Prediction and determination of the effectiveness of lung cancer treatment in patients

Abstract

Methods are disclosed herein for monitoring a lung cancer patient during lung cancer treatment to determine whether the patient is responding to treatment, and for predicting whether a lung cancer patient is likely to respond to treatment prior to initiating lung cancer treatment. These methods include the generation and analysis of hydroxymethylation signatures, where, when monitoring efficacy, the patient 5hmC signature obtained during treatment is compared to a baseline 5hmC signature, whereas when predicting efficacy, the patient 5hmC signature is compared to a reference 5hmC profile. Analysis of 5hmC levels at certain hydroxymethylation biomarker loci indicates whether a patient is likely to benefit from a particular lung cancer treatment or is likely to continue to benefit from a particular lung cancer treatment.

Description

Technical Field

[0001] Technical Field The present invention generally relates to the treatment of cancer, and more specifically to the management of patients undergoing lung cancer treatment such as immunotherapy. The present invention provides methods for predicting the effectiveness of treating lung cancer patients with a particular treatment and for monitoring the response of lung cancer patients undergoing treatment.

Background Art

[0002] Background Methods that enable monitoring of the response to cancer treatment are extremely important for making information-based decisions regarding whether to continue cancer treatment, for avoiding unnecessary exposure of patients to potentially multiple side effects, and for providing the most effective treatment as early as possible during the course of the disease. When monitored with conventional imaging methods, in fact, when the tumor is enlarged due to a favorable drug effect (which may be due to an immune response, for example), it may appear as if the size of the tumor is increasing (a phenomenon known as pseudoprogression). Thus, immunotherapy, in particular, presents difficulties in detecting early the response of patients to treatment. Monitoring immunotherapy response to detect potential resistance is important to be able to discontinue the administration of ineffective immunotherapy (preventing any immunotherapy-related adverse events), thereby enabling a rapid switch to another treatment as early as possible. There is a critical need for an effective method for predicting and monitoring the response of cancer patients to treatment, particularly immunotherapy.

[0003] One approach that others have tried to address this problem is through identifying and monitoring tumor-specific mutations in ctDNA. However, this approach requires prior knowledge of the tumor mutation status and suffers from the problems of low levels of ctDNA and genomic (mutation) plasticity observed in patients.

[0004] Lung cancer is the most common cancer worldwide and is also the leading cause of cancer-related deaths globally, estimated to claim 1.8 million lives each year. See Sung et al., (2021), “Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries,” Ca Cancer J Clin 71, 209-249. Most lung cancer patients are diagnosed with advanced disease, and 85% to 90% of all lung cancer cases are characterized as non-small cell lung cancer (NSCLC). Among the treatment options available for NSCLC, immunotherapy has strengthened its position in the management of advanced disease, achieving better patient outcomes as monotherapy or combination therapy, in both first-line and later settings. See Wang et al., (2021), “Toward personalized treatment approaches for non-small-cell lung cancer,” Nat Med 27, 1345-1356; and Shields et al., (2021), “Immunotherapy for Advanced Non-Small Cell Lung Cancer: A Decade of Progress,” Am Soc Clin Oncol Educ Book 41, e105-e127.

[0005] In the treatment of cancer using immunotherapy approaches, immune checkpoint inhibitors (ICIs) targeting programmed cell death protein 1 (PD-1) and programmed cell death protein ligand 1 (PD-L1) have achieved remarkable success. ICIs act by reversing T cell inhibition induced by the tumor microenvironment, thereby restoring antitumor immunity.

[0006] However, while ICI can induce effective and durable responses, only some patients show any benefit. Measurement of PD-L1 expression in tumors was among the first to be adopted in the clinic early on as a patient selection biomarker for anti-PD1. However, PD-L1 expression often fails to identify patients who can benefit from ICI treatment. This is demonstrated, for example, by the lack of response seen in PD-L1-expressing patients and survival statistics that do not correlate with PD-L1 expression (see Brahmer et al., (2015), “Nivolumab versus Docetaxel in Advanced Squamous-Cell Non-Small-Cell Lung Cancer,” New Engl J Med 373:123-135; and Rodriguez-Abreu et al., (2021), “Pemetrexed plus platinum with or without pembrolizumab in patients with previously untreated metastatic nonsquamous NSCLC: protocol-specified final analysis,” KEYNOTE-189 Ann Oncol 32:881-895).Additional biomarkers that have recently emerged to improve patient selection for ICI therapy include assessment of tumor-infiltrating lymphocytes (Tumeh et al., (2014), “PD-1 blockade induces responses by inhibiting adaptive immune resistance,” Nature 515:568-571); T-cell-inflamed gene-expression profile (Ayers et al., (2017), “IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade,” J. Clin. Invest. 127:2930-2940; Cristescu et al., (2018), “Pan-tumor genomic biomarkers for PD-1 checkpoint blockade-based immunotherapy,” Science 362:eaar3593) and tumor mutational burden (TMB) (Rizvi et al., (2015), “Mutational landscape determines sensitivity to PD-1 blockade in non-small-cell lung cancer,” Science 348:124-128). The combination of such biomarkers serves as molecular readout information for various components of the cancer-immunity cycle (Chen et al., (2013), “Oncology Meets Immunology: The Cancer-Immunity Cycle,” Immunity 39:1-10), thereby providing mechanistic insights that can actually increase the proportion of patients who respond to ICI. However, many current biomarkers are not only insufficient to accurately identify ICI responders, but also present practical challenges associated with the availability of sufficient tumor biopsy material, tumor heterogeneity, and the technical complexity involved in the standardization of assay analysis and interpretation. Combinations of multiple biomarker assays to improve the accuracy of predicting responders also increase the complexity and cost of the process.Furthermore, it may prove challenging for immunotherapy efficacy monitoring to require invasive tumor biopsies in patients. Thus, non-invasive and robust biomarkers that can improve ICI patient selection and efficacy monitoring have not yet been established.

[0007] Liquid biopsy approaches that utilize plasma-derived cell-free DNA (cfDNA) present important advantages for biomarker discovery and development in the context of ICI efficacy prediction and monitoring. Circulating cell-free DNA is composed of DNA fragments found in the blood that are derived from the genomes of dead cells of different tissues and blood cells, reflecting cellular turnover under normal / healthy conditions as well as altered homeostasis caused by disease (Rostami et al., (2020), “Senescence, Necrosis, and Apoptosis Govern Circulating Cell-free DNA Release Kinetics,” Cell Rep. 31(13):107830); and Barefoot et al., “Detection of Cell Types Contributing to Cancer from Circulating, Cell-Free Methylated DNA” Front. Genet. (27 July 2021)). Unlike conventional tissue biopsies that can sample only locally and accessible tumor sites, plasma-derived cfDNA can capture tumor heterogeneity by circulating tumor DNA derived from not only a single but also multiple tumor sites. Liquid biopsies also provide access to cfDNA derived from the tumor microenvironment and immune cells. Furthermore, non-invasive approaches such as liquid biopsies enable the collection of serial samples over time, allowing for the dynamic monitoring of treatment efficacy and resistance as well as disease recurrence.

[0008] Epigenetic modifications in cfDNA, such as 5-methylcytosine (5mC) and 5-hydroxymethylcytosine (5hmC), have been widely studied in liquid biopsy-based diagnostic approaches. 5hmC is a stable epigenetic mark derived from the oxidation of 5mC by the dioxygenase of the ten-eleven translocation (TET) family. Unlike 5mC, 5hmC generally serves as a mark of a transcriptionally permissive chromatin state and has been shown to be positively correlated with transcriptional activation, particularly for tissue-type and cell-type specific genes (see Cui, X.-L. et al., A human tissue map of 5-hydroxymethylcytosines exhibits tissue specificity through gene and enhancer modulation. Nat Commun 11, 6161 (2020). Szulwach, K.E. et al. Integrating 5-hydroxymethylcytosine into the epigenomic landscape of human embryonic stem cells. Plos Genet 7, (2011)). In the context of the immune system, dynamic changes in the 5hmC profile have been observed to be important for the differentiation and function of immune cells; Nestor et al., (2016), “5-Hydroxymethylcytosine Remodeling Precedes Lineage Specification during Differentiation of Human CD4+ T Cells,” Cell Rep. 16(2):559-570; and Tsiouplis et al., (2021), “TET-Mediated Epigenetic Regulation in Immune Cell Development and Disease,” Frontiers Cell Dev. Biology 8:623948. Thus, the 5hmC signature in plasma-derived cfDNA provides a rich means for biomarker discovery for various applications ranging from early detection to treatment selection and efficacy monitoring. There is still an unmet and urgent need in the art for improved methods to predict the efficacy of immunotherapy in lung cancer patients and to monitor the efficacy of lung cancer patients during immunotherapy. The ideal method uses cell-free DNA analysis in combination with predictive hydroxymethylation biomarkers and is reliable and non-invasive.

Prior Art Documents

Non-Patent Documents

[0009]

Non-Patent Document 1

Non-Patent Document 2

Non-Patent Document 3

Non-Patent Document 4

Non-Patent Document 5

Non-Patent Document 6

Non-Patent Document 7

Summary of the Invention

Means for Solving the Problems

[0010] DNA of tumor and normal cells is released into the bloodstream, and cell-free DNA (cfDNA) samples extracted therefrom can be analyzed for genetic signatures and epigenetic signatures. As suggested above, epigenetic signatures include, by way of example, DNA methylation and DNA hydroxymethylation, and herein, the 5hmC profile (or “hydroxymethylome” or “hydroxymethylation signature”) is of particular interest.

[0011] The present invention is based on the discovery of a set of hydroxymethylation biomarkers that, in combination with one or more other types of biomarkers, features and / or patient-specific characteristics as needed, correlate with whether a lung cancer patient is responsive or likely to be responsive to a particular lung cancer treatment, such as immunotherapy. “Biomarker,” as the term is used herein, refers to a characteristic that can be measured as an indicator of responsiveness to an exposure or intervention, including normal biological processes, disease processes or therapeutic interventions.

[0012] In a first embodiment, the present invention provides a method for monitoring a patient having lung cancer to determine the effectiveness of treatment during lung cancer treatment, the method comprising the following steps:

[0013] (a) For a lung cancer patient, before undergoing lung cancer treatment, (i) obtaining a cell-free DNA (cfDNA) sample from the patient, enriching hydroxymethylated DNA in the sample, amplifying the hydroxymethylated DNA, and sequencing the amplified hydroxymethylated DNA in a manner that identifies 5-hydroxymethylcytosine (5hmC)-containing fragments or sites in the DNA; and (ii) measuring the hydroxymethylation level in the sequenced cfDNA at each of a plurality of hydroxymethylated biomarker loci, wherein each hydroxymethylated biomarker locus is selected as showing an increase or decrease in hydroxymethylation in a manner correlated with the presence of lung cancer, thereby obtaining a baseline hydroxymethylation signature;

[0014] (b) Using the baseline hydroxymethylation signature as a first input parameter to a computer-generated prediction model comprising a trained machine learning model, thereby providing a first probability score;

[0015] (c) Repeating the process of (a) during the treatment of the patient by lung cancer treatment to obtain a monitoring hydroxymethylation signature for the lung cancer patient;

[0016] (d) Using the monitoring hydroxymethylation signature as a second input parameter to the computer-generated prediction model, thereby providing a second probability score;

[0017] (e) A step of comparing the second probability score with the first probability score, and a step of deriving a differential probability score that characterizes the likelihood that the patient is responsive to lung cancer treatment.

[0018] Each biomarker locus in the above method is selected as showing an increase or decrease in hydroxymethylation in a manner correlated with the lung cancer tumor burden.

[0019] The baseline probability score and the probability score obtained during treatment ("monitoring probability score") are typically calculated using logistic regression analysis of the differences in hydroxymethylation levels at each of the hydroxymethylation biomarker loci. As is known to those skilled in the art, alternatives to logistic regression, including conventional statistical methods, are also envisioned.

[0020] The hydroxymethylation biomarker locus is selected as showing differential hydroxymethylation with respect to the lung cancer tumor burden if, for example, it can be established by the Wilcoxon rank sum test with a p-value of less than 0.05 and a fold change "FC" of at least 1.5 between lung cancer patients who do not respond to lung cancer treatment and lung cancer patients who respond to lung cancer treatment.

[0021] In one aspect, step (e) further includes combining the differential probability score with additional features for at least one additional feature type to characterize the likelihood that the patient is responsive to lung cancer treatment. Examples of additional feature types include DNA fragment size distribution, copy number polymorphism (CNV), cfDNA concentration, methylation profile, T cell inflammatory gene expression profile, circulating tumor DNA number, serum CA19-9 level, serum CA125 level, LAG3 expression, IDO-1 expression, T cell number, inflammatory gene signature, myeloid-derived suppressor cell number, lymphocyte number, mismatch repair deficiency, tumor gene mutation amount, presence or absence of germline mutations, patient-specific specific clinical parameters, and any combination of the foregoing. Combining the additional features with the hydroxymethylation information is generally performed using stacked ensemble analysis.

[0022] In some embodiments, the lung cancer is non-small cell lung cancer.

[0023] In other embodiments, the lung cancer is selected from adenocarcinoma, squamous cell carcinoma, small cell lung cancer, adenosquamous carcinoma, carcinoid tumor, bronchial adenocarcinoma, and sarcomatoid carcinoma.

[0024] In some embodiments, the lung cancer treatment is immunotherapy.

[0025] This method may further include obtaining an additional patient hydroxymethylation signature during the course of lung cancer treatment and calculating a probability score of whether the patient continues to respond to treatment by comparing the additional hydroxymethylation signature to a baseline signature, a monitoring signature, or both.

[0026] By "favorably respond to lung cancer treatment", "respond to lung cancer treatment", "favorably respond to immunotherapy", or simply "respond to immunotherapy", a lung cancer patient treated with lung cancer treatment shows a complete response (CR), partial response (PR), or stable disease (SD) for at least a period of 6 months, as defined by the RECIST 1.1 guidelines described in Eisenhauer et al. (2009), "New response evaluation criteria in solid tumors: revised RECIST guidelines (version 1.1)", European J. Cancer 45(2):228 - 247), the disclosure of which is incorporated herein by reference. A lung cancer patient showing progressive disease (PD) is considered a non-responder to immunotherapy as the term is also defined in the RECIST 1.1 guidelines.

[0027] In another embodiment, the present invention provides a method for determining the likelihood that a lung cancer patient will respond to treatment with a selected lung cancer treatment, the method comprising the steps of:

[0028] (a)(i) Obtaining a cell-free DNA (cfDNA) sample from a patient, enriching hydroxymethylated DNA in the sample, amplifying the hydroxymethylated DNA, and sequencing the amplified hydroxymethylated DNA in a manner that identifies 5-hydroxymethylcytosine (5hmC)-containing fragments or sites in the DNA to obtain a hydroxymethylation signature for a lung cancer patient;

[0029] (b) Mapping the sequenced hydroxymethylated DNA to each of a plurality of hydroxymethylated biomarker loci in a reference hydroxymethylation profile comprising a mixture of hydroxymethylation signatures for a population of individuals having at least one common characteristic selected from having lung cancer and responding to lung cancer treatment and having lung cancer and not responding to lung cancer treatment;

[0030] (c) Determining the degree and position differences between the patient hydroxymethylation signature and the reference hydroxymethylation profile at each locus;

[0031] (d) Calculating a probability score representing the likelihood that a lung cancer patient will respond to treatment in lung cancer therapy using the degree and position of the differences.

[0032] Each hydroxymethylation signature in the mixture serving as the reference hydroxymethylation profile includes the hydroxymethylation level at each of a plurality of hydroxymethylated biomarker loci, and as before, each hydroxymethylated biomarker locus is selected as showing an increase or decrease in hydroxymethylation in a manner correlated with the lung cancer tumor burden.

[0033] In some embodiments, a plurality of hydroxymethylated biomarker loci are associated with T cell inflammatory genes. The genes include, in some embodiments, genes selected from CXCR6, TIGIT, CD27, PDCD1LG2, CD274, CD8A, LAG3, NKG7, CCL5, CMKLR1, PSMB10, CXCL9, IDO1, HLA-DQA1, CD276, STAT1, HLA-DRB1, CD276, STAT1, HLA-DRB1, and HLA-E.

[0034] In some embodiments, a plurality of hydroxymethylated biomarker loci in a reference hydroxymethylation profile are selected from the hydroxymethylated biomarker loci in the tables of FIGS. 29-38.

[0035] In some embodiments, a plurality of hydroxymethylated biomarker loci in a reference hydroxymethylation profile are selected from the hydroxymethylated biomarker loci in the tables of FIGS. 36-38.

