a function of the number of modifications, such as the number of methylated sites per molecule, with molecules having more methylation eluting under increased salt concentrations. To elute the DNA into distinct populations based on the extent of methylation, one can use a series of elution buffers of increasing NaCl concentration. Salt concentration can range from about 100 nm to about 2500 mM NaCl. In one embodiment, the process results in three (3) partitions. Molecules are contacted with a solution at a first salt concentration and comprising a molecule comprising an agent that recognizes a modified nucleobase, which molecule can be attached to a capture moiety, such as streptavidin. At the first salt concentration a population of molecules will bind to the agent and a population will remain unbound. The unbound population can be separated as a “hypomethylated” population. For example, a first partition enriched in hypomethylated form of DNA is that which remains unbound at a low salt concentration, e.g., 100 mM or 160 mM. A second partition enriched in intermediate methylated DNA is eluted using an intermediate salt concentration, e.g., between 100 mM and 2000 mM concentration. This is also separated from the sample. A third partition enriched in hypermethylated form of DNA is eluted using a high salt concentration, e.g., at least about 2000 mM.

[0356] In some embodiments, a monoclonal antibody raised against 5-methylcytidine (5mC) is used to purify methylated DNA. DNA is denatured, e.g., at 95°C in order to yield single- stranded DNA fragments. Protein G coupled to standard or magnetic beads as well as washes following incubation with the anti-5mC antibody are used to immunoprecipitate DNA bound to the antibody. Such DNA may then be eluted. Partitions may comprise unprecipitated DNA and one or more partitions eluted from the beads.

[0357] In some embodiments, sample DNA (e.g., between 5 and 200 ng) is mixed with methyl binding domain (MBD) buffer and magnetic beads conjugated with MBD proteins and incubated overnight. Methylated DNA (hypermethylated DNA) binds the MBD protein on the magnetic beads during this incubation. Non-methylated (hypomethylated DNA) or less methylated DNA (intermediately methylated) is washed away from the beads with buffers containing increasing concentrations of salt. For example, one, two, or more fractions containing non-methylated, hypomethylated, and / or intermediately methylated DNA may be obtained from such washes. Finally, a high salt buffer is used to elute the heavily methylated DNA (hypermethylated DNA) from the MBD protein. In some embodiments, these washes result in three partitions (hypomethylated partition, intermediately methylated fraction and hypermethylated partition) of DNA having increasing levels of methylation.

[0358] In some embodiments, partitioning procedures may result in imperfect sorting of DNA molecules among the subsamples. For example, a minority of the molecules in an unmethylated or hypomethylated subsample may be highly modified (e.g., hypermethylated), and / or a minority of the molecules in a hypermethylated subsample may be unmodified or mostly unmodified (e.g., unmethylated or mostly unmethylated). Such molecules are considered nonspecifically partitioned.

[0359] In some embodiments, nonspecifically partitioned molecules are removed using a methylation-dependent nuclease, e.g., a methylation dependent restriction enzyme (MDRE), digesting / cleaving the DNA where the restriction enzyme (RE) recognition site contains a methylated nucleotide but not cleaving the DNA where the restriction enzyme (RE) recognition site contains an unmethylated nucleotide. In some embodiments, nonspecifically partitioned molecules are removed using a methylation sensitive nuclease, e.g., a methylation sensitive restriction enzyme (MSRE), digesting / cleaving the DNA where the restriction enzyme (RE) recognition site contains an unmethylated nucleotide but not cleaving the DNA where the restriction enzyme (RE) recognition site contains a methylated nucleotide. For example, in some embodiments, a hypomethylated subsample is contacted with a methylation-dependent nuclease, such as a methylation-dependent restriction enzyme, thereby degrading nonspecifically partitioned DNA, e.g., methylated DNA, in the subsample. Alternatively, or in addition, a hypermethylated subsample is contacted with a methylation-sensitive endonuclease, such as a methylation-sensitive restriction enzyme, thereby degrading nonspecifically partitioned DNA in the subsample.

[0360] Degradation of nonspecifically partitioned DNA in one or more partitioned subsamples may improve the performance of methods that rely on accurate partitioning of DNA on the basis of a cytosine modification. For example, such degradation may provide improved sensitivity and / or simplify downstream analyses. In some embodiments, partitioning DNA on the basis of a modification, such as methylation, then removing nonspecifically partitioned DNA using MDREs and / or MSREs as described herein provides improved efficiency and / or cost over DNA analysis methods comprising procedures that affect a first nucleobase differently from a second nucleobase, such as bisulfite sequencing or bisulfite conversion.

[0361] In some embodiments, one or more nucleases are used to degrade nonspecifically partitioned DNA molecules. In some embodiments, a subsample is contacted with a plurality of nucleases. The subsample may be contacted with the nucleases sequentially or simultaneously. Simultaneous use of nucleases may be advantageous when the nucleases are active under similar conditions (e.g., buffer composition) to avoid unnecessary sample manipulation.Contacting a subsample with more than one methylation-dependent restriction enzyme can more completely degrade nonspecifically partitioned hypermethylated DNA. Contacting a subsample with more than one methylation-sensitive restriction enzyme can more completely degrade nonspecifically partitioned hypomethylated and / or unmethylated DNA.

[0362] In some embodiments, a methylation-dependent nuclease comprises one or more of MspJI, LpnPI, FspEI, or McrBC. In some embodiments, at least two methylation-dependent nucleases are used. In some embodiments, at least three methylation-dependent nucleases are used.

[0363] In some embodiments, a methylation-sensitive nuclease comprises one or more of AatII, AccII, AciI, Aor13HI, Aor15HI, BspT104I, BssHII, BstUI, Cfr10I, ClaI, CpoI, Eco52I, HaeII, HapII, HhaI, Hin6I, HpaII, HpyCH4IV, MluI, MspI, NaeI, NotI, NruI, NsbI, PmaCI, Psp1406I, PvuI, SacII, SalI, SmaI, and SnaBI. In some embodiments, at least two methylation-sensitive nucleases are used. In some embodiments, at least three methylation-sensitive nucleases are used. In some embodiments, the methylation-sensitive nucleases comprise BstUI and HpaII. In some embodiments, the two methylation-sensitive nucleases comprise HhaI and AccII. In some embodiments, the methylation-sensitive nucleases comprise BstUI, HpaII and Hin6I.

[0364] In some embodiments, the partitions of DNA are desalted and concentrated in preparation for enzymatic steps of library preparation. C. Adapter Ligation

[0365] In some embodiments, adapters are added to the DNA. This may be done concurrently with an amplification procedure, e.g., by providing the adapters in a 5’ portion of a primer (where PCR is used, this can be referred to as library prep-PCR or LP-PCR). In some embodiments, adapters are added by other approaches, such as ligation. In some such methods, prior to partitioning or prior to capturing, first adapters are added to the nucleic acids by ligation to the 3’ ends thereof, which may include ligation to single-stranded DNA. The adapter can be used as a priming site for second-strand synthesis, e.g., using a universal primer and a DNA polymerase. A second adapter can then be ligated to at least the 3’ end of the second strand of the now double-stranded molecule. In some embodiments, the first adapter comprises an affinity tag, such as biotin, and nucleic acid ligated to the first adapter is bound to a solid support (e.g., bead), which may comprise a binding partner for the affinity tag such as streptavidin. For further discussion of a related procedure, see Gansauge et al., Nature Protocols 8:737-748 (2013). Commercial kits for sequencing library preparation compatible with single-stranded nucleic acidsare available, e.g., the Accel-NGS® Methyl-Seq DNA Library Kit from Swift Biosciences. In some embodiments, after adapter ligation, nucleic acids are amplified.

[0366] Preferably, the adapters include different tags of sufficient numbers that the number of combinations of tags results in a low probability e.g., 95, 99 or 99.9% of two nucleic acids with the same start and stop points receiving the same combination of tags. Adapters, whether bearing the same or different tags, can include the same or different primer binding sites, but preferably adapters include the same primer binding site.

[0367] In some embodiments, following attachment of adapters, the nucleic acids are subject to amplification. The amplification can use, e.g., universal primers that recognize primer binding sites in the adapters.

