Method and system for measuring cellular state

Abstract

In various aspects of the present disclosure, methods and systems for detecting cellular states in a biological sample are provided. In one aspect of the present disclosure, a method for determining a cell type or cellular state is provided. In some embodiments, the method includes: providing or having previously provided a sample containing DNA or RNA, and generating a methylation profile of the DNA or RNA in the sample, or providing or having previously provided a methylation profile of the DNA or RNA in the sample.

Description

[Technical Field]

[0001] Cross-reference of related applications This application claims priority to U.S. Provisional Application No. 62 / 916,961, filed on 18 October 2020, which is incorporated herein by reference in its entirety.

[0002] Description of research and development funded by the federal government. Not applicable.

[0003] Reference materials Not applicable.

[0004] Field of Invention This disclosure relates, in general, to a method for detecting the cellular state in body fluids or nucleic acid mixtures. [Overview of the project]

[0005] Among the various aspects of this disclosure are the provision of methods and systems for detecting cellular states.

[0006] One aspect of this disclosure provides a method for determining a cell type or cell state. In some embodiments, the method includes providing or being provided with a sample containing DNA or RNA, and generating a methylation profile of the DNA or RNA in the sample, or providing or being provided with a methylation profile of the DNA or RNA in the sample. In some embodiments, the methylation profile includes co-associated CpG methylation patterns and methylated haplotype blocks (MHBs) (tightly bound CpG sites) of the DNA. In some embodiments, the method includes detecting a cell type or cell state by counting co-associated CpG methylation patterns in the DNA, including counting the co-associated CpG methylation patterns containing two or more CpGs in the DNA, or counting MHBs. In some embodiments, the method includes assigning the DNA to a cell type or cell state based on a reference CpG value or reference MHB value, the reference CpG value or reference MHB value being determined from a reference cell type or reference cell state. In some embodiments, the method comprises counting DNA molecules assigned to each reference CpG value or reference MHB value, each reference CpG value or reference MHB value corresponding to a cell type or cellular state. In some embodiments, the method further comprises counting known single CpG methylation profiles to increase sensitivity. In some embodiments, the sample is a blood sample. In some embodiments, the reference values ​​are differentially methylated CpGs derived from DNA of known cell types and known cellular states of bacterial, viral, fungal, or eukaryotic parasitic origin (optional). In some embodiments, the sample is plasma, tissue, or biopsy material. In some embodiments, the sample contains body fluids. In some embodiments, the body fluids are selected from whole blood, plasma, urine, saliva, or stool. In some embodiments, the sample does not contain solid tissue biopsy material. In some embodiments, the DNA or RNA is cell-free DNA or RNA and is derived from plasma.In some embodiments, the method includes determining a cell state-specific signature by the method described in claim 1, or providing or having been provided with a cell state-specific signature of a sample. In some embodiments, the DNA or RNA is cell-free and is rare cell type circulating DNA or RNA. In some embodiments, the sample comprises cell-free DNA (cfDNA) or cell-free RNA (cfRNA), and the sample is collected from the tumor microenvironment. In some embodiments, the tumor microenvironment comprises tumor-infiltrating leukocytes. In some embodiments, the DNA is cell-free tumor ctDNA. In some embodiments, the subject has been immunotherapyd before providing the sample. In some embodiments, the cellular state measured is derived from DNA from circulating cell-free tumor-infiltrating leukocytes (TILs) from the tumor microenvironment (TME). In some embodiments, the method includes profiling TILs according to methylation signatures and / or determining the proportion of different TIL subsets from cell type-specific methylation profiles identified in cell-free DNA. In some embodiments, the DNA is classified as originating from normal leukocytes, tumor-associated cells, or tumor-infiltrating leukocytes. In some embodiments, the method includes administering cancer treatment to a subject (e.g., immunotherapy, chemotherapy, radiation) and measuring cell type and cellular status in a sample as an indicator of treatment response. In some embodiments, if the ctilDNA level is lower than that of a responder to immunotherapy, the subject is determined to be at risk of being a non-responder to immunotherapy. In some embodiments, the sample contains cell-free DNA (cfDNA), and the sample is blood from a subject who has sepsis, is suspected of having sepsis, or is at risk of having sepsis. In some embodiments, the sample is a blood sample from a subject who has sepsis, is suspected of having sepsis, or is at risk of having sepsis. In some embodiments, the cellular status of exhausted lymphocytes is measured. In some embodiments, exhausted T cells are measured. In some embodiments, organ-specific cellular status or organ-specific cell type is measured.In some embodiments, the DNA originates from an organ, a damaged organ, a T cell, a debilitated T cell, an immune cell, a microorganism, a septic tissue, or a site of secondary infection. In some embodiments, if cfDNA analysis detects DNA originating from a microbial pathogen, the subject is diagnosed with an infection or sepsis. In some embodiments, if cfDNA analysis detects reduced cfDNA originating from a microbial pathogen compared to cfDNA originating from the microbial pathogen, and the subject is receiving treatment (e.g., antibiotics), the subject is determined to be responding to treatment. In some embodiments, if cfDNA analysis detects reduced cfDNA from a microbial pathogen compared to previously measured cfDNA analysis, the subject is determined to be responding to treatment or that the infection is improving. In some embodiments, if cfDNA analysis detects elevated cfDNA from organ tissue, the source of infection is determined to be the organ tissue with the elevated detected cfDNA. In some embodiments, if cfDNA analysis detects elevated cfDNA from suspected damaged organ tissue compared to a control, the organ is determined to be damaged. In some embodiments, if cfDNA analysis detects decreased cfDNA from damaged organ tissue compared to previously measured cfDNA analysis, organ damage is determined to be improving. In some embodiments, if cfDNA analysis detects elevated cfDNA from suspected damaged organ tissue compared to a control, the organ is determined to be damaged. In some embodiments, if cfDNA analysis detects elevated cfDNA from multiple organ systems compared to a control, the subject is determined to be at risk of multiple organ failure. In some embodiments, if cfDNA analysis detects elevated cfDNA from exhausted T cells or opportunistic pathogens compared to a control, the subject is determined to be at risk of secondary infection. In some embodiments, the DNA is cell-free DNA. In some embodiments, the method uses RNA instead of DNA.