[0036] Similar to the previous method, the prediction of efficacy for lung cancer treatment can combine additional feature quantities for at least one additional feature type to characterize the likelihood that a patient will respond to lung cancer treatment.

[0037] If it is determined that a lung cancer treatment is potentially useful in treating a patient, or is presumed to be effective in treating a patient who has already received treatment, i.e., the probability score exceeds a predetermined threshold, the patient can initiate or continue the lung cancer treatment. If the analysis indicates that the patient is unlikely to respond to treatment with a particular lung cancer treatment, a decision can be made to pursue a different course of action. This spares the patient unnecessary treatment, potential adverse events resulting from immunotherapy, and the loss of valuable time in the course of the disease.

[0038] As can be inferred from the above, the selected loci that serve as hydroxymethylation biomarkers herein include loci selected for their relevance to treatment efficacy in lung cancer treatment such as immunotherapy. By "relevance," it is meant that a hydroxymethylation biomarker locus, alone or in combination with one or more other hydroxymethylation biomarker loci, tends to show an increase or decrease in hydroxymethylation in a manner that correlates with treatment efficacy in lung cancer patients, for example, with respect to tumor size, stage, invasiveness, malignancy, etc. In general, relevance can be determined by assessing the correlation between a potential biomarker and the likelihood that a lung cancer patient will respond to treatment in lung cancer treatment such as immunotherapy. As briefly suggested above, such correlation generally, but not necessarily always, includes (a) a multiplicative change of at least 1.5 between the hydroxymethylation level at a particular locus in responders and the hydroxymethylation level at the same locus in non-responders, and (b) a p-value of less than 0.05, differential hydroxymethylation between responders and non-responders determined by the Wilcoxon rank sum test (Mann et al., (1947), “On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other,” Annals of Mathematical Statistics 18(1):50-60).

[0039] A reference hydroxymethylation profile is a dataset representing the hydroxymethylation levels of each of a plurality of hydroxymethylation biomarker loci, the dataset being a composite of hydroxymethylation signatures of a plurality of individuals having at least one common characteristic. The reference profile may be a composite of hydroxymethylation signatures of lung cancer patients who responded to treatment in a particular lung cancer treatment, or may be a composite of hydroxymethylation signatures of lung cancer patients who were non-responders to treatment in a particular lung cancer treatment.

[0040] The reference profile may also be a focused reference profile that can enhance the accuracy of the evaluation method. Different types of focused reference hydroxymethylation profiles can be constructed from different population groups, and then an appropriate reference profile can be selected for the evaluation of a specific patient. For example, for assessing female lung cancer patients in their 70s, a narrowed or focused reference profile generated from a group of female lung cancer patients, lung cancer patients aged 70 to 80, or female lung cancer patients aged 70 to 80 who responded well to treatment in the lung cancer treatment of interest is used. It will be understood that additional such focused reference profiles can be constructed according to the attributes of the lung cancer patients undergoing evaluation or monitoring.

[0041] As long as the currently available methods for assessing whether a subject with lung cancer is responding or will respond to treatment in lung cancer treatment such as immunotherapy are mainly based on clinical symptoms and evaluation by X-ray examination or biomarkers that are not sufficiently predictive, the method of the present invention provides an improvement over those methods.

[0042] In a further embodiment, the present invention then provides a method for the following.

[0043] Determining a treatment strategy for a subject with lung cancer,

[0044] Determining whether the immunotherapy used to treat a subject with lung cancer should be discontinued,

[0045] Determining an alternative treatment strategy after discontinuation of the first lung cancer treatment, and

[0046] Reducing the risk that a subject with lung cancer who is unlikely to respond well to immunotherapy receives immunotherapy.

[0047] In another embodiment, the present invention provides a dataset for use in analyzing the efficacy of lung cancer treatment, the dataset comprising a mixture of hydroxymethylation signatures of a plurality of individuals having lung cancer and having at least one common feature selected from being responsive to lung cancer treatment and not being responsive to lung cancer treatment, wherein each hydroxymethylation signature in the mixture comprises hydroxymethylation levels at each of a plurality of hydroxymethylation biomarker loci selected from those in the tables of FIGS. 29-38.

[0048] In some embodiments, the plurality of hydroxymethylation biomarker loci are selected from those in the tables of FIGS. 36-38.

[0049] The present invention also provides a method for determining whether a lung cancer patient is responsive to lung cancer treatment by calculating a 5hmC molecular response score MR from the analysis of 5hmC levels at selected 5hmC biomarker loci, where a positive value generally indicates that the patient is responsive to treatment and a negative value generally indicates that the patient is non-responsive. 5hmC

[0050] Accordingly, in a further embodiment, a method is provided for identifying differential hydroxymethylation sites for use as hydroxymethylation biomarkers in the assessment of whether a lung cancer patient is a responder or non-responder to lung cancer treatment, the method comprising the steps of:

[0051] (a) obtaining cfDNA from each of a plurality of lung cancer patients who are known responders or known non-responders to treatment;

[0052] (b) determining a baseline count T0 in CPM at each of a plurality of candidate hydroxymethylation biomarker loci in the cfDNA obtained from each patient;

[0053] (c) After starting the treatment and confirming the efficacy or non - efficacy of the treatment, determining the subsequent count T in CPM R which is a step of determining T R where T is determined for each of a plurality of candidate hydroxymethylation biomarker loci in cfDNA obtained from each of the patients, the step;

[0054] (d) Selecting candidate hydroxymethylation biomarker loci showing a threshold p - value of less than 0.05 and a difference ζ of at least 1.5 as hydroxymethylation biomarker loci, where ζ=(T R -T0) / T0 is the step.

[0055] In some aspects of the embodiment, the method further includes, at each of the selected hydroxymethylation biomarker loci, calculating the value for log2(T R / T0) and identifying the calculated value as x i at each locus i or y j at each locus j, where x i and y j are positively and negatively correlated with treatment efficacy, respectively.

[0056] In another embodiment, the present invention provides a method for determining whether a lung cancer patient is responsive to lung cancer treatment, the method including the following steps:

[0057] (a) Determining the baseline count T0 at each of the hydroxymethylation biomarker loci selected according to the above - mentioned method for identifying differential hydroxymethylation sites suitable as hydroxymethylation biomarkers in the evaluation of whether a lung cancer patient is responsive or non - responsive to lung cancer treatment in a cfDNA sample obtained from the patient;

[0058] (b) After obtaining the cfDNA sample from the patient, after starting treatment, the subsequent count T at each of the selected hydroxymethylation biomarker loci Q determining step;

[0059] (c) At each of the selected hydroxymethylation biomarker loci log2(T Q / T0) calculating the value for and identifying the calculated value as x at each locus i i or y at each locus j j where x i and y j are positively and negatively correlated with treatment efficacy respectively, step;

[0060] (d) Using the equation MR 5hmC = μ x - μ y calculating the 5hmC molecular genetics efficacy score (MR 5hmC ) for the patient, where μ x is the average of x over the i loci i and μ y is the average of y over the j loci j step;

[0061] (e) When the 5hmC molecular genetics efficacy score is positive, determining that the patient is responsive to treatment.

[0062] The application documents of this patent include at least one drawing created in color. By filing an application and paying the required fees, a copy of this patent with color drawings will be issued by the Patent and Trademark Office.

Brief Description of Drawings

[0063]

Figure 1

[0064]

Figure 2

[0065]

Figure 3

[0066]

Figure 4

[0067]

Figure 5

[0068]

Figure 6

[0069]

Figure 7

[0070]

Figure 8

[0071]

Figure 9

[0072]

Figure 10

[0073]

Figure 11

Figure 12

[0074]

Figure 13

[0075]

Figure 14

[0076]

Figure 15

Figure 16

[0077]

Figure 17

Figure 18

[0078]

Figure 19

Figure 20

[0079]

Figure 21

Figure 22

[0080]

Figure 23

[0081]

Figure 24

[0082]

Figure 25

[0083]

Figure 26

[0084]

Figure 27

[0085]

Figure 28

[0086]

Figure 29-1

Figure 29-2

[0087]

Figure 30-1

Figure 30-2

Figure 30-3

Figure 30-4

Figure 30-5

Figure 30-6

[0088]

Figure 31-1

Figure 31-2

Figure 31-3

Figure 31-4

Figure 31-5

Figure 31-6

Figure 31-7

Figure 31-8

Figure 31-9

Figure 31-10

Figure 31-11

Figure 31-12

[0089]

Figure 32-1

Figure 32-2

Figure 32-3

Figure 32-4

Figure 32-5

Figure 32-6

Figure 32-7

Figure 32-8

[0090]

Figure 33-1

Figure 33-2

Figure 33-3

Figure 33-4

Figure 33-5

Figure 33-6

Figure 33-7

Figure 33-8

[0091]

Figure 34-1

Figure 34-2

Figure 34-3

Figure 34-4

Figure 34-5

Figure 34-6

Figure 34-7

[0092]

Figure 35-1

Figure 35-2

Figure 35-3

Figure 35-4

Figure 35-5

Figure 35-6

Figure 35-7

Figure 35-8

Figure 35-9

Figure 35-10

Figure 35-11

Figure 35-12

Figure 35-13

[0093]

Figure 36-1

Figure 36-2

Figure 36-3

Figure 36-4

Figure 36-5

Figure 36-6

Figure 36-7

Figure 36-8

Figure 36-9

Figure 36-10

Figure 36-11

Figure 36-12

Figure 36-13

[0094]

Figure 37-1

Figure 37-2

Figure 37-3

Figure 37-4

Figure 37-5

Figure 37-6

Figure 37-7

Figure 37-8

Figure 37-9

Figure 37-10

Figure 37-11

Figure 37-12

Figure 37-13

Figure 37-14

[0095]

Figure 38-1

Figure 38-2

Figure 38-3

Figure 38-4

Figure 38-5

Figure 38-6

Figure 38-7

Figure 38-8

Figure 38-9

Figure 38-10

Figure 38-11

Figure 38-12

Figure 38-13

Figure 38-14

Figure 38-15

Figure 38-16

Figure 38-17

Figure 38-18

Figure 38-19

Mode for Carrying Out the Invention

[0096] Detailed Description of the Invention 1. Terms and Overview:

[0097] Unless otherwise defined, all technical and scientific terms used in this specification have the meanings commonly understood by those skilled in the art to which this invention belongs. Specific terms that are particularly important for explaining the present invention are defined below. Other related terms are defined in International Patent Application Publication WO2017 / 176630 (Quake et al., Invention Title "Noninvasive Diagnostics by Sequencing 5-Hydroxymethylated Cell-Free DNA"). The above-mentioned patent gazette and all other patent documents and publications referred to in this specification are hereby incorporated by reference expressly.

[0098] In this specification and the appended claims, the singular forms "a", "an", and "the" include references to the plural unless the context clearly dictates otherwise. Thus, for example, "an adapter" refers to not only a single adapter but also two or more adapters, which may be the same or different, and "a template molecule" refers to not only a single template molecule but also a plurality of template molecules.

[0099] Numeric ranges include the numbers defining the range. Unless otherwise specified, nucleic acids are described from left to right in the 5' to 3' direction, and amino acid sequences are described from left to right in the amino-terminal to carboxy-terminal direction.

[0100] The headings provided herein do not limit the various aspects or embodiments of the invention. Thus, the terms defined immediately hereinafter are more fully defined by reference to the entire specification.

[0101] The term "sample", as used herein, refers to a substance or a mixture of substances, which is typically, but not necessarily, in liquid form and contains one or more analytes of interest.

[0102] The term "biological sample", as used herein, refers to a sample derived from a body fluid, cell, tissue, or organ of a human subject and includes a mixture of biomolecules (including proteins, peptides, lipids, nucleic acids, etc.). Although not necessarily, the sample is generally a blood sample such as a whole blood sample, a serum sample, or a plasma sample.

[0103] As used herein, the term "nucleic acid sample" refers to a biological sample containing nucleic acids. The nucleic acid sample may be a cell-free nucleic acid sample containing nucleosomes, in which case the nucleic acid sample may be referred to herein as a "nucleosome sample". The nucleic acid sample may be composed of cell-free DNA that is substantially free of histones and other proteins in the sample, for example, in the case after cell-free DNA purification. The nucleic acid sample may also include cell-free RNA herein.

[0104] "Sample fraction" refers to a subset of the original biological sample and may be a compositionally equal part of the biological sample, such as when a blood sample is divided into a plurality of equal fractions. Alternatively, the sample fraction may be compositionally different, for example, in the case where a particular component of the biological sample is removed (extraction of cell-free nucleic acids being an example).

[0105] As used herein, the term "cell-free nucleic acid" encompasses both cell-free DNA and cell-free RNA, and this cell-free DNA and cell-free RNA may be present in the cell-free fraction of a biological sample including a body fluid. The body fluid can be blood including whole blood, serum, or plasma. In most instances, the biological sample is a blood sample, and cell-free nucleic acid samples, e.g., cell-free DNA samples, are extracted from those samples by currently commonly used means known to those skilled in the art and / or described in relevant books and literature, and kits for performing cell-free nucleic acid extraction are commercially available (e.g., AllPrep® DNA / RNA Mini Kit and QIAmp DNA Blood Mini Kit (both available from Qiagen), or MagMAX Cell-Free Total Nucleic acid Kit and MagMAX DNA Isolation Kit (available from Thermo Fisher Scientific)). See, e.g., Hui et al., Fong et al. (2009) Clin.Chem. 55(3):587-598 as well.

[0106] The terms "nucleic acid" and "polynucleotide" are used interchangeably herein to describe a polymer of nucleotides, e.g., deoxyribonucleotides or ribonucleotides, of any length, e.g., greater than about 2 bases, greater than about 10 bases, greater than about 100 bases, greater than about 500 bases, greater than 1000 bases, and up to about 10,000 or more bases in length. The nucleic acid may be enzymatically produced, chemically synthesized, or occur naturally.

[0107] The terms "duplex" and "duplexed" are used interchangeably herein to describe two complementary polynucleotides that have formed base pairs, i.e., hybridized to each other. A DNA duplex is referred to herein as "double-stranded DNA" or "dsDNA" and can be an intact molecule or a segment of a molecule. For example, in this specification, the dsDNA called barcoding and adapter ligation is an intact molecule, whereas the dsDNA formed between the nucleic acid terminal sequences of each proximity probe in a proximity extension assay is a dsDNA segment.

[0108] As used herein, the term "strand" refers to one strand of a nucleic acid consisting of nucleotides covalently bonded to each other, e.g., by phosphodiester bonds. In a cell, DNA usually exists in a double-stranded form and itself has two complementary strands referred to herein as the "top" and "bottom" strands. In certain cases, the complementary strands of a chromosomal region may be referred to as the "plus" and "minus" strands, the "positive" and "negative" strands, the "first" and "second" strands, the "coding" and "non-coding" strands, the "Watson" and "Crick" strands, or the "sense" and "antisense" strands. Which strand is designated as the top or bottom strand is arbitrary and does not imply any particular orientation, function, or structure. The nucleotide sequences of the first strands of some exemplary mammalian chromosomal regions (e.g., BACs, assemblies, chromosomes, etc.) are known and may be present, for example, in the NCBI Genbank database.

[0109] "Adapter", as the term is used herein, is a short synthetic oligonucleotide suitable for a particular purpose in biological analysis. An adapter can be single-stranded or double-stranded, but the preferred adapter herein is double-stranded. In one embodiment, the adapter may be a hairpin adapter (i.e., a single molecule that forms base pairs intramolecularly to form a structure with a double-stranded stem and loop, and the 3' and 5' ends of this molecule are each bound to the 5' and 3' ends of a double-stranded DNA molecule). In another embodiment, the adapter may be a Y-shaped adapter. In another embodiment, the adapter may itself be formed from two different oligonucleotide molecules that form base pairs with each other. As will be apparent, the bondable ends of the adapter may be designed to be compatible with the overhangs formed by cleavage with restriction enzymes, or may have blunt ends or 5' T overhangs. The term "adapter" refers to both double-stranded and single-stranded molecules. The adapter may be DNA or RNA, or a mixture of the two. An adapter containing RNA can be cleaved by RNase treatment or alkaline hydrolysis. The adapter may be 15-100 bases, for example 50-70 bases, although adapters outside this range are also contemplated.

[0110] The term "adapter ligation", as used herein, refers to a nucleic acid bound to an adapter. The adapter can bind to the 5' end and / or 3' end of a nucleic acid molecule. As used herein, the term "adapter sequence addition" refers to the act of adding an adapter to the end of a fragment in a sample. This can be done by filling in the ends of the fragment using polymerase, adding a terminal A sequence, and then ligating an adapter containing a T overhang to the fragment having this terminal A sequence. The adapter is usually ligated to a DNA double-strand using ligase, but for RNA, it is covalently or otherwise bound to at least one end of the cDNA double-strand, preferably in the absence of ligase.