[0368] In some embodiments, following attachment of adapters, the DNA is partitioned, comprising contacting the DNA with an agent that preferentially binds to nucleic acids bearing an epigenetic modification. The nucleic acids are partitioned into at least two subsamples differing in the extent to which the nucleic acids bear the modification from binding to the agents. For example, if the agent has affinity for nucleic acids bearing the modification, nucleic acids overrepresented in the modification (compared with median representation in the population) preferentially bind to the agent, whereas nucleic acids underrepresented for the modification do not bind or are more easily eluted from the agent. The nucleic acids can then be amplified from primers binding to the primer binding sites within the adapters. Partitioning may be performed instead before adapter attachment, in which case the adapters may comprise differential tags that include a component that identifies which partition a molecule occurred in.

[0214] In some embodiments, the nucleic acids are linked at both ends to Y-shaped adapters including primer binding sites and tags. The molecules are amplified. D. Tagging

[0369] “Tagging” DNA molecules is a procedure in which a tag is attached to or associated with the DNA molecules. Tags can be molecules, such as nucleic acids, containing information that indicates a feature of the molecule with which the tag is associated. For example, molecules can bear a sample tag (which distinguishes molecules in one sample from those in a different sample) or a molecular tag / molecular barcode / barcode (which distinguishes different molecules from one another (in both unique and non-unique tagging scenarios). For methods that involve a partitioning step, a partition tag (which distinguishes molecules in one partition from those in a different partition) may be included. In some embodiments, adapters added to DNA molecules comprise tags. In certain embodiments, a tag can comprise one or a combination of barcodes. As used herein, the term “barcode” refers to a nucleic acid molecule having a particular nucleotidesequence, or to the nucleotide sequence, itself, depending on context. A barcode can have, for example, between 10 and 100 nucleotides. A collection of barcodes can have degenerate sequences or can have sequences having a certain hamming distance, as desired for the specific purpose. So, for example, a molecular barcode can be comprised of one barcode or a combination of two barcodes, each attached to different ends of a molecule. Additionally, or alternatively, for different partitions and / or samples, different sets of molecular barcodes, or molecular tags can be used such that the barcodes serve as a molecular tag through their individual sequences and also serve to identify the partition and / or sample to which they correspond based the set of which they are a member.

[0370] In some embodiments, two or more partitions, e.g., each partition, is / are differentially tagged. Tags can be used to label the individual polynucleotide population partitions so as to correlate the tag (or tags) with a specific partition. Alternatively, tags can be used in embodiments that do not employ a partitioning step. In some embodiments, a single tag can be used to label a specific partition. In some embodiments, multiple different tags can be used to label a specific partition. In embodiments employing multiple different tags to label a specific partition, the set of tags used to label one partition can be readily differentiated for the set of tags used to label other partitions. In some embodiments, the tags may have additional functions, for example the tags can be used to index sample sources or used as unique molecular identifiers (which can be used to improve the quality of sequencing data by differentiating sequencing errors from mutations, for example as in Kinde et al., Proc Nat’l Acad Sci USA 108: 9530-9535 (2011), Kou et al., PLoS ONE, 11 : e0146638 (2016)) or used as non-unique molecule identifiers, for example as described in US Pat. No.9,598,731. Similarly, in some embodiments, the tags may have additional functions, for example the tags can be used to index sample sources or used as non-unique molecular identifiers (which can be used to improve the quality of sequencing data by differentiating sequencing errors from mutations).

[0371] In some embodiments, partition tagging comprises tagging molecules in each partition with a partition tag. After re-combining partitions (e.g., to reduce the number of sequencing runs needed and avoid unnecessary cost) and sequencing molecules, the partition tags identify the source partition. In some embodiments, the partition tags can serve as identifiers of the source partition and the molecule, i.e., different partitions are tagged with different sets of molecular tags, e.g., comprised of a pair of barcodes. In this way, the one or more molecular barcodes attached to the molecule indicates the source partition as well as being useful to distinguish molecules within a partition. For example, a first set of 35 barcodes can be used totag molecules in a first partition, while a second set of 35 barcodes can be used tag molecules in a second partition.

[0372] In some embodiments, after partitioning and tagging with partition tags, the molecules may be pooled for sequencing in a single run. In some embodiments, a sample tag is added to the molecules, e.g., in a step subsequent to addition of partition tags and pooling. Sample tags can facilitate pooling material generated from multiple samples for sequencing in a single sequencing run.

[0373] Alternatively, in some embodiments, partition tags may be correlated to the sample as well as the partition. As a simple example, a first tag can indicate a first partition of a first sample; a second tag can indicate a second partition of the first sample; a third tag can indicate a first partition of a second sample; and a fourth tag can indicate a second partition of the second sample.

[0374] While tags may be attached to molecules already partitioned based on one or more characteristics, the final tagged molecules in the library may no longer possess that characteristic. For example, while single stranded DNA molecules may be partitioned and tagged, the final tagged molecules in the library are likely to be double stranded. Similarly, while DNA may be subject to partition based on different levels of methylation, in the final library, tagged molecules derived from these molecules are likely to be unmethylated. Accordingly, the tag attached to molecule in the library typically indicates the characteristic of the “parent molecule” from which the ultimate tagged molecule is derived, not necessarily to characteristic of the tagged molecule, itself.

[0375] As an example, barcodes 1, 2, 3, 4, etc. are used to tag and label molecules in the first partition; barcodes A, B, C, D, etc. are used to tag and label molecules in the second partition; and barcodes a, b, c, d, etc. are used to tag and label molecules in the third partition. Differentially tagged partitions can be pooled prior to sequencing. Differentially tagged partitions can be separately sequenced or sequenced together concurrently, e.g., in the same flow cell of an Illumina sequencer.

[0376] After sequencing, analysis of reads can be performed on a partition-by-partition level, as well as a whole DNA population level. Tags are used to sort reads from different partitions. Analysis can include in silico analysis to determine genetic and epigenetic variation (one or more of methylation, chromatin structure, etc.) using sequence information, genomic coordinates length, coverage, and / or copy number. In some embodiments, higher coverage can correlate with higher nucleosome occupancy in genomic region while lower coverage can correlate with lower nucleosome occupancy or a nucleosome depleted region (NDR).E. Enriching / Capturing step; Amplification

[0377] Methods disclosed herein can comprise capturing DNA, such as cfDNA target regions. In some embodiments, the capturing comprises contacting the DNA with probes (e.g., oligonucleotides) specific for the target regions. Enrichment or capture may be performed on any sample or subsample described herein using any suitable approach known in the art.

[0378] In some embodiments, enrichment or capture is performed after attachment of adapters to sample molecules. In some embodiments, enrichment or capture is performed after a partitioning step. In some embodiments, enrichment or capture is performed after an amplification step. In some embodiments, sample molecules are partitioned, then adapters are attached, then sample molecules are amplified, and then the amplified molecules are subjected to enrichment or capture. The enriched or captured molecules may then be subjected to another amplification and then sequenced.

[0379] In some embodiments, the probes specific for the target regions comprise a capture moiety that facilitates the enrichment or capture of the DNA hybridized to the probes. In some embodiments, the capture moiety is biotin. In some such embodiments, streptavidin attached to a solid support, such as magnetic beads, is used to bind to the biotin. Nonspecifically bound DNA that does not comprise a target region is washed away from the captured DNA. In some embodiments, DNA is then dissociated from the probes and eluted from the solid support using salt washes or buffers comprising another DNA denaturing agent. In some embodiments, the probes are also eluted from the solid support by, e.g., disrupting the biotin-streptavidin interaction. In some embodiments, captured DNA is amplified following elution from the solid support. In some such embodiments, DNA comprising adapters is amplified using PCR primers that anneal to the adapters. In some embodiments, captured DNA is amplified while attached to the solid support. In some such embodiments, the amplification comprises use of a PCR primer that anneals to a sequence within an adapter and a PCR primer that anneals to a sequence within a probe annealed to the target region of the DNA.