[0111] As used herein, the term "adapter-bound sample" refers to a sample that is bound to an adapter. As will be appreciated in view of the above definition, a sample bound to an asymmetric adapter includes a strand having non-complementary sequences at its 5' and 3' ends.

[0112] As used herein, the term "amplification" refers to creating one or more copies or "amplicons" of a template nucleic acid and can be carried out using any nucleic acid amplification technique. Nucleic acid amplification techniques are, for example, techniques such as PCR, NASBA, TMA, and SDA methods.

[0113] The terms "enrich" and "enrichment" refer to partially purifying a template molecule having a particular feature (e.g., a nucleic acid containing 5-hydroxymethylcytosine) from an analyte that does not have such a feature (e.g., a nucleic acid not containing 5-hydroxymethylcytosine). Enrichment typically increases the concentration of the analyte having the said feature by at least 2-fold, at least 5-fold, or at least 10-fold relative to the analyte that does not have the said feature. After enrichment, at least 10%, at least 20%, at least 50%, at least 80%, or at least 90% of the analyte in the sample may have the feature used for enrichment. For example, at least 10%, at least 20%, at least 50%, at least 80%, or at least 90% of the nucleic acid molecules in the enriched composition may contain one or more strands having 5-hydroxymethylcytosine modified to contain a capture tag.

[0114] As used herein, the term "sequencing" refers to a method by which at least 10 consecutive nucleotides of a polynucleotide are identified (e.g., at least 20, at least 50, at least 100, or at least 200, or more consecutive nucleotides are identified).

[0115] The term "next-generation sequencing (NGS)" or "high-throughput sequencing", as used herein, refers to the so-called parallel SBS (sequencing-by-synthesis) or SBL (sequencing-by-ligation) platforms currently employed by, for example, Illumina, Life Technologies, Roche, etc. Next-generation sequencing methods include, but are not limited to, nanopore sequencing methods such as those commercialized by Oxford Nanopore Technologies, electronic detection methods such as Ion Torrent technology commercialized by Life Technologies, and single-molecule fluorescence-based methods such as those commercialized by Pacific Biosciences.

[0116] The term "read", as used herein, refers to the raw or processed output of a sequencing system (e.g., ultra-parallel sequencing, etc.). In some embodiments, the output of the methods described herein is a read. In some embodiments, those reads may require trimming, filtering, and alignment, resulting in raw reads, trimmed reads, and aligned reads.

[0117] A "unique feature identifier" (UFI) sequence refers to a relatively short nucleic acid sequence that is useful for identifying the features of a nucleic acid molecule. Nucleic acid template molecules and their amplicons containing a UFI may be referred to herein as "barcoded" template molecules or amplicons. Examples of types of UFI sequences include, but are not limited to, the following.

[0118] A "source identifier array" (or "source UFI" or "source barcode") identifies the originating biological sample (or other source). That is, each DNA molecule in a single sample is tagged with the same source identifier array, thus enabling sample pooling prior to sequencing. These UFIs may also be characterized as "sample identifier arrays", "sample UFIs" or "sample barcodes".

[0119] "Fragment identifier array" (or "fragment UFI" or "fragment barcode"): In a nucleic acid sample in which the nucleic acid has been fragmented, each fragment in the sample is barcoded with a corresponding fragment identifier array. Multiple sequence reads having non-overlapping fragment identifier arrays indicate that the nucleic acid template molecules have different origins, whereas multiple reads having the same fragment identifier array, or fragment identifier arrays that substantially overlap, may indicate fragments of the same template molecule. The unique feature identified here is the template nucleic acid molecule from which the fragment originated.

[0120] The two strands of a DNA duplex are each independently tagged by a "strand identifier array" (or "strand UFI" or "strand barcode"), thereby enabling determination of the strand from which the read originated (i.e., it can be determined as the W strand or the C strand).

[0121] A "5hmC identifier array" (or "5hmC barcode") identifies DNA fragments originating from 5hmC-containing cell-free DNA template molecules in the sample, i.e., "hydroxymethylated" DNA.

[0122] A "5mC identifier array" (or "5mC barcode") identifies DNA fragments originating from 5mC-containing cell-free DNA template molecules that do not contain 5hmC.

[0123] A "molecular UFI sequence" (or "molecular barcode") is of a length sufficient to provide a UFI sequence and is added to all nucleic acid template molecules in a sample and is random such that all nucleic acid template molecules bind to a specific UFI sequence. As is known in the art, molecular UFI sequences are used to correct for errors in amplification and sequencing devices, to allow the user to track duplicates and remove duplicates from downstream analysis, and to allow for counting of molecules and then determination of analyte concentration. See, for example, Casbon et al., (2011) Nuc. Acids Res. 39(12):1-8. The "unique property" here is the identity of the nucleic acid template molecule.

[0124] In some embodiments, the UFI can have a length in the range of 1 to about 35 nucleotides, such as 3 to 30 nucleotides, 4 to 25 nucleotides or 6 to 20 nucleotides. In certain cases, the UFI may be error-detecting and / or error-correcting, which means that even if errors are present (e.g., if there are synthesis errors in the molecular barcode sequence, if there are reading errors in the molecular barcode sequence, or if the molecular barcode sequence is distorted during any of the various processing steps for determining the molecular barcode sequence), nevertheless, the code can be correctly interpreted. The use of error-correcting sequences is described in the literature (see, for example, U.S. Patent Application Publication No. 2010 / 0323348 (Hamati et al.) and U.S. Patent Application Publication No. 2009 / 0105959 (Braverman et al.), both of which are incorporated herein by reference).

[0125] The oligonucleotide that functions as a UFI sequence herein may be incorporated into the DNA molecule by any effective method, and "incorporated into" herein means "added to" as long as the UFI is provided at the end of the DNA molecule, near the end of the DNA molecule, or within the DNA molecule. is used interchangeably with "attached to" and "appended to". For example, multiple UFIs may be attached to the ends of DNA using a selected ligase, in which case only the last attached UFI is present at the "end" of the molecule. Further, in the proximity extension assays and histone modification methods detailed below, the UFI may be included within the nucleic acid tail of a proximity probe, at the end of the nucleic acid tail of the proximity probe, or within the hybridized region generated when the probe binds to a protein target.

[0126] More generally, the term "detecting" is used interchangeably with the terms "determining", "measuring", "evaluating", "assessing", "assaying", and "analyzing" to refer to any form of measurement and also includes determining whether an element is present or not. These terms include both quantitative and / or qualitative determinations. Assessing may be relative or absolute. Thus, "assessing the presence of ~" includes both determining the amount of the portion present and determining whether the portion is present or absent. Assessing the level at a hydroxymethylation biomarker site refers to determining the degree of hydroxymethylation at such a site.

[0127] "Accuracy" refers to the degree of agreement between a measured or calculated quantity (the value recorded in a test) and its correct (or true) value. Clinical accuracy relates to the proportion of true results (true positive (TP) or true negative (TN)) to incorrectly classified results (false positive (FP) or false negative (FN)), and may be referred to as sensitivity, specificity, positive predictive value (PPV), or negative predictive value (NPV), or likelihood, or other metrics, and may also be referred to as the odds ratio.

[0128] "Performance" is a term related to the overall usefulness and quality of a diagnostic or prognostic test, including, inter alia, clinical and analytical accuracy, other analytical and process characteristics such as usage characteristics (e.g., stability, ease of use), healthcare economic value, and the relative costs of the components of the test. Any of these factors can be a source of excellent performance, and thus the usefulness of the test, and can be measured, as appropriate, by appropriate "performance metrics" such as the AUC, time to result, shelf life, etc.

[0129] "Clinical parameter" includes all non-sample biomarkers of these and other characteristics, not limited to, the health status or lesion size of the subject, lesion location, patient age, patient weight, patient gender, patient ethnicity, family history, genetic mutations, and PD-L1 tumor staining results currently used in clinical practice to determine whether anti-PD-1 therapy is appropriate.

[0130] A "formula", "algorithm", or "model" is any mathematical equation, algorithmic, analytical, or programmed process, or statistical technique that receives one or more continuous or categorized inputs and calculates an output value, which may be referred to as a "probability score" or "index value". Non-limiting examples of "formulas" include sums, ratios, and regression operators (e.g., coefficients or exponents), conversion or normalization of biomarker values (including, but not limited to, normalization schemes based on clinical parameters such as gender, age, or ethnicity), rules and guidelines, statistical classification models, neural networks trained on historical populations, and the like.

[0131] It is particularly useful in combining hydroxymethylation levels at various biomarker loci with clinical parameters and, if necessary, further combining with other factors (e.g., non-hydroxymethylated biomarkers) to determine the relationship between the hydroxymethylation levels at biomarker loci detected in patient samples and the likelihood of a patient responding to immunotherapy, through linear and non-linear equations and statistical classification analysis. In the construction of panels and combinations, of particular importance are, among others, structural and syntactic statistical classification algorithms, as well as risk index construction methods utilizing pattern recognition and machine learning characteristics, established techniques such as correlation, principal components analysis (PCA), factor rotation, logistic regression (LogReg), linear discriminant analysis (LDA), Eigengene linear discriminant analysis (ELDA), support vector machine (SVM), random forest (RF), recursive partitioning tree (RPART), and, among others, other related decision tree classification techniques, shrunken centroids (SC), StepAIC, k-nearest neighbors, boosting, decision trees, neural networks, Bayesian networks, and hidden Markov models. Many such algorithmic techniques are further implemented to perform both feature (site) selection and regularization (e.g., in Ridge regression, Lasso, and elastic net, among others). Other techniques, including the Cox model, Weibull model, Kaplan-Meier model, and Greenwood model well-known to those skilled in the art, may be used for hazard analysis of survival and time to event. Many such techniques are useful, including hydroxymethylation biomarker selection techniques (e.g., variable addition method, variable reduction method, or variable increase and decrease method), a complete investigation of all potential biomarker sets or panels of a given size, combined with genetic algorithms, or such techniques themselves include biomarker selection techniques.These methods may be combined with information criteria, such as the Akaike's Information Criterion (AIC) or the Bayes Information Criterion (BIC), to quantify the trade-off between additional biomarkers and model improvement and to minimize overfitting. The resulting predictive model may be validated in other studies using techniques such as Bootstrap, Leave-One-Out (LOO), and 10-Fold cross-validation (CV), or cross-validated in the study where they were originally learned. At various steps, the false discovery rate (FDR) may be estimated by value substitution with techniques known in the art.

[0132] "Likelihood" in the context of one embodiment of the present invention is the probability that a patient is responsive or non-responsive to treatment in lung cancer treatment. In another embodiment, "likelihood" is the probability that a patient will be responsive or non-responsive to treatment in a particular lung cancer treatment.

[0133] "Hydroxymethylation level" refers to the degree of hydroxymethylation within a hydroxymethylation biomarker locus. The degree of hydroxymethylation is typically measured as the ratio of 5hmC residues to the total cytosine, both modified and unmodified, within a nucleic acid region, i.e., hydroxymethylation density. Other measures of hydroxymethylation density, such as the ratio of 5hmC residues to the total nucleotides in a nucleic acid region, are also possible.

[0134] "Hydroxymethylation profile" or "hydroxymethylation signature" refers to a dataset that includes the hydroxymethylation levels at each of a plurality of hydroxymethylation biomarker loci. The hydroxymethylation profile can be a reference hydroxymethylation profile that includes a mixture of hydroxymethylation profiles for a population of individuals having at least one common characteristic, as described elsewhere herein. The hydroxymethylation profile can also be a patient hydroxymethylation signature constructed from measurements of the hydroxymethylation levels at each of a plurality of hydroxymethylation biomarker sites.

[0135] Accordingly, a "reference hydroxymethylation profile" refers to a dataset representing the hydroxymethylation levels of each of a plurality of hydroxymethylation biomarkers, the dataset being a mixture of hydroxymethylation profiles of a plurality of individuals having at least one common characteristic, e.g., lung cancer patients who respond to treatment with immunotherapy as determined by the RECIST1.1 guidelines, or lung cancer patients who do not respond to treatment with immunotherapy as similarly determined.

[0136] As used herein, "hydroxymethylation biomarker" includes loci selected for their relevance to the likelihood that a lung cancer patient will respond to treatment with immunotherapy.

[0137] As used throughout this application, the term "locus" refers to a site on a nucleic acid molecule, which can be single-stranded or double-stranded, and further, an individual locus (or plural "loci") can be of any length and thus includes a single CpG site and full-length genes, or can span larger structures such as topologically related domains including cases where several such loci come together to form a group such as a related sequence motif, other homologies or functional features (regardless of whether adjacent or topologically related). Loci herein can be contained within a gene body, within annotation features outside the gene body such as a promoter, enhancer, transcription start site, transcription stop site or DNA binding site, or combinations thereof, or within an untranslated region, or "UTR" (including 3'UTR and 5'UTR).

[0138] Some of the individual hydroxymethylation biomarkers disclosed herein may not have significant individual importance in the assessment of a patient's responsiveness to immunotherapy, but when used in combination with one or more other types of biomarkers and, optionally, clinical parameters that affect the assessment and monitoring of lung tumors, as required by the methods of the present invention, for example, it should be noted that they become important in differentiating between subjects responsive to immunotherapy and those not responsive to immunotherapy.

[0139] In this application, any two variables are considered to be "highly correlated" if they have a coefficient of determination (R2) of 0.5 or greater. The present invention encompasses such functional and statistical equivalents to the hydroxymethylation biomarkers disclosed herein.

[0140] As used herein with respect to two variables (e.g., two values, two sets of values, a value or set of values and a disease state, a value or set of values and a risk associated with a disease state, etc.), the term "correlates" indicates that the two variables tend to vary together. "Correlation" is a measure of the extent to which two or more variables vary together. A positive correlation indicates the degree to which the variables increase or decrease in parallel. An example of a positive correlation is the relationship between the level of hydroxymethylation at a hydroxymethylation biomarker locus on the one hand and the efficacy of a lung cancer patient to immunotherapy on the other hand, where the level of hydroxymethylation increases as the efficacy of the subject to the immunotherapy increases. Conversely, a negative correlation would exist if the level of hydroxymethylation at the hydroxymethylation biomarker locus decreases as the efficacy of the subject to immunotherapy decreases.

[0141] The present invention relates in part to the discovery that certain biological markers, particularly epigenetic markers associated with DNA hydroxymethylation, correlate with the likelihood that a subject having lung cancer will respond favorably to a lung cancer treatment such as immunotherapy. These methods include measuring the level of hydroxymethylation at each of a plurality of hydroxymethylation biomarker loci to create a hydroxymethylation signature for a patient, and then comparing the patient's hydroxymethylation signature to a reference hydroxymethylation profile at each locus. The biomarkers are differentially hydroxymethylated with respect to the likelihood of response to treatment with checkpoint blockade therapy for patients having non-small cell lung cancer, which is one of several types of "immunotherapy" discussed elsewhere herein.

[0142] The present invention enables a physician to determine the effectiveness of a lung cancer treatment being administered to a subject having lung cancer, diagnose lung cancer in a patient in whom a lung tumor has not yet been identified, determine whether a lung tumor identified via imaging diagnosis is non-small cell lung cancer, assess the stage of a identified lung tumor, predict whether a healthy individual is at risk of developing lung cancer, particularly non-small cell lung cancer, identify the risk that a identified lung tumor will progress to cancer, and also identify changes in the size, stage, malignancy or degree of invasiveness of a cancerous lung tumor.

[0143] As used herein, the term "lung cancer" refers to any cancer of the lung, including non-small cell lung cancer, such as adenocarcinoma, squamous cell carcinoma and large cell carcinoma; small cell lung cancer; adenosquamous carcinoma; carcinoid tumor; bronchioloalveolar carcinoma; and sarcomatoid carcinoma. Non-small cell lung cancer (NSCLC) is the most common form of lung cancer, and adenocarcinoma is the most common among non-smokers. Lung cancer patients evaluated using the present method may be in the early or late stages of the disease, have tumors showing strong or weak PD-L1 staining, show different spirometry results, and the like.

[0144] As used herein, the term "lung cancer patient" or simply "patient" refers to any living individual diagnosed with lung cancer via imaging diagnosis, biopsy or other known means, and refers to the intended recipient of an immunotherapy treatment discussed in detail herein.

[0145] As used herein, the term "immunotherapy" refers to any method for treating a disease by activating or suppressing the immune system. Examples of immunotherapy include, but are not limited to, cell therapies such as dendritic cell therapy; antibody therapy; and cytokine therapy (e.g., treatment with interferon or interleukin). Most commonly, the immunotherapy treatments discussed herein involve the administration of therapeutic antibodies that bind to and block immune checkpoint receptor proteins such as CTLA-4, PD-1. PD-1 is an important immune checkpoint receptor expressed by activated T cells and B cells and mediates immunosuppression. Two cell surface glycoprotein ligands for PD-1, programmed death ligand-1 (PD-L1) and programmed death ligand-2 (PD-L2), have been identified, and these are expressed on many human cancers in addition to antigen-presenting cells and have been shown to downregulate T cell activation and cytokine secretion upon binding to PD-1. Representative antibodies that target PD-1, PD-1 ligands (e.g., PD-L1), or other immune checkpoint receptors or their ligands and are encompassed by the immunotherapy referred to herein include, but are not limited to, atezolizumab (Tecentriq®, Genentech); necitumumab (Portrazza®, Eli Lilly); nivolumab (Opdivo®, Bristol-myers Squibb); cemiplimab (Libtayo, Regeneron Pharmaceuticals); avelumab (Bavencio, EMD Serono): durvalumab (Imfinzi®, AstraZeneca); dostarlimab (Jemperli®, GlaxoSmithKline); retifanlimab (Zynyz®, Incyte); and pembrolizumab (Keytruda®, Merck). However, it should be understood that the invention is not limited in this regard and that "immunotherapy" as used herein refers to any therapeutic treatment that activates or suppresses the immune system and is potentially useful in the treatment of lung cancer.

[0146] It should be noted that the methodology of the present invention is not necessarily limited to lung cancer immunotherapy and can also be applied to other lung cancer treatments.

[0147] 2. Determination and Use of Baseline Hydroxymethylation Signature:

[0148] In a first embodiment, a method for monitoring a patient having lung cancer during lung cancer treatment is provided to determine the effectiveness of the lung cancer treatment. In a first step of this method, a baseline hydroxymethylation signature is obtained for the patient. The baseline hydroxymethylation signature includes the hydroxymethylation levels at each of a plurality of hydroxymethylation biomarker loci. The selected hydroxymethylation biomarker loci are differentially hydroxymethylated with respect to the likelihood that the patient will respond or not respond to the intended immunotherapy treatment, such that a hydroxymethylation level higher or lower than the hydroxymethylation level in the baseline hydroxymethylation signature obtained during treatment correlates with the likelihood that the patient will respond to treatment in the selected lung cancer treatment.

[0149] To generate a baseline hydroxymethylation signature for a patient, a cfDNA sample is obtained from the patient. Extraction of cfDNA from a blood sample can be performed using any suitable technique, for example, using a commercially available kit referred to in the previous section. Next, the cfDNA is enriched such that its concentration is significantly increased, which is virtually necessary because the levels of cfDNA typically obtained are very low. A generally preferred enrichment technique is described in the international patent application publication WO2017 / 176630 by Quake et al., which is hereby incorporated by reference in its entirety: an affinity tag is added to 5hmC residues in a sample of cfDNA, and then the tagged DNA molecules are selectively removed by binding to a functionalized solid support. An example of a method as described by Quake et al. involves first modifying blunt-ended adapter-ligated double-stranded DNA fragments in a cell-free sample to covalently bond biotin as an affinity tag to 5hmC residues. This can be done by selectively glucosylating 5hmC residues with uridine diphosphate (UDP) glucose functionalized at the 6-position with an azide moiety, followed by a spontaneous 1,3-cycloaddition reaction with alkynyl-functionalized biotin via a "click chemistry" reaction. The DNA fragments containing biotinylated 5hmC residues are then adapter-ligated dsDNA template molecules that can be captured in an enrichment step with a solid support functionalized with a biotin-binding protein (e.g., avidin or streptavidin).

[0150] The captured cfDNA is then amplified without being released from the support, resulting in a plurality of amplicons. Any suitable amplification technique (e.g., PCR, NASBA, TMA, SDA) can be used, but PCR is preferred.

[0151] Next, the patient cfDNA is sequenced in a manner that identifies 5hmC-containing fragments or sites in the cfDNA, as described below.

[0152] Next, the hydroxymethylation levels in the sequenced cfDNA are measured at each of a plurality of hydroxymethylation biomarker loci, each hydroxymethylation biomarker locus being selected to show an increase or decrease in hydroxymethylation in a manner correlated with the presence of lung cancer or response to immunotherapy.

[0153] The resulting baseline hydroxymethylation signature is used as a first input parameter into a computer-generated prediction model that includes a trained machine learning model, thereby providing a first probability score. Next, a second hydroxymethylation signature, also referred to herein as a “monitoring hydroxymethylation signature,” is determined for a lung cancer patient during treatment with a selected lung cancer treatment. This second hydroxymethylation signature is determined in the same manner as the baseline hydroxymethylation signature discussed above. This second hydroxymethylation signature is used as a second input parameter into the computer-generated prediction model to provide a second probability score. Finally, the second probability score is compared to the first probability score to derive a differential probability score that characterizes the likelihood that the patient is responding to the lung cancer treatment.

[0154] If the second probability score is greater than the first probability score such that the differential probability score is positive, it may be determined that the lung cancer treatment is likely ineffective, i.e., the patient is not responding to the treatment. An alternative system may also be configured such that a negative differential probability score indicates non-response to the treatment.

[0155] In an embodiment of determining whether a lung cancer patient who has not yet received treatment is likely to be effectively treated with a particular treatment, the baseline hydroxymethylation signature is compared to a reference hydroxymethylation profile instead of the hydroxymethylation signature obtained during the course of the treatment.

[0156] That is, the identified 5hmC residues are mapped to each of a plurality of loci in the reference hydroxymethylation profile, where each locus serves as a hydroxymethylation biomarker that is differentially hydroxymethylated with respect to the likelihood that a lung cancer patient will respond or not respond to treatment with a particular lung cancer treatment, as described above. Thus, information regarding the hydroxymethylation level is inferred from the sequence reads obtained. That is, the sequence reads are analyzed to provide a quantitative determination of which sequences in the cfDNA are hydroxymethylated and the level of hydroxymethylation. This can be done, for example, by counting the sequence reads or, prior to amplification, by counting the number of original starting molecules based on the fragmentation cleavage points of the original starting molecules and / or whether the original starting molecules contain the same molecular UFI. The use of molecular UFI sequences (sometimes referred to as "molecular barcodes") in combination with other features of the fragments (e.g., the end sequences of the fragments that define the cleavage points) to distinguish between fragments is known. See, inter alia, Casbon (2011) Nucl. Acids Res. 22 e81 and Fu et al., (2011) Proc. Natl. Acad. Sci. USA 108:9026-31). Molecular barcodes are also described in various other publications, in addition to U.S. Patent Application Publication Nos. 2015 / 0044687, 2015 / 0024950, and 2014 / 0227705, and U.S. Patents Nos. 8,835,358 and 7,537,897.

[0157] The molecular UFI sequence is preferably incorporated into an adapter that is ligated to the ends of the cfDNA after extraction of the cfDNA. The adapter can be constructed to include additional UFI sequences, such as a sample UFI sequence, a strand identifier UFI sequence, or both.

[0158] Other methods for identifying hydroxymethylation signatures of DNA in cell-free nucleic acid samples are described in International Patent Application Publication WO2019 / 160994 by Arensdorf et al. for "Methods for the Epigenetic Analysis of DNA, particularly Cell-Free DNA"; U.S. Patent Applications Nos. 16 / 275,237 and 17 / 118,234 by Arensdorf et al. filed concurrently; and Liu et al., (2019) Nature Biotech. 37:424-29, all of which are incorporated herein by reference. These references are also useful in combination with embodiments of the present invention in which the cfDNA methylation profile of a patient is identified in addition to the cfDNA hydroxymethylation profile of the patient.

[0159] Mapping the identified 5hmC residues to each of a plurality of loci in a reference hydroxymethylation profile enables determination of the differences between the patient's hydroxymethylation profile and the reference hydroxymethylation profile with respect to both the degree and location of the differences.

[0160] Selected loci in the methods described herein are loci identified herein as hydroxymethylation biomarkers, i.e., loci that are differentially hydroxymethylated in a manner related to the likelihood that a lung cancer patient will respond to treatment with immunotherapy and / or the likelihood that a lung cancer patient is responding to ongoing treatment with immunotherapy. These biomarkers are shown in FIGS. 29-32 and FIGS. 36-38.

[0161] When genome-wide coverage through shotgun sequencing is not required or desirable (generally for cost reasons), either a targeted detection approach or a non-sequencing detection approach can be used after enrichment to quantify specific hydroxymethylation biomarkers and loci of interest. For example, after enrichment of 5hmC, targeted PCR amplicons that cover only specific regions are generated from the 5hmC-enriched template, utilized as a narrower genome coverage approach, and can be used as input for sequencing or detected directly.

[0162] If there is interest in a smaller number of discrete loci, a combination of these post-enrichment approaches with target amplification can also be an efficient way to reduce the number of sequencing reads (and the cost of sequencing) required for each sample, enabling further sample multiplexing per sequencing run and further reducing the sequencing cost required for each sample). In non-sequencing approaches, quantitative PCR or hybridization assays themselves (e.g., using direct fluorescent nucleotide labeling and microarray or other substrate capture and binding) can be used as quantitative readout information for hydroxymethylation biomarkers, and such approaches are well known in the art and are often scaled up to hundreds or thousands of short amplicons.

[0163] In this method, after amplification, pooling, and sequencing, a 5hmC UFI sequence is added to the ends of the captured adapter-ligated dsDNA template molecules so that information regarding the hydroxymethylation profile can be estimated from the resulting sequence reads. That is, the sequence reads are analyzed to quantitatively determine the sequences that are hydroxymethylated in cfDNA. This may be done by counting the sequence reads or, alternatively, by counting the number of original starting molecules based on the fragmentation cleavage points of the sequences and / or whether the sequences have a similar molecular UFI prior to amplification. The identification of each fragment is known using a molecular UFI sequence (or “molecular barcode” as it is sometimes referred to) along with other features of the fragment (e.g., the end sequences of the fragment that define the cleavage points). See Casbon (2011) Nucl. Acids Res. 22 e81 and Fu et al. (2011) Proc. Natl Acad. Sci. USA 108:9026-31. Molecular barcodes are also described in U.S. Patent Application Publications Nos. 2015 / 0044687, 2015 / 0024950, and 2014 / 0227705, and U.S. Patents Nos. 8,835,358 and 7,537,897, as well as various other publications.

[0164] Other methods for determining the hydroxymethylation profile of DNA in cell-free nuclear samples are described in International Patent Publication WO 2019 / 160994 A1 (Arensdorf et al., invention title “Methods for the Epigenetic analysis of DNA, particularly Cell-Free DNA”), and U.S. Patent Application Publication No. 2017 / 0298422 (Song et al.), which are hereby incorporated by reference. These references are also useful with respect to embodiments of the present invention where the 5-hydroxymethylation determination and analysis further includes the detection of cfDNA methylation profiles in addition to the cfDNA hydroxymethylation profile.

[0165] The methodology of Arensdorf et al. described in International Publication No. 2019 / 160994 can be implemented as follows.

[0166] Dual biotin technology: After the cell-free nucleic acid sample is extracted from the biological sample and the cfDNA is adapter-ligated, the 5hmC residues in the cfDNA are selectively labeled with an affinity tag, e.g., a biotin moiety as described hereinabove. Biotinylation can be carried out, as described above, by glucosylation with uridine diphosphate glucose-6-azide catalyzed by βGT, followed by selective functionalization of the 5hmC residues by a click chemistry reaction to covalently attach an alkyne-functionalized biotin moiety. Next, an avidin or streptavidin surface (e.g., in the form of streptavidin beads) is used to retrieve all of the dsDNA template molecules biotinylated at the 5hmC positions, which are then placed in a separate container for attaching the UFI sequence during amplification. The remaining dsDNA template molecules in the supernatant are fragments that have 5mC residues or no modifications (the latter group includes cDNA generated from cfRNA). Next, a TET protein is used to oxidize the 5mC residues in the supernatant to 5hmC, and in this case, a TET mutant protein is used to ensure that the oxidation of 5mC does not proceed beyond hydroxylation. Suitable TET mutant proteins for this purpose are described in Liu et al., (2017) Nature Chem. Bio. 13:181-191, which is incorporated herein by reference. Then, glucosylation catalyzed by βGT and subsequent biotin functionalization are repeated. The fragments biotinylated at each of the original 5mC positions thus marked are captured with streptavidin beads. The DNA fragments bound to the beads are then barcoded with a UFI sequence during amplification that is different from the 5mC UFI sequence used in the first step. Unmodified DNA fragments, i.e., fragments that do not contain modified cytosine residues, remain in the supernatant at this point. If desired, sequence-specific probes can be used to hybridize to the non-methylated DNA strands. The resulting hybridized complexes can be retrieved during amplification, as described above, and tagged with additional UFI sequences.

[0167] Pyridine borane method: This is an alternative to the dual biotinylation technique and is also a bisulfite-free process. This method relies on the use of pyridine borane or an equivalently effective alternative organic borane to convert 5-carboxylcytosine (5caC) and 5-formylcytosine (5fC) (both of which can be generated from 5mC and 5hmC) to dihydrouracil (DHU). Since DHU residues are read as thymine (T), while 5mC and 5hmC are read as C, the difference between parallel array reads enables the determination of the DHU positions, which then indicates the positions of the 5mC and 5hmC positions.

[0168] In one embodiment, the pyridine borane method enables the identification of 5hmC positions in adapter-ligated target DNA in a cell-free sample. First, the target DNA is oxidized with an oxidizing reagent that converts 5hmC to 5caC or 5fC, and the selected oxidizing reagent has no effect on 5mC. The oxidation can be carried out enzymatically, but in this embodiment, a chemical oxidizing reagent is preferred. Examples of suitable chemical oxidants for use in performing the foregoing conversion include inorganic or organic perruthenate anions in the form of metal perruthenate salts such as potassium perruthenate (KRuO4), tetraalkylammonium perruthenates such as tetrapropylammonium perruthenate (TPAP) and tetrabutylammonium perruthenate (TBAP), and polymer-supported perruthenate (PSP); and inorganic peroxo compounds and compositions such as the combination of peroxotungstate or copper(II) perchlorate / TEMPO (2,2,6,6-tetramethyl-1-piperidinyloxy), but are not limited thereto. The modified DNA containing 5caC or 5fC instead of 5hmC is then treated with an organic borane effective to reduce, deaminate, decarboxylate or deformylate the oxidized 5hmC and give DHU in place of the oxidized 5hmC. As long as the sequence reads can be readily compared to the standard sequence reads obtained for the target DNA, the DHU-containing DNA is amplified and sequenced to provide sequence reads indicative of 5hmC, and the change from C in the standard sequence reads to T in the sequence reads indicative of 5hmC indicates the 5hmC position.

[0169] In another embodiment, the pyridine borane method is used to identify the 5mC positions in adapter-bound target DNA in a cell-free sample. In this case, 5hmC residues in the target DNA are first tagged with an affinity tag that enables removal of 5hmC-containing fragments from the sample. For example, the 5hmC residues can be selectively glucosylated with uridine diphosphate (UDP) glucose functionalized with an azide moiety at the 6-position, and following this step, a spontaneous 1,3-cycloaddition reaction with alkynyl-functionalized biotin via a "click chemistry" reaction follows. The resulting biotinylated DNA target molecules can then be separated from the sample using a solid support functionalized with a biotin-binding protein (e.g., avidin or streptavidin). The remaining DNA in the sample contains 5mC but not 5hmC. In the next step, the 5mC residues in the remaining DNA are enzymatically oxidized to 5caC or 5fC, followed by treatment with pyridine borane to convert the oxidized 5mC residues to DHU. Preferred enzymes useful as oxidants are enzymes of the ten-eleven translocation (TET) family, or "TET catalytically active fragments" as defined in U.S. Patent No. 9,115,386, the disclosure of which is incorporated herein by reference. The preferred TET enzyme in this context is TET2; see Ito et al., (2011) Science 333(6047):1300-1303. Since the change from C to T (resulting from the replacement of 5mC with DHU) indicates the 5mC position, comparison of the standard sequence reads to the target DNA with the obtained sequence reads following amplification and sequencing indicates the positions of the 5mC residues in the sample DNA.