[0380] In some embodiments, the methods herein comprise enriching for or capturing DNA comprising epigenetic and / or sequence-variable target regions. Such regions may be captured from an aliquot of a sample (e.g., a sample that has undergone attachment of adapters and amplification), while the step of partitioning the DNA with an agent that recognizes a modified cytosine, such as methyl cytosine, is performed on a separate aliquot of the sample. Enriching for or capturing DNA comprising epigenetic and / or sequence-variable target regions may comprise contacting the DNA with a first or second set of target-specific probes. Such target-specific probesmay have any of the features described herein for sets of target-specific probes, including but not limited to in the embodiments set forth above and the sections relating to probes below. Capturing may be performed on one or more subsamples prepared during methods disclosed herein. In some embodiments, DNA is captured from the first subsample or the second subsample, e.g., the first subsample and the second subsample. In some embodiments, the subsamples are differentially tagged (e.g., as described herein) and then pooled before undergoing capture. Exemplary methods for capturing DNA comprising epigenetic and / or sequence-variable target regions can be found in, e.g., WO 2020 / 160414, which is hereby incorporated by reference.

[0381] The capturing step may be performed using conditions suitable for specific nucleic acid hybridization, which generally depend to some extent on features of the probes such as length, base composition, etc. Those skilled in the art will be familiar with appropriate conditions given general knowledge in the art regarding nucleic acid hybridization. In some embodiments, complexes of target-specific probes and DNA are formed.

[0382] In some embodiments, methods described herein comprise capturing a plurality of sets of target regions of cfDNA obtained from a subject. The target regions may comprise differences depending on whether they originated from a tumor or from healthy cells or from a certain cell type. The capturing step produces a captured set of cfDNA molecules. In some embodiments, cfDNA molecules corresponding to a sequence-variable target region set are captured at a greater capture yield in the captured set of cfDNA molecules than cfDNA molecules corresponding to an epigenetic target region set. In some embodiments, a method described herein comprises contacting cfDNA obtained from a subject with a set of target-specific probes, wherein the set of target-specific probes is configured to capture cfDNA corresponding to the sequence-variable target region set at a greater capture yield than cfDNA corresponding to the epigenetic target region set. For additional discussion of capturing steps, capture yields, and related aspects, see W02020 / 160414, which is incorporated herein by reference for all purposes.

[0383] It can be beneficial to capture cfDNA corresponding to the sequence-variable target region set at a greater capture yield than cfDNA corresponding to the epigenetic target region set because a greater depth of sequencing may be necessary to analyze the sequence- variable target regions with sufficient confidence or accuracy than may be necessary to analyze the epigenetic target regions. The volume of data needed to determine fragmentation patterns (e.g., to test for perturbation of transcription start sites or CTCF binding sites) or fragment abundance (e.g., in hypermethylated and hypomethylated partitions) is generally less than the volume of data needed to determine the presence or absence of cancer-related sequence mutations. Capturing the target region sets at different yields can facilitate sequencing the targetregions to different depths of sequencing in the same sequencing run (e.g., using a pooled mixture and / or in the same sequencing cell).

[0384] In some embodiments, the DNA is amplified. In some embodiments, amplification is performed before the capturing step. In some embodiments, amplification is performed after the capturing step. In some embodiments, amplification is performed before and after the capturing step. In various embodiments, the methods further comprise sequencing the captured DNA, e.g., to different degrees of sequencing depth for the epigenetic and sequence-variable target region sets, consistent with the discussion herein.

[0385] In some embodiments, a capturing step is performed with probes for a sequence- variable target region set and probes for an epigenetic target region set in the same vessel at the same time, e.g., the probes for the sequence-variable and epigenetic target region sets are in the same composition. This approach provides a relatively streamlined workflow. In some embodiments, the concentration of the probes for the sequence-variable target region set is greater that the concentration of the probes for the epigenetic target region set.

[0386] Alternatively, a capturing step is performed with a sequence-variable target region probe set in a first vessel and with an epigenetic target region probe set in a second vessel, or a contacting step is performed with a sequence-variable target region probe set at a first time and a first vessel and an epigenetic target region probe set at a second time before or after the first time. This approach allows for preparation of separate first and second compositions comprising captured DNA corresponding to a sequence-variable target region set and captured DNA corresponding to an epigenetic target region set. The compositions can be processed separately as desired (e.g., to partition based on methylation as described herein) and pooled in appropriate proportions to provide material for further processing and analysis such as sequencing.

[0387] In some embodiments, adapters are included in the DNA as described herein. In some embodiments, tags, which may be or include barcodes, are included in the DNA. In some embodiments, such tags are included in adapters. Tags can facilitate identification of the origin of a nucleic acid. For example, barcodes can be used to allow the origin (e.g., subject) whence the DNA came to be identified following pooling of a plurality of samples for parallel sequencing. This may be done concurrently with an amplification procedure, e.g., by providing the barcodes in a 5’ portion of a primer, e.g., as described herein. In some embodiments, adapters and tags / barcodes are provided by the same primer or primer set. For example, the barcode may be located 3’ of the adapter and 5’ of the target-hybridizing portion of the primer. Alternatively, barcodes can be added by other approaches, such as ligation, optionally together with adapters in the same ligation substrate.

[0388] Additional details regarding amplification, tags, and barcodes are discussed herein, which can be combined to the extent practicable with any of these embodiments. F. Procedures that affect a first nucleobase in the DNA differently from a second nucleobase in the DNA or methylation-sensitive conversion methods

[0389] In some embodiments, methods disclosed herein comprise a step of subjecting DNA, or a subsample thereof, to a procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA, wherein the first nucleobase is a modified or unmodified nucleobase, the second nucleobase is a modified or unmodified nucleobase different from the first nucleobase, and the first nucleobase and the second nucleobase have the same base pairing specificity. In some embodiments, the procedure chemically converts the first or second nucleobase such that the base pairing specificity of the converted nucleobase is altered. In some embodiments, DNA is subjected to a procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA before library preparation using the DNA, before a first amplification of the DNA, before dividing the DNA into a plurality of subsamples, or any combination thereof. In certain embodiments, the DNA is subjected to the procedure before or after contacting the DNA with a methylation-sensitive nuclease.

[0390] In some embodiments, the procedure that affects a first nucleobase of the DNA differently from a second nucleobase of the DNA is performed prior to the sequencing and / or (a) prior to or after the selectively depleting the target nucleic acid comprising the wild-type sequence, the target nucleic acid comprising the converted nucleotide, or the target nucleic acid that does not comprise the converted nucleotide; (b) prior to the amplifying the selectively digested population of target nucleic acids; (c) prior to or after the partitioning the population of target nucleic acids into a plurality of subsamples; and / or (d) prior to or after a step of enriching for one or more sets of target regions of DNA.

[0391] In some embodiments, if the first nucleobase is a modified or unmodified adenine, then the second nucleobase is a modified or unmodified adenine; if the first nucleobase is a modified or unmodified cytosine, then the second nucleobase is a modified or unmodified cytosine; if the first nucleobase is a modified or unmodified guanine, then the second nucleobase is a modified or unmodified guanine; and if the first nucleobase is a modified or unmodified thymine, then the second nucleobase is a modified or unmodified thymine (where modified and unmodified uracil are encompassed within modified thymine for the purpose of this step).

[0392] In some embodiments, the first nucleobase is a modified or unmodified cytosine, then the second nucleobase is a modified or unmodified cytosine. For example, first nucleobase may comprise unmodified cytosine (C) and the second nucleobase may comprise one or more of 5-methylcytosine (mC) and 5-hydroxymethylcytosine (hmC). Alternatively, the second nucleobase may comprise C and the first nucleobase may comprise one or more of mC and hmC. Other combinations are also possible, such as where one of the first and second nucleobases comprises mC and the other comprises hmC.

[0393] In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA comprises bisulfite conversion. Treatment with bisulfite converts unmodified cytosine and certain modified cytosine nucleotides (e.g. 5-formyl cytosine (fC) or 5-carboxylcytosine (caC)) to uracil whereas other modified cytosines (e.g., 5- methylcytosine, 5-hydroxylmethylcystosine) are not converted. Thus, where bisulfite conversion is used, the first nucleobase comprises one or more of unmodified cytosine, 5-formyl cytosine, 5- carboxylcytosine, or other cytosine forms affected by bisulfite, and the second nucleobase may comprise one or more of mC and hmC, such as mC and optionally hmC. Sequencing of bisulfite- treated DNA identifies positions that are read as cytosine as being mC or hmC positions. Meanwhile, positions that are read as T are identified as being T or a bisulfite-susceptible form of C, such as unmodified cytosine, 5-formyl cytosine, or 5-carboxylcytosine. Performing bisulfite conversion, such as on a DNA sample as described herein, facilitates identifying positions containing mC or hmC using the sequence reads obtained from the exemplary sample. For an exemplary description of bisulfite conversion, see, e.g., Moss et al., Nat Commun.2018; 9: 5068.

[0394] In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA comprises oxidative bisulfite (Ox-BS) conversion. This procedure first converts hmC to fC, which is bisulfite susceptible, followed by bisulfite conversion. Thus, when oxidative bisulfite conversion is used, the first nucleobase comprises one or more of unmodified cytosine, fC, caC, hmC, or other cytosine forms affected by bisulfite, and the second nucleobase comprises mC. Sequencing of Ox-BS converted DNA identifies positions that are read as cytosine as being mC positions. Meanwhile, positions that are read as T are identified as being T, hmC, or a bisulfite-susceptible form of C, such as unmodified cytosine, fC, or hmC. Performing Ox-BS conversion, such as on a DNA sample as described herein, thus facilitates identifying positions containing mC using the sequence reads obtained from the sample. For an exemplary description of oxidative bisulfite conversion, see, e.g., Booth et al., Science 2012; 336: 934-937.

[0395] In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA comprises Tet-assisted bisulfite (TAB) conversion. In TAB conversion, hmC is protected from conversion and mC is oxidized in advance of bisulfite treatment, so that positions originally occupied by mC are converted to U while positions originally occupied by hmC remain as a protected form of cytosine. For example, as described in Yu et al., Cell 2012; 149: 1368-80, β-glucosyl transferase can be used to protect hmC (forming 5-glucosylhydroxymethylcytosine (ghmC)), then a TET protein such as mTet1 can be used to convert mC to caC, and then bisulfite treatment can be used to convert C and caC to U while ghmC remains unaffected.

[0396] Alternatively, a carbamoyltransferase enzyme, such as 5-hydroxymethylcytosine carbamoyltransferase as described in Yang et al., Bio-protocol, 2023; 12(17): e4496, can be used to protect hmC (by converting hmC to 5-carbamoyloxymethylcytosine (5cmC)), then a TET protein such as mTet1 or a TET2 comprising a T1372S mutation, can be used to convert mC to caC, and then bisulfite treatment can be used to convert C and caC to U while 5cmC remains unaffected. Thus, when TAB conversion is used, the first nucleobase comprises one or more of unmodified cytosine, fC, caC, mC, or other cytosine forms affected by bisulfite, and the second nucleobase comprises hmC. Sequencing of TAB-converted DNA identifies positions that are read as cytosine as being hmC positions. Meanwhile, positions that are read as T are identified as being T, mC, or a bisulfite-susceptible form of C, such as unmodified cytosine, fC, or caC. Performing TAB conversion, such as on a DNA sample as described herein, thus facilitates identifying positions containing hmC using the sequence reads obtained from the sample.

[0397] In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA comprises Tet-assisted conversion with a substituted borane reducing agent, optionally wherein the substituted borane reducing agent is 2- picoline borane, borane pyridine, tert-butylamine borane, or ammonia borane. In Tet-assisted pic- borane conversion with a substituted borane reducing agent conversion, a TET protein is used to convert mC and hmC to caC, without affecting unmodified C. caC, and fC if present, are then converted to dihydrouracil (DHU) by treatment with 2-picoline borane (pic-borane) or another substituted borane reducing agent such as borane pyridine, tert-butylamine borane, or ammonia borane, also without affecting unmodified C. See, e.g., Liu et al., Nature Biotechnology 2019; 37:424–429 (e.g., at Supplementary Fig.1 and Supplementary Note 7). Thus, when this type of conversion is used, the first nucleobase comprises one or more of 5mC, 5fC, 5caC, or 5hmC, and the second nucleobase comprises unmodified cytosine. DHU is read as a T in sequencing. Thus, when this type of conversion is used, the first nucleobase comprises one or more of mC, fC, caC,or hmC, and the second nucleobase comprises unmodified cytosine. Sequencing of the converted DNA identifies positions that are read as cytosine as being unmodified C positions. Meanwhile, positions that are read as T are identified as being T, mC, fC, caC, or hmC. Performing TAP conversion, such as on a DNA sample as described herein, thus facilitates identifying positions containing unmodified C using the sequence reads obtained from the sample. This procedure encompasses Tet-assisted pyridine borane sequencing (TAPS), described in further detail in Liu et al.2019, supra.

[0398] Alternatively, protection of hmC (e.g., using βGT or 5-hydroxymethylcytosine carbamoyltransferase) can be combined with Tet-assisted conversion with a substituted borane reducing agent, e.g. as described above. In this method (TAPS-β), 5hmC can be protected from conversion, for example through glucosylation using β-glucosyl transferase (βGT), forming 5- glucosylhydroxymethylcytosine (5ghmC), or through carbamoylation using 5- hydroxymethylcytosine carbamoyltransferase, forming 5cmC. This is described in Yu et al., Cell 2012; 149: 1368-80. Treatment with a TET protein, such as mTet1 or a TET2 comprising a T1372S mutation, then converts mC to caC but does not convert C, 5ghmC, or 5cmC.5caC is then converted to DHU by treatment with pic-borane or another substituted borane reducing agent such as borane pyridine, tert-butylamine borane, or ammonia borane, also without affecting ghmC, 5cmC, or unmodified C. Thus, when Tet-assisted conversion with a substituted borane reducing agent is used, the first nucleobase comprises mC, and the second nucleobase comprises one or more of unmodified cytosine or hmC, such as unmodified cytosine and optionally hmC, fC, and / or caC. Sequencing of the converted DNA identifies positions that are read as cytosine as being either hmC or unmodified C positions. Meanwhile, positions that are read as T are identified as being T, fC, caC, or mC. Performing TAPSβ conversion, such as on a DNA sample as described herein, thus facilitates distinguishing positions containing unmodified C or hmC on the one hand from positions containing mC using the sequence reads obtained from the sample. For an exemplary description of this type of conversion, see, e.g., Liu et al., Nature Biotechnology 2019; 37:424–429.5-hydroxymethylcytosine carbamoyltransferase is described in Yang et al., Bio-protocol, 2023; 12(17): e4496.

[0399] In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA comprises chemical-assisted conversion with a substituted borane reducing agent, optionally wherein the substituted borane reducing agent is 2- picoline borane, borane pyridine, tert-butylamine borane, or ammonia borane. In chemical- assisted conversion with a substituted borane reducing agent, an oxidizing agent such as potassium perruthenate (KRuO4) (also suitable for use in ox-BS conversion) is used to specificallyoxidize hmC to fC. Treatment with pic-borane or another substituted borane reducing agent such as borane pyridine, tert-butylamine borane, or ammonia borane converts fC and caC to DHU but does not affect mC or unmodified C. Thus, when this type of conversion is used, the first nucleobase comprises one or more of hmC, fC, and caC, and the second nucleobase comprises one or more of unmodified cytosine or mC, such as unmodified cytosine and optionally mC. Sequencing of the converted DNA identifies positions that are read as cytosine as being either mC or unmodified C positions. Meanwhile, positions that are read as T are identified as being T, fC, caC, or hmC. Performing this type of conversion, such as on a DNA sample as described herein, thus facilitates distinguishing positions containing unmodified C or mC on the one hand from positions containing hmC using the sequence reads obtained from the sample. For an exemplary description of this type of conversion, see, e.g., Liu et al., Nature Biotechnology 2019; 37:424–429.

[0400] In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA comprises APOBEC-coupled epigenetic (ACE) conversion. In ACE conversion, an AID / APOBEC family DNA deaminase enzyme such as APOBEC3A (A3A) is used to deaminate unmodified cytosine and mC without deaminating hmC, fC, or caC. Thus, when ACE conversion is used, the first nucleobase comprises unmodified C and / or mC (e.g., unmodified C and optionally mC), and the second nucleobase comprises hmC. Sequencing of ACE-converted DNA identifies positions that are read as cytosine as being hmC, fC, or caC positions. Meanwhile, positions that are read as T are identified as being T, unmodified C, or mC. Performing ACE conversion on a DNA sample as described herein thus facilitates distinguishing positions containing hmC from positions containing mC or unmodified C using the sequence reads obtained from the sample. For an exemplary description of ACE conversion, see, e.g., Schutsky et al., Nature Biotechnology 2018; 36: 1083–1090.