[0170] In a further embodiment, the pyridine borane technique can be implemented to detect the positions of both 5mC and 5hmC residues in a single cell-free DNA sample. The method involves, for a first fraction of a cell-free DNA sample containing adapter-bound target DNA,

[0171] (a) blocking the 5hmC residues with a blocking reagent to obtain blocked 5hmC residues,

[0172] (b) enzymatically oxidizing 5mC residues to provide oxidized 5mC residues selected from 5caC, 5fC, and combinations thereof,

[0173] (c) treating with pyridine borane to convert the oxidized 5mC residues to DHU, thereby providing a first fraction of DNA containing blocked 5hmC residues and DHU at the 5mC position, and

[0174] (d) amplifying and sequencing the first fraction of DNA to provide a first fraction sequence read in which blocked 5hmC residues are read as C and DHU is read as T comprising.

[0175] Glucosylation is effective as a blocking technique, in which case the blocking reagent can be β-glucosyltransferase and the resulting blocking group on the 5hmC residue is glucose.

[0176] For a second fraction of the same sample, the method comprises

[0177] (e) oxidizing 5hmC residues with an oxidizing reagent effective to convert 5hmC residues to oxidized 5hmC residues without modifying 5mC residues, wherein the oxidized 5hmC residues are selected from 5caC, 5fC, and combinations thereof,

[0178] (f) treating with pyridine borane to convert the oxidized 5hmC residues to DHU, thereby providing a second fraction of DNA containing unmodified 5mC residues and DHU at the 5hmC position,

[0179] (g) amplifying and sequencing the second fraction of DNA to provide a second fraction sequence read in which unmodified 5mC residues are read as C and DHU is read as T, and

[0180] (h) Comparing the first allocation sequence read with the second allocation sequence read to identify 5mC and 5hmC positions in the template DNA further comprises.

[0181] See, for example, Liu et al., (2019) Nature Biotech. 37:424-429.

[0182] Organic boranes can be characterized as complexes of borane with nitrogen-containing compounds selected from nitrogen heterocycles and tertiary amines. The nitrogen heterocycle can be monocyclic, bicyclic or polycyclic, but is typically monocyclic in the form of a 5- or 6-membered ring containing a nitrogen heteroatom and, optionally, one or more additional heteroatoms selected from N, O and S. The nitrogen heterocycle can be aromatic or alicyclic. Preferred nitrogen heterocycles herein include 2-pyrroline, 2H-pyrrole, 1H-pyrrole, pyrazolidine, imidazolidine, 2-pyrazoline, 2-imidazoline, pyrazole, imidazole, 1,2,4-triazole, 1,2,4-triazole, pyridazine, pyrimidine, pyrazine, 1,2,4-triazine and 1,3,5-triazine, any of which can be unsubstituted or substituted with one or more non-hydrogen substituents. Typical non-hydrogen substituents are alkyl groups, especially lower alkyl groups such as methyl, ethyl, n-propyl, isopropyl, n-butyl, isobutyl, t-butyl and the like. Exemplary compounds include pyridine borane, 2-methylpyridine borane (also called 2-picoline borane) and 5-ethyl-2-pyridine. Further information regarding these organic boranes and the reactions of these organic boranes to convert oxidized 5mC residues to DHU can be found in the Arensdorf patent application publication cited above.

[0183] Biotin / Native 5mC Enrichment Method: This is an alternative to the dual biotin technology and starts with the biotinylation of 5hmC residues in adapter-ligated DNA fragments, followed by avidin or streptavidin capture. However, here, instead of modifying the methylated DNA remaining in the supernatant, anti-5mC antibodies or MBD proteins are used to capture and enrich native 5mC-containing fragments. This technique is less preferred herein unless it results in the generation of dsDNA template molecules that can be amplified, pooled, and sequenced along with other dsDNA template molecules from the same sample.

[0184] 3. Hydroxymethylation Analysis:

[0185] Next, using the degree and location of the difference between the patient cfDNA hydroxymethylation signature taken before treatment, i.e., the baseline hydroxymethylation signature, and the hydroxymethylation signature or reference hydroxymethylation profile taken during treatment, a probability score is calculated that represents the likelihood that a lung cancer patient will respond to a particular lung cancer treatment or that a lung cancer patient undergoing lung cancer treatment is benefiting from the treatment. This can also be done using the 5hmC molecular genetic response score approach described above herein and discussed in the examples.

[0186] In related embodiments, a method for monitoring a lung cancer patient during lung cancer treatment, the method comprising: (i) obtaining a cell-free DNA (cfDNA) sample from the patient, enriching hydroxymethylated DNA in the sample, amplifying the hydroxymethylated DNA, and sequencing the amplified hydroxymethylated DNA in a manner that identifies 5-hydroxymethylcytosine (5hmC)-containing fragments or sites in the DNA; and (ii) measuring the hydroxymethylation level in the sequenced cfDNA at each of a plurality of hydroxymethylated biomarker loci, wherein each hydroxymethylated biomarker locus exhibits an increase or decrease in hydroxymethylation in a manner that correlates with the likelihood that the patient is responding to lung cancer treatment, thereby, first, obtaining hydroxymethylation monitoring data for the lung cancer patient. The measured hydroxymethylation levels are then input into a computer-generated prediction model that includes a trained machine learning model, and finally, the prediction model is used to generate a probability score, i.e., a score representative of the likelihood that the patient is responding to lung cancer treatment. As described above, in addition to, or as an alternative to, the foregoing method, a 5hmC molecular genetics response score approach may also be used.

[0187] More specifically, in order to calculate a probability score or a 5hmC-based molecular genetic efficacy score, the method of the present invention involves statistical analysis and mathematical modeling used to analyze high-dimensional and multimodal biomedical data, namely data obtained using the present method for comparing hydroxymethylation profiles. These methods include one or more objective algorithms, models, and analysis methods, such as mathematical analysis based on topographic, pattern recognition-based protocols, for example, support vector machine (SVM), linear discriminant analysis (LDA), naive Bayes (NB), K-nearest neighbor (KNN) protocol, and other supervised learning algorithms and models such as decision trees, perceptrons, and regularized discriminant analysis (RDA), and similar models and algorithms well-known in the art (Gallant S I, “Perceptron-based learning algorithms,” Perceptron-based learning algorithms 1990;1(2):179-91).

[0188] Statistical analysis includes determining an average (M), such as a geometric mean, a standard deviation (SD), a fold change (FC), etc. Whether the difference in hydroxymethylation levels is considered significant can be determined by well-known statistical approaches, typically by specifying a threshold for a particular statistical parameter, such as a threshold p-value (e.g., p < 0.05), a threshold S-value (e.g., ±0.4, S ≤ -0.4 or S > 0.4), or other values, for example, when the level of biomarker hydroxymethylation in a hydroxymethylation profile is considered to have significantly increased or decreased, respectively, compared to the hydroxymethylation level at the same hydroxymethylation biomarker locus in a reference hydroxymethylation profile.

[0189] In one aspect, the methods of the invention apply mathematical formulations, algorithms, or models to distinguish normal samples from cancerous samples and to distinguish various subtypes, stages, and other aspects of a disease or disease outcome. In another aspect, these methods are used for prediction, classification, prognosis, and monitoring and design of treatment.

[0190] For comparison of hydroxymethylation levels or other values, the data are compressed. Compression is typically performed by principal component analysis (PCA) or similar techniques for visualizing the structure of high-dimensional data. PCA is used to reduce the dimensionality of the data (e.g., measured expression values) to uncorrelated principal components (PCs) that explain or represent most of the variance in the data, such as about 50, 60, 70, 75, 80, 85, 90, 95, or 99% of the variance. PCA enables visualization of biomarker levels and comparison of hydroxymethylation profiles, such as between normal samples or reference samples and test samples. PCA mapping, e.g., three-component PCA mapping, is used to map the data into three-dimensional space for visualization, such as by assigning the first, second, and third PCs to the x-axis, y-axis, and z-axis, respectively.

[0191] In some embodiments, there is a linear correlation between the hydroxymethylation levels of two or more biomarkers. Pearson's correlation (PC) coefficient can be used to assess the linear relationship (correlation) between pairs of values, such as between the hydroxymethylation levels of biomarkers. This analysis can be used to linearly separate the distribution of expression patterns by calculating the PC coefficient for individual pairs of biomarkers (plotted on the x-axis and y-axis of an individual similarity matrix). Thresholds can be set for various degrees of linear correlation, such as a threshold for a highly linear correlation (R 2 > 0.50, or 0.40). A linear classifier can be applied to the dataset. In one example, the correlation coefficient is 1.0.

[0192] In some embodiments, feature selection (FS) is applied to remove the most redundant features from a dataset, such as a hydroxymethylation biomarker dataset. FS enhances the generalization ability, accelerates the learning process, and improves the interpretability of the model. In one aspect, FS is utilized using a "greedy forward" selection approach to select a subset of the most relevant features for a robust learning model. (Peng H, Long F, Ding C, “Feature selection based on mutual information: criteria of max-dependency, max-relevance, and min-redundancy,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2005;27(8):1226-38). In some embodiments, the SVM algorithm is used for data classification by increasing the margin between n datasets (Cristianini N, Shawe-Taylor J. An Introduction to Support Vector Machines and other kernel-based learning methods. Cambridge: Cambridge University Press, 2000).

[0193] The analytical classification of hydroxymethylation biomarkers in this specification can be performed according to a predictive modeling method that sets a threshold for determining the probability that a sample (e.g., a cfDNA sample obtained from a patient) belongs to a given class (e.g., an increased likelihood that a lung cancer patient will respond to immunotherapy). The probability is preferably at least 50%, or at least 60%, or at least 70%, or at least 80% or higher. Classification can also be performed by determining whether the comparison between the acquired dataset and the reference dataset gives a statistically significant difference. In that case, the sample from which the dataset was obtained is classified as not belonging to the reference dataset class. Conversely, if such a comparison is not statistically significantly different from the reference dataset, the sample from which the dataset was obtained is classified as belonging to the reference dataset class.

[0194] The predictive ability of the model can be evaluated according to its ability to give a quality assessment metric of a particular value or range of values, such as AUROC (area under the ROC curve) or accuracy. The area under the curve scale is useful for comparing the accuracy of identifiers over the entire data range. An identifier with a larger AUC has a higher ability to correctly classify the unknowns between the two groups of interest. In some embodiments, the desired quality threshold is a predictive model that classifies samples with an accuracy of at least about 0.7, at least about 0.75, at least about 0.8, at least about 0.85, at least about 0.9, at least about 0.95, or higher. As an alternative measure, the desired quality threshold can refer to a predictive model that classifies samples with an AUC of at least about 0.7, at least about 0.75, at least about 0.8, at least about 0.85, at least about 0.9, or higher.

[0195] As is known in the art, the relative sensitivity and specificity of a prediction model can be adjusted to prioritize either a selectivity measure or a sensitivity measure, and the two measures have an inverse correlation. The limits of the above model can be adjusted to give a selected sensitivity or specificity level depending on the particular requirements of the test being performed. One or both of sensitivity and specificity can be at least about 0.7, at least about 0.75, at least about 0.8, at least about 0.85, at least about 0.9, at least about 0.95, at least about 0.98, at least about 0.99, or higher.

[0196] Raw data can first be analyzed by measuring the hydroxymethylation level for each biomarker. The data can be manipulated, for example, the raw data can be transformed using a standard curve, and when the average of multiple measurements is determined, the average of the multiple measurements can be used to calculate the mean and standard deviation for each patient. Next, the data is input into a selected prediction model, and the prediction model classifies the samples. The information obtained as a result can typically be communicated to the patient or healthcare provider in the form of a written report.

[0197] In one embodiment, hierarchical clustering is performed in the derivation of the prediction model, where Pearson correlation is used as the clustering measure. One approach is to consider the data set as the "learning sample" in a "supervised learning" task. CART is a standard in medical applications (Singer, Recursive Partitioning in the Health Sciences (Springer, 1999)), which can be modified by converting qualitative features to quantitative features, sorting by the achieved significance level, and then applying a selected regularization method (e.g., elastic net or lasso).

[0198] In some embodiments, the predictive model includes a decision tree that maps observations about an item to a conclusion about its target value (Zhang et al., “Recursive Partitioning in the Health Sciences,” in Statistics for Biology and Health (Springer, 1999)). The leaves of the tree represent classifications, and the branches represent combinations of features that devolve to the individual classifications.

[0199] The predictive model and algorithm may further include a perceptron, which forms a feed-forward neural network and is a linear classifier that maps input variables to binary classifiers (Gallant (1990), “Perceptron-based learning algorithms,” in IEEE Transactions on Neural Networks 1(2):179-191). In this model, the learning rate is a constant that adjusts the learning speed. A lower learning rate improves the classification model but increases the time to process variables (Markey et al., (2002) Comput Biol Med 32(2):99-109).

[0200] As previously described herein, the present invention provides a method for determining the likelihood that an individual having lung cancer is likely to respond to immunotherapy, and a method for determining the likelihood that a lung cancer patient undergoing immunotherapy is benefiting from the treatment. Thus, the present invention encompasses diagnostic, prognostic, and predictive uses of hydroxymethylation profiles, as well as uses in patient monitoring, evaluation of treatment options, and evaluation of treatment efficacy, wherein in each method of use, the generated hydroxymethylation profile is combined, as needed, with additional biomarker information and / or clinical parameters. All of these methods include the generation of a hydroxymethylation profile that includes the measurement of the hydroxymethylation level at each of a plurality of hydroxymethylation biomarker loci, which loci are selected to exhibit differential hydroxymethylation in lung cancer patients who are responders or non-responders to treatment with immunotherapy.

[0201] Among the provided prediction and diagnosis / monitoring methods, there are methods that utilize statistical analysis, biomathematical algorithms, and prediction models to analyze the detected hydroxymethylation information. Some embodiments include methods and systems for analyzing hydroxymethylation information in classification, staging, prognosis, treatment design, evaluation of treatment options, prediction of outcomes (e.g., predicting the occurrence of metastasis), etc.

[0202] Methods are also provided that use the evaluation of hydroxymethylation levels at biomarker loci in treatment efficacy prediction and patient monitoring, including the evaluation of treatment efficacy in patients and the evaluation of patient-specific or individualized treatment strategies. In some embodiments, these methods are used in conjunction with treatment, for example, by creating hydroxymethylation profiles weekly or monthly before, after, or at the time of treatment. Since the hydroxymethylation levels at specific biomarker loci correlate with disease progression, treatment ineffectiveness or effectiveness, and / or disease recurrence or lack of recurrence, it is useful to periodically create hydroxymethylation profiles during long-term monitoring or treatment periods. In some aspects, the information obtained may indicate that different treatment strategies are preferred. Accordingly, treatment methods are provided herein in which biomarker evaluation is performed before treatment and then used to monitor the treatment effect.

[0203] More specifically, at various time points after treatment is initiated or restarted, significant changes in the hydroxymethylation levels at one or more of the biomarker loci can be seen, indicating whether a treatment strategy, such as immunotherapy, is successful or not, whether the patient is likely to benefit from immunotherapy, or whether a change in the treatment approach is recommended. In some embodiments, the treatment strategy is changed after hydroxymethylation analysis by, for example, adding different therapeutic interventions in addition to or instead of a previous approach, increasing or decreasing the aggressiveness or frequency of the approach, or stopping or restarting the treatment regimen.

[0204] 4. Analysis of multiple signatures:

[0205] The method of the present invention may also include consideration of one or more additional feature types in combination with the above-described 5-hydroxymethylation analysis. That is, the probability score representing the likelihood that a patient will respond to lung cancer treatment or is responding to lung cancer treatment takes into account not only the patient's 5-hydroxymethylation level at specific 5hmC biomarker loci but also one or more additional feature types that correlate with the likelihood that the patient will respond to lung cancer treatment, particularly treatment with immunotherapy. The additional features may be additional types of biological markers. That is, in addition to being analyzed for hydroxymethylation levels at various loci, cell-free DNA samples obtained from a patient are analyzed for methylation levels; DNA fragment size and fragment size distribution; cell-free DNA concentration in the patient sample corresponding to cfDNA plasma concentration ([p-cfDNA]); RNA analysis such as T cell inflammatory gene expression profile (GEP) (see Cristescu et al. (2018), cited above); changes in the number of circulating tumor DNA (ctDNA); surrogate markers for tumor neoantigen high-frequency microsatellite instability (MSI-H), mismatch repair deficiency (dMMR), and the amount of tumor gene mutations that can be measured from tissue samples or plasma samples (Thompson et al., (2021), “Serial Monitoring of Circulating Tumor DNA by Next Generation Gene Sequencing as a Biomarker of Response and Survival in Patients with Advanced NSCLC Receiving Pembrolizumab-Based Therapy,” JCO Precision Oncology 5:510-524) and Bindal et al., (2021), “Biomarkers of therapeutic response with immune checkpoint inhibitors,” Ann. Transl Med 9(12):1040).); The expression of immunosuppressive biomarkers LAG3 and IDO-1; the number of regulatory T cells (Treg); the number of myeloid-derived suppressor cells; the inflammatory gene signature; tumor-infiltrating effector cells; the number of lymphocytes; the microbiome composition; germline mutations, etc. can also be analyzed for biomarkers.