[0401] In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA comprises enzymatic conversion of the first nucleobase, e.g., as in EM-Seq. See, e.g., Vaisvila R, et al. (2019) EM-seq: Detection of DNA methylation at single base resolution from picograms of DNA. bioRxiv; DOI: 10.1101 / 2019.12.20.884692, available at www.biorxiv.org / content / 10.1101 / 2019.12.20.884692v1. For example, TET2 and T4-βGT or 5- hydroxymethylcytosine carbamoyltransferase (described in Yang et al., Bio-protocol, 2023; 12(17): e4496) can be used to convert 5mC and 5hmC into substrates that cannot be deaminated by a deaminase (e.g., APOBEC3A), and then a deaminase (e.g., APOBEC3A) can be used to deaminate unmodified cytosines converting them to uracils.

[0402] In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA comprises enzymatic conversion of the first nucleobase using a non-specific, modification-sensitive double-stranded DNA deaminase, e.g., as in SEM-seq. See, e.g., Vaisvila et al. (2023) Discovery of novel DNA cytosine deaminase activities enables a nondestructive single-enzyme methylation sequencing method for base resolution high-coverage methylome mapping of cell-free and ultra-low input DNA. bioRxiv; DOI: 10.1101 / 2023.06.29.547047, available at https: / / www.biorxiv.org / content / 10.1101 / 2023.06.29.547047v1. SEM-Seq employs a non- specific, modification-sensitive double-stranded DNA deaminase (MsddA) in a nondestructive single-enzyme 5-methylctyosine sequencing (SEM-seq) method that deaminates unmodified cytosines. Accordingly, SEM-seq does not require the TET2 and T4-βGT or 5- hydroxymethylcytosine carbamoyltransferase protection and denaturing steps that are of use, e.g., in APOEC3A-based protocols. Additionally, MsddA does not deaminate 5-formylated cytosines (5fC) or 5-carboxylated cytosines (5caC). In SEM-seq, unmodified cytosines in the DNA are deaminated to uracil and is read as “T” during sequencing. Modified cytosines (e.g., 5mC) are not converted and are read as “C” during sequencing. Cytosines that are read as thymines are identified as unmodified (e.g., unmethylated) cytosines or as thymines in the DNA. Performing SEM-seq conversion thus facilitates identifying positions containing 5mC using the sequence reads obtained. In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA comprises enzymatic conversion of the first nucleobase using MsddA.

[0403] In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA of the first subsample converts a modified nucleoside. In some embodiments, the conversion procedure which converts a modified nucleosides comprises enzymatic conversion, such as DM-seq, for example, as described in WO2023 / 288222A1. In DM-seq, unmodified cytosines in the DNA are enzymatically protected from a subsequent deamination step wherein 5mC in 5mCpG is converted to T. The enzymatically protected unmodified (e.g., unmethylated) cytosines are not converted and are read as “C” during sequencing. Cytosines that are read as thymines (in a CpG context) are identified as methylated cytosines in the DNA. Thus, when this type of conversion is used, the first nucleobase comprises unmodified (such as unmethylated) cytosine, and the second nucleobase comprises modified (such as methylated) cytosine. Sequencing of the converted DNA identifies positions that are read as cytosine as being unmodified C positions. Meanwhile, positions that are read as T are identifiedas being T or 5mC. Performing DM-seq conversion thus facilitates identifying positions containing 5mC using the sequence reads obtained.

[0404] Exemplary cytosine deaminases for use herein include APOBEC enzymes, for example, APOBEC3A. Generally, AID / APOBEC family DNA deaminase enzymes such as APOBEC3A (A3A) are used to deaminate (unprotected) unmodified cytosine and 5mC. For an exemplary description of APOBEC conversion, see, e.g., Schutsky et al., Nature Biotechnology 2018; 36: 1083–1090.

[0405] The enzymatic protection of unmodified cytosines in the DNA comprises addition of a protective group to the unmodified cytosines. Such protective groups can comprise an alkyl group, an alkyne group, a carboxyl group, a carboxyalkyl group, an amino group, a hydroxymethyl group, a glucosyl group, a glucosylhydroxymethyl group, an isopropyl group, or a dye. For example, DNA can be treated with a methyltransferase, such as a CpG-specific methyltransferase, which adds the protective group to unmodified cytosines. The term methyltransferase is used broadly herein to refer to enzymes capable of transferring a methyl or substituted methyl (e.g.,carboxymethyl) to a substrate (e.g., a cytosine in a nucleic acid). In some embodiments, the DNA is contacted with a CpG-specific DNA methyltransferase (MTase), such as a CpG-specific carboxymethyltransferase (CxMTase), and a substituted methyl donor, such as a carboxymethyl donor (e.g., carboxymethyl-S-adenosyl-L-methionine). See, e.g., WO2021 / 236778A2. In particular embodiments, the CxMTase can facilitate the addition of a protective carboxymethyl group to an unmethylated cytosine. In some embodiments, the unmethylated cytosine is unmodified cytosine. The carboxymethyl group can prevent deamination of the cytosine during a deamination step (such as a deamination step using an APOBEC enzyme, such as A3A). Substituted methyl or carboxymethyl donors useful in the disclosed methods include but are not limited to, S-adenosyl-L-methionine (SAM) analogs, optionally wherein the SAM analog is carboxy-S-adenosyl-L-methionine (CxSAM). SAM analogs are described, for example, in WO2022 / 197593A1. The MTase may be, for example, a CpG methyltransferase from Spiroplasma sp. strain MQ1 (M.SssI), DNA-methyltransferase 1 (DNMT1), DNA- methyltransferase 3 alpha (DNMT3A), DNA-methyltransferase 3 beta (DNMT3B), or DNA adenine methyltransferase (Dam). The CxMTase may be a CpG methyltransferase from Mycoplasma penetrans (M.MpeI). In a particular embodiment, the methyltransferase enzyme is a variant of M.MpeI, wherein the amino acid corresponding to position 374 is R or K.

[0406] In one embodiment, the methyltransferase enzyme is a variant of M.MpeI having an N374R substitution or an N374K substitution. The methyltransferase variant can further comprise one or more amino acid substitutions selected from a) substitution of one or bothresidues T300 and E305 with S, A, G, Q, D, or N; b) substitution of one or more residues A323, N306, and Y299 with a positively charged amino acid selected from K, R or H; and / or c) substitution of S323 with A, G, K, R or H, which may enhance the activity of the enzyme.

[0407] Optionally, the conversion procedure further includes enzymatic protection of 5hmCs, such as by glucosylation of the 5hmCs (e.g., using βGT) or by carbamoylation of the 5hmCs (e.g., using 5-hydroxymethylcytosine carbamoyltransferase), in the DNA prior to the deamination of unprotected modified cytosines. In this method, 5hmC can be protected from conversion, for example through glucosylation using β-glucosyl transferase (βGT), forming (5- glucosylhydroxymethylcytosine) 5ghmC, or through carbamoylation using 5- hydroxymethylcytosine carbamoyltransferase, forming 5cmC. This is described, for example, in Yu et al., Cell 2012; 149: 1368-80, and in Yang et al., Bio-protocol, 2023; 12(17): e4496. Glucosylation or carbamoylation of 5hmC can reduce or eliminate deamination of 5hmC by a deaminase such as APOBEC3A. Treatment with an MTase or CxMTase then adds a protecting group to unmodified (unmethylated) cytosines in the DNA. 5mC (but not protected, unmodified cytosine and not 5ghmC or 5cmC) is then deaminated (converted to T in the case of 5mC) by treatment with a deaminase, for example, an APOBEC enzyme (such as APOBEC3A). Sequencing of the converted DNA identifies positions that are read as cytosine as being either 5hmC or unmodified C positions. Meanwhile, positions that are read as T are identified as being T or 5mC. Performing DM-seq conversion with glucosylation of 5hmC on a sample as described herein thus facilitates distinguishing positions containing unmodified C or 5hmC on the one hand from positions containing 5mC using the sequence reads obtained.

[0408] Also provided herein are methods in which alternative base conversion schemes are used. For example, unmethylated cytosines can be left intact while methylated cytosines and hydroxymethylcytosines are converted to a base read as a thymine (e.g., uracil, thymine, or dihydrouracil).

[0409] In some embodiments, methylating a cytosine in at least one first complementary strand or second complementary strand comprises contacting the cytosine with a methyltransferase such as DNMT1 or DNMT5. In such embodiments, the step of oxidizing a 5- hydroxymethylated cytosine to 5-formylcytosine (such as by contacting the 5-hydroxymethyl cytosine in a first strand and a second strand with KRuO4) can be optional.

[0410] In some embodiments, converting the modified cytosine in at least one first or second strand to a thymine or a base read as thymine comprises oxidizing a hydroxymethyl cytosine, e.g., the hydroxymethyl cytosine is oxidized to formylcytosine. In some embodiments,oxidizing the hydroxymethyl cytosine to formylcytosine comprises contacting the hydroxymethyl cytosine with a ruthenate, such as potassium ruthenate (KRuO4).

[0411] In some embodiments, the modified cytosine is converted to thymine, uracil, or dihydrouracil. In any such embodiments, amplification methods may comprise uracil- and / or dihydrouracil-tolerant amplification methods, such as PCR using a uracil- and / or dihydrouracil- tolerant DNA polymerase.

[0412] In some embodiments, the method comprises converting a formylcytosine and / or a methylcytosine to carboxylcytosine as part of converting the modified cytosine in at least one first or second strand to a thymine or a base read as thymine. For example, converting the formylcytosine and / or the methylcytosine to carboxylcytosine can comprise contacting the formylcytosine and / or the methylcytosine with a TET enzyme, such as TET1, TET2, TET3, or a TET2 comprising a T1372S mutation. In some embodiments, the method comprises reducing the carboxylcytosine as part of converting the modified cytosine in at least one first or second strand to a thymine or a base read as thymine, and / or the carboxylcytosine is reduced to dihydrouracil. In some embodiments, reducing the carboxylcytosine comprises contacting the carboxylcytosine with a borane or borohydride reducing agent.

[0413] In some embodiments, the borane or borohydride reducing agent comprises pyridine borane, 2-picoline borane, borane, tert-butylamine borane, ammonia borane, sodium borohydride, sodium cyanoborohydride (NaBH3CN), lithium borohydride (LiBH4), ethylenediamine borane, dimethylamine borane, sodium triacetoxyborohydride, morpholine borane, 4-methylmorpholine borane, trimethylamine borane, dicyclohexylamine borane, or a salt thereof. In other embodiments, the reducing agent comprises lithium aluminum hydride, sodium amalgam, amalgam, sulfur dioxide, dithionate, thiosulfate, iodide, hydrogen peroxide, hydrazine, diisobutylaluminum hydride, oxalic acid, carbon monoxide, cyanide, ascorbic acid, formic acid, dithiothreitol, beta-mercaptoethanol, or any combination thereof.

[0414] Various TET enzymes may be used in the disclosed methods as appropriate. In some embodiments, the one or more TET enzymes comprise TETv. TETv is described in US Patent 10,260,088. In some embodiments, the one or more TET enzymes comprise TETcd. TETcd is described in US Patent 10,260,088. In some embodiments, the one or more TET enzymes comprise TET1. In some embodiments, the one or more TET enzymes comprise TET2. TET2 may be expressed and used as a fragment comprising TET2 residues 1129-1480 joined to TET2 residues 1844-1936 by a linker as described, e.g., in US Patent 10,961,525. In some embodiments, the one or more TET enzymes comprise TET1 and TET2. In some embodiments, the one or more TET enzymes comprise a V1900 TET mutant, such as a V1900A, V1900C,V1900G, V1900I, or V1900P TET mutant. In some embodiments, the one or more TET enzymes comprise a V1900 TET2 mutant, such as a V1900A, V1900C, V1900G, V1900I, or V1900P TET2 mutant. It can be beneficial to use a TET enzyme that maximizes formation of 5-carboxylcytosine (5-caC) relative to less oxidized modified cytosines, particularly 5-formylcytosine, because 5-caC is not a substrate for enzymatic deamination, e.g., by APOBEC enzymes such as APOBEC3A. Maximizing formation of 5-caC thus reduces the risk of false calls in which a base is identified as unmethylated because it underwent deamination even though it was methylated (or hydroxymethylated) in the original sample. Accordingly, in some embodiments, the TET enzyme comprises a mutation that increases formation of 5-caC. In some embodiments, the one or more TET enzymes comprise a TET2 enzyme comprising a T1372S mutation, such as TET2-CS- T1372S and TET2-CD-T1372S. A TET2 comprising a T1372S mutation is described in US Patent 10,961,525 and may be expressed and used as a fragment comprising TET2 residues 1129-1480 joined to TET2 residues 1844-1936 by a linker. Position 1372 of TET2 corresponds to position 258 of SEQ ID NO: 21 (wild type TET2 catalytic domain) of US Patent 10,961,525. Thus, the sequence of a T1372S TET2 catalytic domain may be obtained by changing the threonine at position 258 of SEQ ID NO: 21 of US Patent 10,961,525 to serine. TET2 comprising a T1372S mutation is also described in Liu et al., Nat Chem Biol. 2017 February; 13(2): 181–187. As demonstrated in Liu et al., TET2 comprising a T1372S mutation can more efficiently oxidize 5mC to produce 5-carboxylcytosine (5caC) than other versions of TET2 such as TET2 lacking a T1372S mutation. In some embodiments, the TET2 enzyme is a human TET2 enzyme comprising a T1372S mutation. Exemplary mutations are set forth above. “A mutation that increases formation of 5-caC” means that the TET enzyme having the mutation produces more 5-caC than a TET enzyme that lacks the mutation but is otherwise identical.5-caC production can be measured as described, e.g., in Liu et al., Nat Chem Biol 13:181-187 (2017) (see Online Methods section, TET reactions in vitro subsection, “driving” conditions). Any variants and / or mutants described in Liu et al. (2017) can be used in the disclosed methods as appropriate.

[0415] Provided herein is a method comprising contacting DNA contacting DNA with a mutant TET2 enzyme (e.g. comprising a V1900A, V1900C, V1900G, V1900I, V1900P, or T1372S mutation) to oxidize 5-methylcytosine (5mC) and / or 5-hydroxymethylcytosine (5hmC) present in the DNA to 5-carboxycytosine (5caC), subsequently contacting at least a portion of the DNA with a substituted borane reducing agent, thereby converting 5-caC in the DNA to dihydrouracil (DHU), thereby producing treated DNA, and sequencing at least a portion of the treated DNA.

[0416] In some embodiments, the procedure that affects a first nucleobase in the DNA differently from a second nucleobase in the DNA comprises separating DNA originally comprisingthe first nucleobase from DNA not originally comprising the first nucleobase. In some such embodiments, the first nucleobase is hmC. DNA originally comprising the first nucleobase may be separated from other DNA using a labeling procedure comprising biotinylating positions that originally comprised the first nucleobase. In some embodiments, the first nucleobase is first derivatized with an azide-containing moiety, such as a glucosyl-azide containing moiety. The azide-containing moiety then may serve as a reagent for attaching biotin, e.g., through Huisgen cycloaddition chemistry. Then, the DNA originally comprising the first nucleobase, now biotinylated, can be separated from DNA not originally comprising the first nucleobase using a biotin-binding agent, such as avidin, neutravidin (deglycosylated avidin with an isoelectric point of about 6.3), or streptavidin. An example of a procedure for separating DNA originally comprising the first nucleobase from DNA not originally comprising the first nucleobase is hmC-seal, which labels hmC to form β-6-azide-glucosyl-5-hydroxymethylcytosine and then attaches a biotin moiety through Huisgen cycloaddition, followed by separation of the biotinylated DNA from other DNA using a biotin-binding agent. For an exemplary description of hmC-seal, see, e.g., Han et al., Mol. Cell 2016; 63: 711-719. This approach is useful for identifying fragments that include one or more hmC nucleobases.

[0417] In some embodiments, following such a separation, the method further comprises differentially tagging each of the DNA originally comprising the first nucleobase, the DNA not originally comprising the first nucleobase. The method may further comprise pooling the DNA originally comprising the first nucleobase and the DNA not originally comprising the first nucleobase following differential tagging. The DNA originally comprising the first nucleobase and the DNA not originally comprising the first nucleobase may then be used in downstream analyses. For example, the pooled DNA originally comprising the first nucleobase and the DNA not originally comprising the first nucleobase may be sequenced in the same sequencing cell (such as after being subjected to further treatments, such as those described herein) while retaining the ability to resolve whether a given read came from a molecule of DNA originally comprising the first nucleobase or DNA not originally comprising the first nucleobase using the differential tags.