[0206] In combination with hydroxymethylation analysis, patient-specific clinical parameters can also be considered. These covariates include factors such as lesion size, lesion malignancy, lesion stage, lesion location, patient age, patient weight, patient body mass index (BMI), patient gender, patient ethnicity, smoking history, and exposure or lack of exposure to known carcinogens.

[0207] In one embodiment, multiple datasets and machine learning techniques are combined to predict the likelihood that a lung cancer patient will respond to treatment in a specific lung cancer treatment or to determine whether a lung cancer patient who has received treatment is responding to the lung cancer treatment used, using an ensemble model, such as a stacked ensemble model. This model uses 5hmC count data in regions of the genome with various annotations such as gene body, promoter, 5’UTR, 3’UTR, enhancer, intron, exon, LINE, SINE, etc. Each annotated region is considered a feature set and incorporated into the stacked ensemble. In addition to the 5hmC features, in this embodiment, additional feature quantities determined from one or more additional feature types are used. For example, the feature quantities can be determined from additional feature types such as cfDNA fragment size and size distribution, copy number polymorphisms, and cell-free DNA plasma concentration from a WGS library constructed for a cell-free DNA sample obtained from a patient.

[0208] In a representative embodiment, a WGS library derived from a patient cfDNA sample is GC-corrected and processed to determine the number of fragments within two or more different size ranges, e.g., two, three, four, five or more different size ranges within the resulting fragment size distribution, within a window of about 1MB, 2MB, 4MB, 5MB, or even 8MB. Examples include two size ranges of 100-150bp and 150-220bp; two size ranges of 100-150bp and 150-300bp; two size ranges of 100-150bp and 150-400bp; two size ranges of 120-155bp and 155-200bp; 50-150bp and 150-400bp; three size ranges of 100-160bp, 160-200bp and 200-220bp; three size ranges of 50-152bp, 153-240bp and 241-1000bp, etc.

[0209] In another representative embodiment, instead of or in addition to a WGS library derived from a patient cfDNA sample, the number of fragments within two or more different fragment size ranges is obtained from only the fragments having 5hmC sites, and the fragments having 5hmC sites can be isolated from the patient cfDNA sample as described in Part 2 of this section. That is, the above description relates to fragment size evaluation in a WGS library, but 5hmC-containing fragments can be evaluated in the same way and used in addition to or instead of WGS fragment size analysis.

[0210] The ratio of the number of large fragments to the number of small fragments can be utilized as a single feature. However, for example, in the case of two size ranges, the number of fragments in the first size range (e.g., 100 - 150 bp) functions as the first feature, and the number of fragments in the second size range (e.g., 150 - 220 bp) functions as the second feature. It is preferable to use the absolute number of fragments in a specific size range as individual features. As another example, in the case of three size ranges, the number of fragments in the first size range (e.g., 100 - 160 bp), the number of fragments in the second size range (e.g., 160 - 200 bp), and the number of fragments in the third size range (e.g., 200 - 220 bp) function as three different features.

[0211] The aforementioned features, namely the number of fragments in each of two or more specific size ranges, can be combined with at least one other feature type in subsequent analysis, in addition to the patient's hydroxymethylation profile. Copy number variation (CNV) is one such additional feature type. This can be readily determined from a GC - corrected WGS library. For example, to assist in the detection of CNV, the number of reads of length 50 - 1000 bp or another selected length can be mapped along the genome into individual windows, such as windows of 100 kb. The concentration of cell - free DNA in the patient sample can serve as an additional feature to be combined with the hydroxymethylation profile as well as at least one of CNV and the number of fragments in bins of different sizes. The concentration of cfDNA in the patient sample can be readily determined by methods described in the relevant literature or methods known to those skilled in the art. See, for example, Chen et al., (2021) Nature Portfolio 11:5040, which is incorporated herein by reference.

[0212] After normalizing for each feature type by the total count, i.e., the number of fragments in each of two or more size ranges, an elastic net regression model is constructed using glmnet. The elastic net mixing ratio α can be optimized, for example, using k-fold (e.g., 5-fold, 10-fold, more than 10-fold) cross-validation for each feature set (e.g., set to 0.01, 0.1, 0.5, etc.). The regularization parameter λ is also optimized at runtime for each feature set using the same k-fold (e.g., 5-fold, 10-fold, or more than 10-fold) cross-validation.

[0213] Models constructed for all feature types, e.g., the number of fragments in each of two or more size ranges, CNV, plasma cfDNA concentration, and 5hmC profiles, are combined together in a stacked ensemble fashion using a final elastic net fit with a predetermined elastic net mixing ratio (e.g., α = 0.01, 0.1, 0.5, etc.). The stacked ensemble combines the models by using the individual prediction scores from each separate model as a feature vector and then fitting to coefficients that weight the scores from each model. By way of non-limiting example, non-zero coefficients from the individual models can generally include 60 - 90% of the hydroxymethylation profile and 10 - 40% of the number of fragments within at least two size ranges. When CNV and cfDNA concentration are included, the relative weighting can, by way of example, be 60 - 90% of the hydroxymethylation profile, 1 - 20% of the number of fragments within at least two size ranges, 1 - 20% of CNV, and 1 - 20% of cfDNA concentration. In one specific example where cfDNA concentration is omitted, the relative weighting can be 75% - 85% of the hydroxymethylation profile, 14% - 24% of CNV, and 1% fragmentation.

[0214] Next, the new samples can be scored as follows. First, the 5hmC and WGS libraries are processed to prepare the feature vectors used by the individual elastic net models that are input into the fully stacked ensemble model. The individual elastic net model prediction scores are calculated from the appropriate (5hmC or WGS) feature vectors. These scores are then passed as input to the fully stacked ensemble model to generate the final probability scores.

Example

[0215] A. Study design and methods:

[0216] (i) Clinical cohort and study design:

[0217] This study was conducted using plasma obtained from subjects with non-small cell lung cancer who received pembrolizumab or nivolumab monotherapy. Written informed consent was obtained for the storage of archival records and the use of blood specimens for retrospective analysis, and 31 patients were enrolled at participating facilities in Germany. A total of 151 blood samples were collected as approved by the institutional review board (IRB) responsible at each facility. Submission of the study protocol, IRB approval, and specimen handling across all facilities were managed by Indivumed GmbH (Hamburg, Germany). To identify changes induced in the 5-hydroxymethylome of plasma-derived cfDNA, a cohort of non-small cell lung cancer (NSCLC) patients who received the anti-PD-1 immunotherapy agents pembrolizumab or nivolumab as monotherapy was assembled (Figure 1). The median age of the study cohort was 71. Female subjects comprised 58.1% of the cohort. The majority of patients were at an advanced stage (96.8%). Adenocarcinoma comprised 77.4% of the study cohort, and squamous cell carcinoma comprised 22.6%. Blood samples were collected at baseline and after the start of treatment at 4- to 6-week intervals. The timelines for both the dosing of anti-PD1 immunotherapy and blood sampling are shown in Figure 1. The complete cohort consisted of a total of 31 patients and 150 blood samples. Response to treatment was measured by radiographic diagnosis.

[0218] To identify the potential of 5hmC-based biomarkers to provide information on immunotherapy response in lung cancer patients, first, patient cfDNA was isolated from plasma and then subjected to a 5hmC enrichment assay in which 5-hydroxymethylated cfDNA fragments were captured by a highly specific and sensitive chemical click reaction, followed by DNA library preparation. Whole-genome libraries were prepared from the same input cfDNA material. Genome regions enriched for 5hmC were determined by peak detection using MACS2 (https: / / github.com / taoliu / MACS). The details of the procedure used are as follows.

[0219] (ii) Plasma collection: Whole blood samples were obtained by routine venipuncture into Streck Cell-Free DNA BCT (registered trademark) (Streck, La Vista, Nebraska) tubes according to the manufacturer's protocol. The tubes were maintained at 15°C to 25°C until plasma isolation. Whole blood was centrifuged at 1600×g for 10 minutes at room temperature, and then the plasma layer was transferred to a new tube for centrifugation at 1600×g for 10 minutes to isolate plasma within 24 hours of venipuncture. The plasma was then divided into equal volumes and stored at -80°C.

[0220] (iii) Cell-free DNA isolation: Cell-free DNA was isolated using the MagMAX (registered trademark) Cell-Free DNA Isolation Kit (Thermo Fisher Scientific, Waltham, Massachusetts) according to the manufacturer's protocol with automated execution on a HAMILTON STAR liquid handling device (HAMILTON Company, Reno, Nevada) using MagMAX magnetic beads. During this procedure, plasma was incubated with proteinase K and 20% SDS at 60°C for 20 minutes and then cooled. Next, cfDNA was bound to magnetic beads and washed with wash buffer and 80% ethanol, which is under the right of Thermo Fisher Scientific. Finally, cfDNA was eluted with 75 μl of elution buffer. All cfDNA eluates were quantified using the PicoGreen (registered trademark) dsDNA quantification assay (Thermo Fisher Scientific) using Molecular Devices' Spectramax (registered trademark) Plate Readers. TapeStation (registered trademark) 4200 capillary electrophoresis (Agilent Technologies, Santa Clara, California) was used to ensure the absence of contaminating high molecular weight DNA resulting from leukocyte lysis.

[0221] (iv) 5-Hydroxymethylcytosine (5hmC)-enrichment assay and 5hmC / WGS library preparation: Using the cell-free "5hmC-Seal" method described in International Patent Application Publication WO2017 / 176630 by Quake et al., Song et al., (2011) 29:68 - 72, and Han et al., (2016) Mol. Cell 63:711 - 19, 5hmC-enriched libraries were prepared, and these disclosures are incorporated herein by reference. Briefly, hMe-Seal is a low-input whole-genome cell-free 5hmC sequencing method based on selective chemical labeling. To capture 5hmC-containing DNA fragments for sequencing, β-glucosyltransferase is used to selectively label 5hmC with a biotin moiety via azide-modified glucose. When performing hMe-Seal in this example, for each assay, cfDNA was normalized to a total input of 10 ng, and after binding to sequencing adapters, 5hmC was selectively labeled with β-GT and affinity-enriched by selectively capturing DNA fragments containing biotin-labeled 5hmC by binding to Dynabeads M270 streptavidin (Thermo Fisher Scientific). Then, to minimize sample loss during purification, PCR was performed directly on the beads. All libraries were quantified using the PicoGreen® dsDNA quantification assay (Thermo Fisher Scientific) by Molecular Devices' SpectraMax Plate Readers and normalized to 1 ng / μl before pooling. The library pool was quantified by the Qubit dsDNA High Sensitivity Assay (Thermo Fisher Scientific) and normalized for sequencing.

[0222] (v) DNA Sequencing and Alignment: Using the NovaSeq® instrument (Illumina, San Diego, CA) with Version 2 reagent chemistry, 75-base pair paired-end sequencing was performed according to the manufacturer's recommendations for DNA sequencing. Data were collected using the NovaSeq System Suite 2.2.04. Raw data processing and demultiplexing were performed using Illumina bcl2fastq software version 2.20.0.422 to generate sample-specific FASTQ output. Sequencing reads were aligned to the hg38 reference genome using BWA-MEM (Li and Durbin (2010), “Fast and accurate long-read alignment with Burrows-Wheeler transform,” Bioinformatics 26:589-595) with default parameters. The quality of the sequencing data was assessed using the Picard Toolkit (Broad Institute).

[0223] (vi) Peak detection: Using BWA-MEM read alignment, regions or peaks of dense read accumulation marking the positions of hydroxymethylated cytosine residues were identified. Prior to identifying peaks, the BAM file containing the positions of the aligned reads was filtered for poorly mapped (MAPQ < 30) and improperly paired reads using SAMtools (Li et al., (2009), “The Sequence Alignment / Map format and SAMtools,” Bioinform Oxf Engl 25:2078-9). MACS2 (https: / / github.com / taoliu / MACS) was used to perform 5hmC peak calling with a p-value cutoff of 1.00e-5. Identified 5hmC peaks that were present in the “blacklist regions” (https: / / sites.google.com / site / anshulkundaje / projects / blacklists) defined elsewhere and those present on the X, Y chromosomes and mitochondrial genomes were also removed using Bedtools (Quinlan et al., (2010), “BEDTools: a flexible suite of utilities for comparing genomic features,” Bioinformatics 26:841-842). Calculation of genome feature-enriched overlapping 5hmC peaks was performed using the software HOMER (http: / / homer.ucsd.edu / homer / ) with default parameters.

[0224] (vii) Differential 5-hydroxymethylation analysis:

[0225] For the purpose of identifying peaks with differential 5hmC signals, first, a bed file with the merged peaks of all identified peaks was created and overlapping peaks were combined. The raw counts for the peaks were normalized by converting them to log2(counts per million). Weakly expressed peaks (CPM > 2 in less than 10 samples) were removed prior to differential analysis. To identify differentially induced 5-hydroxymethylation by immunotherapy treatment, the Wilcoxon signed-rank test was used to compare plasma collected at the baseline time point with plasma collected while the patient was undergoing treatment. To identify differentially hydroxymethylated regions in responders compared to non-responders, the Wilcoxon sum-rank test was applied to compare plasma hydroxymethylation profiles obtained at the baseline time point in responders versus non-responders. For further analysis, peaks with a p-value less than 0.05 and a log2 fold change of at least 1.5 (either up or down) were retained.

[0226] At each 5hmC biomarker locus, log2(T R / T0) starting from the evaluation of, 5hmC-based molecular genetic response (MR 5hmC ) was also calculated, where T0 is the baseline CPM at each locus and T Q is the CPM observed during treatment monitoring at time Q at each locus. The value of log2(T Q / T0) obtained at each locus is designated as either x i (a positive value if response exists or a negative value if non-response exists) or y i (a positive value if non-response exists or a negative value if response exists). That is, x i and y i represent genomic loci showing increased 5hmC that correlate positively or negatively with treatment response, respectively. Then, MR 5hmC as the mean difference: MR 5hmC = μ x - μy is calculated, where μ x and μ y are the averages of all x i , i = 1, ..., n i , and y j , j = 1, ..., n j respectively (in this example, n i = 129 and n j = 154). In this analysis, the selected loci have a p-value of less than 0.05 and a difference ζ of at least 1.5, ζ = (t Q - t0) / t0.

[0227] T Q represents the CPM at each 5hmC biomarker locus at a certain time point Q during treatment, while the term T R is used to refer to a specific time point at which image diagnosis is performed and treatment efficacy is evaluated according to the RECIST guidelines (i.e., responders = CR / PR or non-responders = PD).

[0228] (viii) Prediction modeling:

[0229] To assess the feasibility of detecting patient response to treatment using the above-described 5hmC profiling assay, the characteristics of cfDNA in plasma obtained from cancer patients were elucidated. The intent was to use the samples to provide a set suitable for training a machine learning model to detect signals of progression. An ensemble of binary models (“base learners”) was trained on early versus late stage samples, and the feature vectors for the various binary models were based on different genomic features derived from the inventors' assay. For example, the feature vector used for one of the binary models consisted of the cpm counts of fragments mapped to the gene body. Each base learner binary model was trained using elastic net regularization, a means of performing feature reduction when the number of features exceeds the number of samples. For an explanation of the general elastic net procedure, see Friedman et al., (2010) J. Stat. Software 33(1):1-22. A software implementation of these methods can be found at https: / / cran.r-project.org / web / packages / glmnet / index.html. An initial filtering procedure was performed to remove features with low variance before fitting the base learners with elastic net. Elastic net logistic regression fitting was performed using glmnet40 with the alpha, mixing parameter set to 0.01. The value of lambda, the elastic net regularization parameter, was set using cv.glmnet, which uses cross-validation to select the optimal value of lambda. After fitting individual base learner binomials to each feature type vector, alpha was set to 0.5 and the value of lambda was determined via cv.glmnet as well to train an elastic net binary ensemble model using the scores from the individual base learners.

[0230] B. Results:

[0231] (i) RECIST assessment:

[0232] As described above, the complete cohort consisted of 31 patients, and a total of 150 blood samples were collected. The blood samples were collected as described in A(i). The efficacy of each patient against anti-PD1 immunotherapy was determined by radiological imaging diagnosis. Among the 31 patients, 18 showed radiological efficacy of partial response or complete response (PR / CR) according to the "Response Evaluation Criteria in Solid Tumors" (RECIST 1.1) guideline (Eisenhauer et al., (2009), "New response evaluation criteria in solid tumours: Revised RECIST guideline (version 1.1)" European J. Cancer 45: 228-247), while 13 patients showed progressive disease (Figure 2).