[0418] In some embodiments, the first nucleobase is a modified or unmodified adenine, and the second nucleobase is a modified or unmodified adenine. In some embodiments, the modified adenine is N6-methyladenine (mA). In some embodiments, the modified adenine is one or more of N6-methyladenine (mA), N6-hydroxymethyladenine (hmA), or N6-formyladenine (fA).

[0419] Techniques comprising partitioning based on methylation status or methylated DNA immunoprecipitation (MeDIP) can be used to separate DNA containing modified bases such as mC, mA, caC (which may be generated by oxidation of mC or hmC with Tet2, e.g., beforeenzymatic conversion of unmodified C to U, e.g., using a deaminase such as APOBEC3A), or dihydrouracil from other DNA. See, e.g., Kumar et al., Frontiers Genet.2018; 9: 640; Greer et al., Cell 2015; 161: 868-878. An antibody specific for mA is described in Sun et al., Bioessays 2015; 37:1155-62. Antibodies for various modified nucleobases, such as mC, caC, and forms of thymine / uracil including dihydrouracil or halogenated forms such as 5-bromouracil, are commercially available. Various modified bases can also be detected based on alterations in their base pairing specificity. For example, hypoxanthine is a modified form of adenine that can result from deamination and is read in sequencing as a G. See, e.g., US Patent 8,486,630; Brown, Genomes, 2nd Ed., John Wiley & Sons, Inc., New York, N.Y., 2002, chapter 14, “Mutation, Repair, and Recombination.” G. Captured Set; Target Regions

[0420] In some embodiments, nucleic acids captured or enriched using a method described herein comprise captured DNA, such as one or more captured sets of DNA. In some embodiments, the captured DNA comprise target regions that are differentially methylated in different immune cell types. In some embodiments, the immune cell types comprise rare or closely related immune cell types, such as activated and nai...

Claims

CLAIMS WHAT IS CLAIMED IS:

1. A method comprising: obtaining first training data from first samples derived from first subjects in which a tumor is detected, the first training data including first nucleotide sequence representations corresponding to first nucleic acid molecules included in the first samples; obtaining second training data from second samples derived from second subjects in which a tumor is not detected, the second training data including second nucleotide sequence representations corresponding to second nucleic acid molecules included in the second samples; performing a training process for one or more machine learning models, wherein the training process includes: determining first quantitative measures based on first values for features of the first nucleotide sequence representations; determining second quantitative measures based on second values for the features of the second nucleotide sequence representations; and determining, based on the first quantitative measures and the second quantitative measures, one or more threshold quantitative measures to identify nucleotide sequence representations to provide to a machine learning classification model to determine an indication of a tumor-related biological condition; obtaining additional nucleotide sequence representations derived from an additional subject and corresponding to additional nucleic acid molecules included in an additional sample obtained from the additional subject; determining additional quantitative measures based on additional values for the features of the additional nucleotide sequence representations; determining a subset of the additional nucleotide sequence representations for which the additional quantitative measures correspond to the one or more threshold quantitative measures; and providing the subset of the additional nucleotide sequence representations to the machine learning classification model to determine the indication of the tumor-related biological condition for the additional subject.

2. The method of claim 1, comprising:for individual genomic regions, determining, based on differences between the first values and the second values, probabilities that specified sets of values of the features for nucleotide sequence representations correspond to samples obtained from subjects in which the tumor-related biological condition is present.

3. The method of claim 2, comprising: determining, based on the probabilities, that nucleic acid molecules aligned with one or more genomic regions have less than a threshold probability of identifying samples obtained from subjects in which the tumor-related biological condition is present; and producing a diagnostic test that does not enrich the one or more genomic regions.

4. The method of claim 2, comprising: generating a probability map that indicates the probabilities for the individual genomic regions.

5. The method of any one of claims 2-4, wherein: the first training data includes first methylation data that indicates a first number of methylated cytosine-guanine dinucleotides (CpGs) present in the first nucleic acid molecules; and the second training data includes second methylation data that indicates a second number of methylated CpGs present in the second nucleic acid molecules; and the method comprises: determining the first values for the features with respect to the first nucleotide sequence representations based on first characteristics of the first nucleotide sequence representations and the first methylation data; and determining the second values for the features with respect to the second nucleotide sequence representations based on second characteristics of the second nucleotide sequence representations and the second methylation data.

6. The method of claim 5, wherein determining the first values for the features includes determining a first number of CpGs within one or more regions of the first nucleotide sequence representations and determining the second values for the features includes determining a second number of CpGs within the one or more regions of the second nucleotide sequence representations.

7. The method of claim 6, wherein determining the first values for the features includes determining a first amount of methylated CpGs within the one or more regions of the first nucleotide sequence representations and determining the second values for the features includes determining a second amount of methylated CpGs within the one or more regions of the second nucleotide sequence representations.

8. The method of claim 7, wherein the first amount of methylated CpGs within the one or more regions includes a first count of individual methylated CpGs for the first nucleotide sequence representations within the one or more regions and the second amount of methylated CpGs within the one or more regions includes a second count of individual methylated CpGs for the second nucleotide sequence representations within the one or more regions.

9. The method of claim 7, wherein the first amount of methylated CpGs within the one or more regions for the first nucleotide sequence representations corresponds to a first range of methylated CpGs and the second amount of methylated CpGs within the one or more regions for the second nucleotide sequence representations corresponds to a second range of methylated CpGs.

10. The method of claim 5, wherein determining the first values for the features includes determining a first number of CpGs within individual first nucleic acid molecules and determining the second values for the features includes determining a second number of CpGs within individual second nucleic acid molecules.

11. The method of claim 5, wherein determining the first values for the features includes determining a first amount of methylated CpGs within individual first nucleic acid molecules and determining the second values for the features includes determining a second amount of methylated CpGs within individual second nucleic acid molecules.

12. The method of claim 11, wherein the first amount of methylated CpGs includes a first count of individual methylated CpGs within the first nucleic acid molecules and the second amount of methylated CpGs includes a second count of individual methylated CpGs within the second nucleic acid molecules.

13. The method of claim 11, wherein the first amount of methylated CpGs corresponds to a first range of methylated CpGs within the first nucleic acid molecules and the second amount of methylated CpGs corresponds to a second range of methylated CpGs within the second nucleic acid molecules.

14. The method of any one of claims 5-13, wherein determining the first values for the features and the second values for the features includes determining, a first number of first sequencing reads that correspond to individual first nucleic acid molecules included in the first samples and determining a number of second sequencing reads that correspond to individual second nucleic acid molecules included in the second samples.

15. The method of any one of claims 5-14, wherein determining the first values for the features includes determining a first number of restriction enzyme cut sites for the first nucleotide sequence representations and determining the second values for the features includes determining a second number of restriction enzyme cut sites for the second nucleotide sequence representations.

16. The method of any one of claims 5-15, wherein determining the first values for the features includes determining first lengths for individual first nucleotide sequence representations and determining the second values for the features includes determining second lengths for individual second nucleotide sequence representations.

17. The method of any one of claims 5-16, wherein determining the first values for the features and the second values for the features includes determining, for individual first nucleotide sequence representations and individual second nucleotide sequence representations, an offset of a start position with respect to a genomic region in which the individual first nucleotide sequence representations and individual second nucleotide sequence representations are located.

18. The method of any one of claims 5-17, wherein determining the first values for the features and the second values for the features includes determining, for a given genomic region, a number of the first nucleic acid molecules and a number of the second nucleic acid molecules aligned with the given genomic region in relation to a number of the first nucleic acidmolecules and a number of the second nucleic acid molecules aligned with one or more control genomic regions.

19. The method of any one of claims 1-18, wherein the machine learning classification model includes a random forests model.

20. The method of any one of claims 1-19, wherein the first quantitative measures and the second quantitative measures are determined using a logistic regression model.

21. The method of any one of claims 1-20, wherein the first values for the features with respect to the first nucleotide sequence representations, the second values for the features with respect to the second nucleotide sequence representations, and the additional values for the features with respect to the additional nucleotide sequence representations are represented as vectors that are provided to the machine learning classification model.