[0233] (ii) 5hmC profile of cfDNA samples in the study cohort:

[0234] After plasma collection, 5hmC enrichment, DNA sequencing and alignment, and 5hmC enrichment assay described in Part A, Sections (ii)-(v), peak detection was performed using the methodology described in Part A, Section (vi). Genomic regions enriched for 5hmC were determined by peak detection using MACS2. The number of peaks found per million unique reads in a given sample ranged from 2,102 to 10,090, with a median of 5,635. See Figure 3, which illustrates the 5hmC landscape for each sample as a bar graph showing the number of observed 5hmC peaks (upper panel) and the number of hydroxymethylated genes / promoters (lower panel). Most of the peaks were located in introns and intergenic regions, at 60% and 30% respectively (Figure 4). Consistent with previous publications (Garcia-Murillas et al., (2015), “Mutation tracking in circulating tumor DNA predicts relapse in early breast cancer.” Sci Transl Med 7:302ra133; Guler et al., (2020) “Detection of early stage pancreatic cancer using 5-hydroxymethylcytosine signatures in circulating cell free DNA.” Nat Commun 11:5270), 5hmC peaks were most statistically enriched in gene regions when determined by the ratio of 5hmC bases in each genomic feature to the total number of bases in the same features across the genome (Figure 5).

[0235] (iii) Differential 5-hydroxymethylation of plasma cfDNA in anti-PD-1 responders and non-responders:

[0236] To determine whether responders had different 5hmC profiles of plasma cfDNA compared to non-responders, plasma-derived 5hmC profiles at the baseline time point (before the start of treatment) were compared. Patients with progressive disease as determined by radiological diagnosis according to RECIST 1.1 criteria were classified as non-responders, while complete responders and partial responders were classified as responders. This differential 5hmC analysis identified 482 genes with p-values less than <0.05 (Figure 6). Among the genes with increased 5hmC in responders compared to non-responders, 5hmC was higher across genes with known functions in the immune response, such as sialic acid-binding immunoglobulin-like lectin 5 (SIGLEC5) (Vuchkovska et al., (2022) “Siglec-5 is an inhibitory immune checkpoint molecule for human T cells,” Immunology) and CXCL1 (Figure 6). On the other hand, non-responders had increased 5hmC across genes related to metastasis / epithelial-mesenchymal transition or resistance to anti-PD-1 therapy, such as FGFR2 (Ranieri, D. et al., (2015) “Expression of the FGFR2 mesenchymal splicing variant in epithelial cells drives epithelial-mesenchymal transition,” Oncotarget 7:5440-5460) and IL6 (Huseni, M.A. et al., (2023) “CD8+T cell-intrinsic IL-6 signaling promotes resistance to anti-PD-L1 immunotherapy,” Cell Reports Medicine 4:100878). (Figure 6).

[0237] Next, gene set enrichment analysis (GSEA) was performed to identify biological processes that could be distinguished by comparing the pre-treatment cfDNA 5hmC profiles in responders to the pre-treatment cfDNA 5hmC profiles in non-responders. This analysis revealed that several gene sets related to inflammatory immune states such as allograft rejection, inflammatory response, and TNFα signaling via NF-κB were upregulated in responders compared to non-responders (Figure 7). The most significantly enriched pathways determined by GSEA that are associated with response to immunotherapy (as evidenced by increased 5hmC levels in responders) were identified as follows (see also Figures 7 and 32):

[0238] HALLMARK_ALLOGRAFT_REJECTION;

[0239] HALLMARK_INFLAMMATORY_RESPONSE;

[0240] HALLMARK_G2M_CHECKPOINT;

[0241] HALLMARK_COMPLEMENT;

[0242] HALLMARK_TNFA_SIGNALING_VIA_NFKB; and

[0243] HALLMARK_E2F_TARGETS

[0244] The following 16 gene sets were found to be enriched in non-responders compared to responders as determined by GSEA (Figures 7 and 32).

[0245] HALLMARK_COAGULATION;

[0246] HALLMARK_XENOBIOTIC_METABOLISM;

[0247] HALLMARK_BILE_ACID_METABOLISM;

[0248] HALLMARK_CHOLESTEROL_HOMEOSTASIS;

[0249] HALLMARK_HEME_METABOLISM;

[0250] HALLMARK_MYOGENESIS;

[0251] HALLMARK_FATTY_ACID_METABOLISM;

[0252] HALLMARK_UV_RESPONSE_UP;

[0253] HALLMARK_ESTROGEN_RESPONSE_EARLY;

[0254] HALLMARK_KRAS_SIGNALING_DN;

[0255] HALLMARK_ESTROGEN_RESPONSE_LATE;

[0256] HALLMARK_OXIDATIVE_PHOSPHORYLATION;

[0257] HALLMARK_ADIPOGENESIS;

[0258] HALLMARK_UNFOLDED_PROTEIN_RESPONSE;

[0259] HALLMARK_PANCREAS_BETA_CELLS; and

[0260] HALLMARK_EPITHELIAL_MESENCHYMAL_TRANSITION

[0261] Notably, among the genes upregulated in non-responders, there were epithelial-mesenchymal transition genes previously associated with resistance to anti-PD-1 treatment. See Mariathasan et al., (2018), “TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells,” Nature 554:544-548. Furthermore, genes downregulated upon KRAS activation were downregulated in responders, suggesting KRAS activation in responders. Interestingly, KRAS mutations have previously been correlated with response and better outcomes during immunotherapy treatment. See Liu et al., (2020), “The superior efficacy of anti-PD-1 / PD-L1 immunotherapy in KRAS-mutant non-small cell lung cancer that correlates with an inflammatory phenotype and increased immunogenicity,” Cancer Lett 470:95-105; and Lauko et al., (2021), “Impact of KRAS mutation status on the efficacy of immunotherapy in lung cancer brain metastases,” Sci Rep-uk 11:18174.

[0262] Previous RNA analyses of tissue samples obtained from patients who received anti-PD-1 treatment revealed a T cell inflammatory gene expression profile (GEP) incorporating the RNA abundance of 18 genes to distinguish responders from non-responders (see Ayers et al., (2017) and Cristescu et al. (2018) cited above). Given that a majority of differentially hydroxymethylated genes are known to be associated with immune response and drug resistance, we next evaluated 5hmC levels across these previously identified genes within the T cell inflammatory GEP. In particular, several genes important in regulating T cells, such as PDCD1 (encoding the PD-1 protein), IDO-1, LAG-3, and CXCL11, were differentially hydroxymethylated (Figure 8). Furthermore, the average 5hmC level of T cell inflammatory GEP genes was significantly higher in responders compared to non-responders (Figure 9). Previous studies mentioned above demonstrated that increased expression of these 18 T cell inflammatory GEP genes in tumor tissue and in the tumor microenvironment can be used to predict immune therapy response, but this study did not address 5hmC levels in plasma cfDNA.

[0263] To determine whether patients with better overall survival could be distinguished, the inventors evaluated 5hmC levels in plasma cfDNA across 18 T cell inflammatory GEP genes. A subset of patients was followed for at least 36 months after treatment initiation (total n = 19). For survival analysis, these patients were divided into two groups according to the mean 5hmC values of these patients across 18 T cell inflammatory GEP genes, and patients above or below the median of the complete cohort were assigned to the high 5hmC (n = 9) or low 5hmC (n = 10) groups, respectively. As expected, based on the results showing higher 5hmC levels across T cell inflammatory GEP genes in responders compared to non-responders (Figure 9), a tendency to separate the two groups was observed with respect to overall survival, but in this subset of patients, with a p-value of 0.11, statistical significance was not reached (Figure 10). Overall, these results indicate that the expression of these tissue-related genes in predicting treatment response can also be captured by increased hydroxymethylation using this plasma cfDNA analysis without the need for tissue samples. Collectively, these findings indicate that 5hmC profiles can be used to identify immune-related signatures in plasma cfDNA that are consistent with immunotherapy response.

[0264] (iv) Differential 5-hydroxymethylation in cfDNA during anti-PD-1 treatment in responsive and non-responsive patients:

[0265] To examine whether there were any changes in the plasma-derived 5hmC profile after anti-PD-1 treatment, at the time point of radiographic response (T R) were used to perform paired differential analysis by comparing the 5hmC fragment counts per gene with the baseline time point (T0) collected before treatment initiation. Patients with progressive disease as determined by radiological imaging according to the RECIST 1.1 criteria were classified as non-responders, while complete responders and partial responders were classified as responders. This differential analysis identified 530 genes in responders with a p-value < 0.05 (Figure 11). Among the genes with increased 5hmC counts were genes related to tumor immunity or immune cell activation such as HLA-DQB1 and CD69, while genes with decreased 5hmC counts were enriched for cancer or drug resistance genes such as IGF1, TWIST1, and MMP16. In non-responding patients, a similar comparison of treatment-time samples with paired pre-treatment (T0) samples from the time point of radiological response (TR) identified 715 genes with differential 5hmC (Figure 12). In contrast to responding patients, genes with increased 5hmC included genes related to resistance to checkpoint inhibition based on previous studies of tissue samples such as FGF signaling (Adachi et al., (2021) “Inhibition of FGFR Reactivates IFNγ Signaling in Tumor Cells to Enhance the Combined Antitumor Activity of Lenvatinib with Anti-PD-1 Antibodies,” Cancer Res 82:292-306), while genes with decreased 5hmC included CD74, which has a known function in the immune response (Su et al., (2017) “The biological function and significance of CD74 in immune diseases,” Inflamm.Res. 66, 209-216).

[0266] Among the genes differentially hydroxymethylated after treatment, only 25 genes were shared between responders and non-responders (Figure 13). Most of these genes (17 out of 25) showed opposite trends between responders and non-responders (Figure 14). Genes with increased 5hmC in non-responders and decreased 5hmC in responders included multiple cancer-related genes such as MMP16 and TWIST1. Collectively, the minimal overlap between differentially hydroxymethylated genes in non-responder and responder cohorts suggests that these changes detected in plasma cfDNA after treatment are specific to anti-PD-1 responsiveness / resistance.

[0267] To determine whether the 5hmC profile in plasma-derived cfDNA can provide information regarding biological efficacy against anti-PD-1 treatment, the inventors performed gene set enrichment analysis (GSEA) using 5hmC counts across the gene body. In responders, the top enriched pathways were highly immune-related, such as interferon gamma (IFNg) response, inflammatory response, interferon alpha (IFNa) response, and TNFa signaling via NFkB (Figure 15). The IFNg response, the most significantly enriched pathway in 5hmC analysis of plasma cfDNA, has been previously associated with anti-PD-1 response by gene expression analysis of tumor tissue samples collected at baseline from patients treated continuously with pembrolizumab (see Ayers et al., (2017) and Cristescu et al., (2018), previously cited herein). For non-responders, fewer gene sets changed significantly upon treatment. Among the top enriched gene sets identified in non-responders was the epithelial-mesenchymal transition (EMT) gene set (Figure 16), which is known to be associated with resistance to several drugs, including ICIs. See Mariathasan et al., (2018), “TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells,” Nature 554:544-548. Hugo, W. et al., (2016) “Genomic and Transcriptomic Features of Response to Anti-PD-1 Therapy in Metastatic Melanoma.” Cell 165, 35-44. These findings indicate that the inflammatory T cell and EMT signatures found in the tumor microenvironment of responders and non-responders, respectively, can be identified in plasma cfDNA by 5hmC analysis.

[0268] (v) Differentially hydroxymethylated loci identified in cfDNA when discriminating normal lung tissue from lung cancer tissue:

[0269] The differentially hydroxymethylated genes in cfDNA identified in this study were previously associated with ICI response and resistance in tissue transcriptome analysis. Therefore, the inventors sought to ask whether the changes observed in the cfDNA hydroxymethylome were at least partially derived from the hydroxymethylome of tumor-derived DNA. For this purpose, the inventors first identified differentially hydroxymethylated regions (DhMRs) during anti-PD-1 treatment by comparing the 5hmC profiles at T Q in responders and non-responders to T0. In responders (Figure 17) and non-responders (Figure 18) respectively, 152 and 301 DhMRs with p-values less than 0.05 and fold changes greater than 1.5 were identified by the Wilcoxon rank sum test. These DhMRs identified separately in responders and non-responders were specific and did not overlap (Figures 21 and 22). To examine whether these DhMRs identified in plasma provided useful information regarding changes in the 5hmC profile in tumor tissue, 5hmC profiles were generated for 15 normal lung tissues and 18 lung cancer tissues. The top responder (Figure 17) and non-responder (Figure 18) DhMRs identified in plasma were both able to distinguish between tumor lung tissue and normal lung tissue as visualized by the t-distributed stochastic neighbor embedding method (t-SNE) (Figures 19 and 20). These results suggest that the DhMRs identified in plasma are consistent with changes in the 5hmC profile of lung tumor tissue.

[0270] (vi) Monitoring of anti-PD-1 treatment efficacy using plasma-derived 5hmC profiles:

[0271] The observation that 5hmC changes induced by anti-PD-1 treatment did not significantly overlap between responders and non-responders indicates that such regions can be used to monitor treatment efficacy. To test this idea, the treatment-induced fold change in 5hmC occupancy normalized to baseline ((T RCalculate (T - T0) / T0 (i.e., ζ, as defined previously), and then identify the region with the most statistically significant change by applying a threshold p-value of 0.05 and a difference ζ of at least 1.5. First, determine the genomic loci with the most significant change between the responder cohort and the non-responder cohort. This led to the identification of 154 regions with decreased 5hmC and 129 regions with increased 5hmC in responders compared to non-responders.

[0272] Most of these loci were mapped to gene features, as visualized using the genomic browser IGV for two representative 5hmC peak loci (Figure 24). Examination at all study time points showed that changes in 5hmC levels observed at the time of radiological response could be detected as early as 4 - 6 weeks after treatment initiation (Figure 25). Based on this observation, as identified in Figure 23, the molecular genetic response induced by anti-PD-1 treatment on plasma hydroxymethylome was evaluated by measuring the change in 5hmC count across the top differential regions whose hydroxymethylation levels were negatively or positively correlated with response. The differential molecular genetic response based on 5hmC levels was able to distinguish responders from non-responders at the first time point, as early as 4 - 6 weeks after treatment initiation (Figure 26). These results indicate that the 5hmC profile obtained from plasma contains treatment-induced changes that can serve as molecular genetic response markers with the potential for earlier detection than radiological response. Collectively, our findings demonstrate that the 5hmC-based changes identified in plasma samples from NSCLC patients can be utilized for anti-PD-1 treatment response monitoring.

[0273] To test whether 5hmC profiles can be used as a surrogate for tumor burden to assess treatment efficacy, we constructed a predictive model using lung cancer samples to classify plasma samples from patients with early disease and lower tumor burden and patients with advanced progressive disease with higher tumor burden. The training set used to construct this predictive model was unique and did not overlap with any of the patients in the anti-PD-1 treatment cohort. Applying this model to score plasma samples from patients who received anti-PD-1 treatment but did not respond (non-responders), the predicted score for the plasma sample at the time of treatment increased compared to the predicted score for the baseline plasma sample taken from the same patient (Figure 27, left panel). On the other hand, the predicted score for the plasma sample at the time of treatment decreased compared to baseline for most of the patients who belonged to the responder group (Figure 27, right panel). Comparison of the predicted scores at the time of radiographic response with baseline revealed that for 73% of the patients, the 5hmC profile could predict changes in tumor burden in agreement with RECIST (Figure 28). Collectively, our results suggest that 5hmC-based biomarkers can be observed in plasma samples of NSCLC patients for anti-PD-1 treatment efficacy monitoring.

[0274] Immune checkpoint inhibitors can exert significant and durable anti-tumor effects, albeit in a subset of patients. In the experiment above, we investigated the potential of using 5hmC-based changes detectable in plasma-derived cfDNA to predict and monitor anti-PD-1 treatment response in NSCLC patients. These results indicate that cfDNA 5hmC profiles contain loci that correlate with anti-PD-1 response that can be utilized for biomarker discovery for patient selection and treatment efficacy monitoring for immune checkpoint inhibitors in the treatment of lung cancer.