22. The method of any one of claims 1-21, wherein the indication of the tumor-related biological condition indicates that a tumor is present or a tumor is absent in a subject.

23. The method of any one of claims 1-21, wherein the indication of the tumor-related biological condition corresponds to tumor fraction for a given sample.

24. The method of any one of claims 1-21, wherein the indication of the tumor-related biological condition corresponds to reoccurrence of cancer within a subject.

25. The method of any one of claims 1-21, wherein the indication of the tumor-related biological condition indicates a responsiveness of a subject to one or more treatments for the tumor-related biological condition.

26. The method of any one of claims 1-25, wherein the first quantitative measures and the second quantitative measures indicate differences between nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present and nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is not present.

27. The method of claim 26, wherein the first quantitative measures and the second quantitative measures indicate probabilities of nucleotide sequence representations corresponding to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present.

28. The method of claim 1, wherein the one or more threshold quantitative measures comprise molecule selection data indicative of nucleic acid molecules derived from subjects in which a tumor-related biological condition is present; and the method comprises: obtaining a plurality of further nucleotide sequence representations derived from one or more further samples of a further subject, the plurality of further nucleotide sequence representations corresponding to further nucleic acid molecules included in the one or more further samples; determining further quantitative measures based on further values for one or more features of the plurality of further nucleotide sequence representations; determining a subset of the further plurality of nucleotide sequence representations for which the further quantitative measures correspond to the one or more threshold quantitative measures; and executing a computational model to determine to determine an indication of a somatic variant being present in the one or more further samples, wherein weightings of the subset of the further plurality of nucleotide sequence representations within the computational model are modified to determine the indication of the somatic variant being present.

29. The method of claim 28, wherein the one or more threshold quantitative measures include a threshold quantitative measure indicating at least a threshold probability of a tumor- being present in a subject.

30. The method of claim 29, wherein the subset of the further quantitative measures are greater than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the further plurality of nucleotide sequence representations are upweighted with respect to initial weightings of the further plurality of nucleotide sequence representations.

31. The method of claim 29, wherein the subset of the further quantitative measures are less than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the further plurality of nucleotide sequence representations are downweighted with respect to initial weightings of the further plurality of nucleotide sequence representations 32. The method of any one of claims 28-31, wherein the somatic variant includes a copy number deletion.

33. The method of any one of claims 28-32, wherein the somatic variant includes a copy number duplication.

34. The method of any one of claims 28-33, wherein the somatic variant includes a single nucleotide variant.

35. A method comprising: obtaining molecule selection data generated from a training process of one or more machine learning models, the molecule selection data including one or more threshold quantitative measures indicative of nucleic acid molecules derived from subjects in which a tumor-related biological condition is present; obtaining a plurality of nucleotide sequence representations derived from one or more samples of a subject, the plurality of nucleotide sequence representations corresponding to nucleic acid molecules included in the one or more samples; determining quantitative measures based on values for features of the plurality of nucleotide sequence representations; determining a subset of the plurality of nucleotide sequence representations for which the quantitative measures correspond to the one or more threshold quantitative measures; providing the subset of the plurality of nucleotide sequence representations to a machine learning classification model; and determining, by the machine learning classification model, an indication of the tumor- related biological condition for the subject.

36. The method of claim 35, comprising obtaining methylation data that indicates a number of methylated cytosine-guanine dinucleotides (CpGs) present in the nucleic acid molecules; and the method comprises: determining the values for the features with respect to the plurality of nucleotide sequence representations based on characteristics of the plurality of nucleotide sequence representations and the methylation data.

37. The method of claim 36, wherein determining the values for the features includes determining a number of CpGs within one or more regions of the plurality of nucleotide sequence representations.

38. The method of claim 37, wherein determining the values for the features includes determining an amount of methylated CpGs within the one or more regions of the plurality of nucleotide sequence representations.

39. The method of claim 38, wherein the amount of methylated CpGs within the one or more regions includes a count of individual methylated CpGs for the plurality of nucleotide sequence representations within the one or more regions.

40. The method of claim 38, wherein the amount of methylated CpGs within the one or more regions for the plurality of nucleotide sequence representations corresponds to a range of methylated CpGs.

41. The method of claim 35, wherein determining the values for the features includes determining a number of CpGs within individual nucleic acid molecules.

42. The method of claim 35, wherein determining the values for the features includes determining an amount of methylated CpGs within individual nucleic acid molecules.

43. The method of claim 42, wherein the amount of methylated CpGs includes a count of individual methylated CpGs within the nucleic acid molecules.

44. The method of claim 40, wherein the amount of methylated CpGs corresponds to a range of methylated CpGs within the nucleic acid molecules.

45. The method of any one of claims 36-44, wherein determining the values for the features includes determining a number of sequencing reads that correspond to individual nucleic acid molecules included in the one or more samples.

46. The method of any one of claims 36-25, wherein determining the values for the features includes determining a number of restriction enzyme cut sites for the plurality of nucleotide sequence representations.

47. The method of any one of claims 36-46, wherein determining the values for the features includes determining lengths for individual nucleotide sequence representations.

48. The method of any one of claims 36-47, wherein determining the values for the features includes determining, for individual nucleotide sequence representations an offset of a start position with respect to a genomic region in which the individual nucleotide sequence representations are located.

49. The method of any one of claims 36-48, wherein determining the values for the features includes determining, for a given genomic region, a number of the nucleic acid molecules aligned with the given genomic region in relation to a number of the nucleic acid molecules.

50. The method of any one of claims 35-49, wherein the machine learning classification model includes a random forests model.

51. The method of any one of claims 35-50, wherein the quantitative measures are determined using a logistic regression model.

52. The method of any one of claims 35-51, wherein the values for the features with respect to the plurality of nucleotide sequence representations are represented as vectors that are provided to the machine learning classification model.

53. The method of any one of claims 35-52, wherein the indication of the tumor-related biological condition indicates that a tumor is present or a tumor is absent in the subject.

54. The method of any one of claims 35-53, wherein the indication of the tumor-related biological condition corresponds to tumor fraction for the one or more samples.

55. The method of any one of claims 35-54, wherein the indication of the tumor-related biological condition corresponds to reoccurrence of cancer within the subject.

56. The method of any one of claims 35-55, wherein the indication of the tumor-related biological condition indicates a responsiveness of the subject to one or more treatments for the tumor-related biological condition.

57. The method of any one of claims 35-56, wherein the quantitative measures indicate differences between nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present and nucleotide sequence representations that correspond to nucleic acid molecules derived from subjects in which the tumor-related biological condition is not present.

58. The method of claim 57, wherein the quantitative measures indicate probabilities of nucleotide sequence representations corresponding to nucleic acid molecules derived from subjects in which the tumor-related biological condition is present.

59. The method of claim 35, comprising: obtaining a plurality of additional nucleotide sequence representations derived from one or more additional samples of an additional subject, the plurality of additional nucleotide sequence representations corresponding to additional nucleic acid molecules included in the one or more additional samples; determining additional quantitative measures based on additional values for one or more features of the plurality of additional nucleotide sequence representations; determining a subset of the additional plurality of nucleotide sequence representations for which the additional quantitative measures correspond to the one or more threshold quantitative measures; andexecuting a computational model to determine to determine an indication of a somatic variant being present in the one or more additional samples, wherein weightings of the subset of the additional plurality of nucleotide sequence representations within the computational model are modified to determine the indication of the somatic variant being present.

60. The method of claim 59, wherein the one or more threshold quantitative measures include a threshold quantitative measure indicating at least a threshold probability of a tumor- being present in a subject.

61. The method of claim 60, wherein the subset of the additional quantitative measures are greater than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the additional plurality of nucleotide sequence representations are upweighted with respect to initial weightings of the additional plurality of nucleotide sequence representations.

62. The method of claim 60, wherein the subset of the additional quantitative measures are less than a threshold quantitative measure of the one or more threshold quantitative measures, and the weightings of the subset of the additional plurality of nucleotide sequence representations are downweighted with respect to initial weightings of the additional plurality of nucleotide sequence representations 63. The method of any one of claims 59-62, wherein the somatic variant includes a copy number deletion.

64. The method of any one of claims 59-63, wherein the somatic variant includes a copy number duplication.

65. The method of any one of claims 59-64, wherein the somatic variant includes a single nucleotide variant.

Images