[0275] Among the patient selection biomarkers currently used in diagnosis and treatment, there is PD-L1 immunohistochemistry using tumor biopsy samples. However, it has been reported that patients with PD-L1 immunostaining below the established threshold can still benefit from anti-PD-1 treatment. Furthermore, a lack of efficacy has also been observed in PD-L1 positive patients (Wang et al., (2021), “Toward personalized treatment approaches for non-small-cell lung cancer,” Nat Med 27:1345-1356, Gibney (2016), “Predictive biomarkers for checkpoint inhibitor-based immunotherapy,” Lancet Oncol 17:e542-e551). Tumor sampling at a single tumor site may not accurately reflect the overall PD-L1 expression status in a patient's tumor(s), so both intra-tumor and inter-tumor heterogeneity with respect to PD-L1 expression can lead to such conflicting results. A second important variable is the poor uniformity in PD-L1 immunohistochemistry antibodies and the different thresholds for PD-L1 positivity used (Gibney (2016), “Predictive biomarkers for checkpoint inhibitor-based immunotherapy,” Lancet Oncol 17:e542-e551). These limitations can be circumvented by the methods provided herein that capture DNA released not only by cancer cells from different tumor sites but also by the tumor microenvironment and immune cells. In particular, the analysis of plasma cfDNA 5hmC, which depicts active biological pathways, can provide a better understanding of drug response in tumors and immune cells using a single analyte. Indeed, the inventors have found herein that plasma cfDNA 5hmC profiles are significantly different in responders compared to non-responders both pre-treatment (Figures 11-16) and post-treatment (Figures 6-9). This analysis revealed 5hmC enrichment across immune-related genes and indicated increased immune gene activation in responders.Considering that 60% of both cohorts were PD-L1 positive (Figure 2), these results indicate that 5hmC profiling can provide an independent measure of immune status that gives useful information about potential responsiveness to immunotherapy.

[0276] The plasma-derived cfDNA analysis provided herein presents unparalleled advantages as a source of candidate biomarkers for ICI responsiveness and treatment monitoring. Non-invasive biomarkers avoid the issues associated with tumor biopsies such as low rates of patient compliance, the potential for complications, and insufficient material. Furthermore, non-invasive biomarkers can facilitate serial sampling, which enables the monitoring of dynamic changes during treatment that may be associated with immune activation or acquired resistance. In this study, anti-PD-1 treatment responsiveness induced distinct changes in the plasma cfDNA 5hmC profiles of responders compared to non-responders, which were observed as early as 4-6 weeks after treatment initiation (Figures 25-26). Our results indicate that 5hmC analysis provides a novel non-invasive method for the serial monitoring of anti-PD-1 treatment responsiveness in plasma.

[0277] Plasma-derived cfDNA is a pool of DNA that originates from multiple tissue sources. Immune cells constitute the majority of cfDNA, particularly in healthy subjects, but in cancer patients, tumor cells and other components of the tumor microenvironment also shed DNA into the blood. Methods that are limited to the analysis of tumor-derived cfDNA, such as mutation analysis utilizing tumor information, not only need to address sensitivity issues arising from the dilution of tumor-derived DNA in cfDNA samples, but also cannot provide information about the tumor microenvironment. Tissue-specific marks in cfDNA, particularly 5hmC as described herein, are useful for addressing the issues associated with cfDNA heterogeneity. Consistent with this, our 5hmC analysis of cfDNA identified IFN responses, inflammatory responses in responders, and epithelial-mesenchymal transition in non-responders, indicating both immune cell-derived biology and tumor cell-derived biology.

Claims

1. A computer implementation method for monitoring a patient with lung cancer in order to determine the effectiveness of the treatment during lung cancer treatment, wherein the method is: (a) A step of using a baseline hydroxymethylation signature for a lung cancer patient before treatment as a first input parameter to a computer-generated prediction model including a trained machine learning model, thereby providing a first probability score, wherein the baseline hydroxymethylation signature is obtained by (i) enriching hydroxymethylated DNA in a cell-free DNA (cfDNA) sample obtained from the patient, amplifying the hydroxymethylated DNA, and sequencing the amplified hydroxymethylated DNA in a manner that identifies 5-hydroxymethylcytosine (5hmC)-containing fragments or sites in the DNA; and (ii) measuring the level of hydroxymethylation in the sequenced cfDNA at each of a plurality of hydroxymethylation biomarker loci, wherein each hydroxymethylation biomarker locus is selected to show an increase or decrease in hydroxymethylation in a manner that correlates with the presence of lung cancer, (b) Providing a second probability score by using a monitoring hydroxymethylation signature for the lung cancer patient as a second input parameter to the computer-generated prediction model, wherein the monitoring hydroxymethylation signature is obtained by repeating the process of (a) during the treatment of the patient with lung cancer, (c) A step of comparing the second probability score with the first probability score to derive a differential probability score. Includes, Herein, the differential probability score indicates the likelihood that the patient is responding to the lung cancer treatment, a computer implementation method.

2. The computer implementation method according to claim 1, wherein it is determined that the lung cancer treatment is ineffective if the second probability score is greater than the first probability score.

3. The computer implementation method according to claim 1, wherein each hydroxymethylation biomarker gene locus is selected to show an increase or decrease in hydroxymethylation in a manner that correlates with the response to immunotherapy.

4. The computer implementation method according to claim 1, wherein each hydroxymethylation biomarker gene locus is selected to show an increase or decrease in hydroxymethylation in a manner that correlates with lung cancer tumor volume.

5. The computer implementation method according to claim 1, wherein the baseline probability score and the monitoring probability score are calculated using logistic regression analysis of the difference in hydroxymethylation levels at each of the hydroxymethylated biomarker loci.

6. The computer implementation method according to claim 1, wherein each hydroxymethylation biomarker gene locus is selected to show differential hydroxymethylation with a p-value of less than 0.05 and a change of at least 1.5 between lung cancer patients who do not respond to the lung cancer treatment and lung cancer patients who do respond to the lung cancer treatment, as determined by the Wilcoxon rank-sum test.

7. (c) further comprising combining the differential probability score with an additional feature for at least one additional feature type to characterize the likelihood that the patient is responding to the lung cancer treatment, the computer implementation method according to claim 1.

8. The computer implementation method according to claim 7, wherein the additional feature types include DNA fragment size distribution, copy number variation, cfDNA concentration, methylation profile, T cell inflammatory gene expression profile, circulating tumor DNA count, serum CA19-9 level, serum CA125 level, LAG3 expression, IDO-1 expression, T cell count, inflammatory gene signature, myeloid-derived suppressor cell count, lymphocyte count, mismatch repair deficiency, tumor gene mutation load, presence or absence of germline mutations, patient-specific clinical parameters, and any combination thereof.

9. The aforementioned additional feature type is, The number of cfDNA fragments in each of at least two non-overlapping size ranges, Copy number variations in the aforementioned cfDNA sample, The concentration of cfDNA in the aforementioned cfDNA sample, Patient-specific clinical parameters, and Any of the above combinations The computer implementation method according to claim 7, including the method described in claim 7.

10. The computer implementation method according to claim 9, wherein the patient-specific clinical parameters are selected from lesion size, lesion malignancy, lesion stage, lesion location, patient age, patient weight, patient sex, patient ethnicity, smoking status, and exposure to or lack of exposure to known carcinogens.

11. The computer implementation method according to any one of claims 7 to 10, wherein the combination includes ensemble analysis.

12. The computer implementation method according to claim 11, wherein the ensemble analysis is a stacked ensemble analysis.

13. The computer implementation method according to claim 2, further comprising indicating that the lung cancer treatment should be discontinued after it has been determined that the lung cancer treatment is ineffective.

14. The computer implementation method according to claim 13, further indicating that the lung cancer treatment should be changed to a different lung cancer treatment.

15. The computer implementation method according to claim 14, wherein the different lung cancer treatments include administering higher doses of drugs, administering different drugs, or changing the treatment method.

16. The computer implementation method according to claim 15, wherein the different lung cancer treatments are determined using the baseline hydroxymethylation profile, the monitoring hydroxymethylation profile, or both the baseline hydroxymethylation profile and the monitoring hydroxymethylation profile.

17. A computer implementation method according to any one of claims 1 to 10 and 13 to 16, wherein the response to immunotherapy includes exhibiting a partial response, complete response, or stable disease as defined in RECIST Guideline 1.

1.

18. The computer implementation method according to any one of claims 1 to 10 and 13 to 16, wherein the lung cancer is non-small cell lung cancer.

19. The computer implementation method according to any one of claims 1 to 10 and 13 to 16, wherein the lung cancer is selected from adenocarcinoma, squamous cell carcinoma, small cell lung cancer, adenosquamous carcinoma, carcinoid tumor, bronchial adenocarcinoma, and sarcomatoid carcinoma.

20. The computer implementation method according to any one of claims 1 to 10 and 13 to 16, wherein the lung cancer treatment is immunotherapy.

21. The computer implementation method according to any one of claims 1 to 10 and 13 to 16, wherein the plurality of hydroxymethylated biomarker gene loci are selected from those in the table in Figures 29 to 38.

22. The computer implementation method according to any one of claims 1 to 10 and 13 to 16, wherein the plurality of hydroxymethylated biomarker gene loci are selected from those in the table in Figures 36 to 38.

23. A method for calculating a probability score as an indicator of the likelihood that a lung cancer patient will respond to a selected lung cancer treatment, wherein the method is: (a) A step of obtaining a hydroxymethylation signature for a lung cancer patient by enriching the hydroxymethylated DNA in a cell-free DNA (cfDNA) sample obtained from the patient, amplifying the hydroxymethylated DNA, and sequencing the amplified hydroxymethylated DNA in a manner that identifies 5-hydroxymethylcytosine (5hmC)-containing fragments or sites in the DNA, (b) The steps of mapping the sequenced hydroxymethylated DNA to each of a plurality of hydroxymethylated biomarker gene loci in a reference hydroxymethylation profile which includes a mixture of hydroxymethylation signatures for a population of individuals having lung cancer and responding to lung cancer treatment and having lung cancer and not responding to the lung cancer treatment, (c) The step of determining the difference in degree and location between the patient hydroxymethylation signature and the reference hydroxymethylation profile at each gene locus, (d) A step of calculating the probability score using the degree and location of the difference; Includes, Herein, the probability score is a method that indicates the likelihood that the lung cancer patient will respond to treatment for lung cancer.

24. The method according to claim 23, wherein each hydroxymethylation signature in the composite comprises a hydroxymethylation level at each of a plurality of hydroxymethylation biomarker gene loci.

25. The method according to claim 24, wherein the plurality of hydroxymethylation biomarker gene loci in the reference hydroxymethylation profile are selected from those in the table in Figures 29 to 38.

26. The method according to claim 23, wherein each hydroxymethylation biomarker gene locus is selected as one that exhibits increased or decreased hydroxymethylation in a manner that correlates with the response to immunotherapy.

27. The method according to claim 23, wherein each hydroxymethylation biomarker gene locus is selected to show an increase or decrease in hydroxymethylation in a manner that correlates with lung cancer tumor volume.

28. The method according to claim 23, wherein the probability score is calculated using logistic regression analysis of the difference in hydroxymethylation levels between the patient hydroxymethylation signature and the reference hydroxymethylation profile at each hydroxymethylation biomarker locus.

29. The method according to claim 23, wherein each hydroxymethylation biomarker gene locus is selected to exhibit differential hydroxymethylation with a p-value of less than 0.05 and a change of at least 1.5 between lung cancer patients who do not respond to the lung cancer treatment and lung cancer patients who do respond to the lung cancer treatment, as determined by the Wilcoxon rank-sum test.

30. The method according to claim 23, further comprising (d) combining the probability score with an additional feature for at least one additional feature type to characterize the likelihood that the patient will respond to the lung cancer treatment.

31. The method according to claim 30, wherein the additional feature types include DNA fragment size distribution, copy number variation, cfDNA concentration, methylation profile, T cell inflammatory gene expression profile, circulating tumor DNA count, serum CA19-9 level, serum CA125 level, LAG3 expression, IDO-1 expression, T cell count, inflammatory gene signature, myeloid-derived suppressor cell count, lymphocyte count, mismatch repair deficiency, tumor gene mutation load, presence or absence of germline mutations, patient-specific clinical parameters, and any combination thereof.

32. The aforementioned additional feature type is, The number of cfDNA fragments in each of at least two non-overlapping size ranges, Copy number variations in the aforementioned cfDNA sample, The concentration of cfDNA in the aforementioned cfDNA sample, Patient-specific clinical parameters, and Any of the above combinations The method according to claim 30, including the method described in claim 30.

33. The method according to claim 32, wherein the patient-specific clinical parameters are selected from lesion size, lesion malignancy, lesion stage, lesion location, patient age, patient weight, patient sex, patient ethnicity, smoking status, and exposure to or lack of exposure to known carcinogens.

34. The method according to any one of claims 30 to 33, wherein the combination includes ensemble analysis.

35. The method according to claim 34, wherein the combination includes stacked ensemble analysis.

36. The method according to any one of claims 23 to 33, wherein the response to immunotherapy includes exhibiting a partial response, complete response, or stable disease as defined in RECIST Guideline 1.

1.

37. The method according to any one of claims 23 to 33, wherein the lung cancer is non-small cell lung cancer.

38. The method according to any one of claims 23 to 33, wherein the lung cancer is selected from adenocarcinoma, squamous cell carcinoma, small cell lung cancer, adenosquamous cell carcinoma, carcinoid tumor, bronchial adenocarcinoma, and sarcomatoid carcinoma.

39. The method according to any one of claims 23 to 33, wherein the lung cancer treatment is immunotherapy.

40. A dataset for use in lung cancer treatment response analysis, wherein the dataset comprises a mixture of hydroxymethylation signatures of multiple individuals having at least one common characteristic selected from having lung cancer and responding to lung cancer treatment and having lung cancer and not responding to the lung cancer treatment, wherein each hydroxymethylation signature in the mixture comprises the hydroxymethylation level at each of multiple hydroxymethylation biomarker gene loci selected from the tables in Figures 29 to 35.

41. The dataset according to claim 40, wherein the plurality of hydroxymethylated biomarker gene loci are selected from those shown in the table in Figures 33 to 35.

42. A method for identifying a differential hydroxymethylation site for use as a hydroxymethylation biomarker, as an indicator for evaluating the response to treatment in lung cancer patients, wherein the method is: (a) In each of several candidate hydroxymethylated biomarker gene loci in cfDNA obtained from each of several lung cancer patients who are known to have responded to or not responded to the said treatment, baseline count T in CPM ０ The steps to determine, (b) After initiating the treatment and confirming the response or non-response to the treatment, the subsequent count T in CPM Ｒ This is the step of determining T Ｒ The steps include determining each of the multiple candidate hydroxymethylated biomarker gene loci in the cfDNA obtained from each of the patients, (c) A step of selecting a candidate hydroxymethylated biomarker locus that exhibits a threshold p-value of less than 0.05 and a difference ζ of at least 1.5 as a hydroxymethylated biomarker locus, wherein, ζ=(T Ｒ -T ０ ) / T ０ The steps and Methods that include...

43. In each of the selected hydroxymethylated biomarker gene loci, log ２ (T Ｒ / T ０ ) The steps of calculating a value for x at each locus i, and the calculated value ｉ or y at each gene locus j ｊ The further step includes identifying as, where x ｉ and y ｊ The method according to claim 41, wherein the parameters are positively and negatively correlated with the efficacy of the treatment, respectively.

44. A method for calculating the 5hmC molecular genetic response score (MR 5hmC) as an indicator of whether a lung cancer patient is responding to lung cancer treatment, (a) In the cfDNA sample obtained from the patient, baseline count T in each of the hydroxymethylated biomarker gene loci selected in claim 41 ０ The steps to determine, (b) In the post-cfDNA sample obtained from the patient, the post-count T at each of the hydroxymethylated biomarker loci selected in claim 41 after the commencement of the treatment. Ｑ The steps to determine, (c) In each of the selected hydroxymethylated biomarker gene loci, log ２ (T Ｑ / T ０ ) The value for is calculated, and the calculated value is used for x at each gene locus i. ｉ or y at each gene locus j ｊ A step to identify as, where, x ｉ and y ｊ These are steps and, respectively, positively and negatively correlated with treatment efficacy. (e) Equation MR ５ｈｍＣ =μ ｘ -m ｙ Using the 5hmC molecular genetic response score (MR) for the patient, ５ｈｍＣ The step of calculating μ, where ｘ This is the x that spans the i locus. ｉ It is the average of μ ｙ This is the aforementioned y across the j locus. ｊ The average, step and Includes, A method in which a positive 5hmC molecular genetic response score indicates that the patient is responding to the treatment.

45. The method according to any one of claims 41 to 44, wherein the lung cancer treatment is immunotherapy.

Images