[0007] Another aspect of the present disclosure provides a computer-aided method for detecting at least one abundance of at least one cell identification information in a biological sample, the sample comprising DNA. In some embodiments, the method comprises providing a plurality of reads, each read comprising a DNA sequence and associated methylation state. In some embodiments, the method comprises providing a CpG library comprising a plurality of entries, each entry comprising a CpG site and corresponding cell identification information, each CpG site comprising a co-associated CpG site, and each corresponding cell identification information comprising a cell type or cell state. In some embodiments, the method comprises using a computing device to convert a plurality of reads into a plurality of read assignments according to at least one assignment rule, each read assignment comprising one of cell identification information, cell-associated identification information, and irrelevant identification information. In some embodiments, the method comprises using a computing device to convert a plurality of read assignments into at least one abundance, each abundance corresponding to one cell identification information, and each abundance comprising the total number of read assignments comprising one cell identification information. In some embodiments, at least one assignment rule includes, if the read contains one or fewer CpG sites from multiple entries in a CpG library, using a computing device to convert the read into cell-related identification information; if the read contains at least two CpG sites from multiple entries in a CpG library having the same corresponding cell identification information, using a computing device to convert the read into cell identification information; and / or if the read does not contain any CpG sites from multiple entries in a CpG library, using a computing device to convert the read into irrelevant identification information. In some embodiments, the method includes using a computing device to convert each abundance into at least one of relative abundance and absolute abundance.In some embodiments, each relative abundance includes the abundance of one cell identification information normalized by the sum of all abundances of all cell identification information, and / or each absolute abundance includes the abundance of one cell identification information normalized by the sum of the abundance and the total number of read assignments. In some embodiments, providing multiple reads further includes performing bisulfite sequencing or microarray methylation profiling on DNA. In some embodiments, each CpG site is differentially methylated within the cell of one cell identification information, and each co-associated CpG site includes a sequence position proximal to at least one additional CpG site having the same corresponding cell identification information. In some embodiments, providing a CpG library includes providing a plurality of isolated DNAs corresponding to one cell identification piece; performing bisulfite sequencing or microarray methylation profiling on the plurality of isolated cfDNAs to obtain a plurality of isolated reads, wherein each isolated read includes the isolated sequence and associated methylation state of the isolated DNA; performing differential methylation region analysis on the plurality of isolated reads to identify a plurality of candidate CpG sites; and / or assigning a candidate CpG site as an entry in the CpG library for one cell identification piece if the candidate CpG site includes a proximal sequence position of at least one additional candidate CpG site. In some embodiments, the biological sample includes a body fluid. In some embodiments, the body fluid is selected from whole blood, plasma, urine, saliva, or feces. In some embodiments, the biological sample does not include solid tissue biopsy material. In some embodiments, the DNA is cell-free DNA. In some embodiments, instead of DNA, the method uses RNA.

[0008] A further aspect of the present disclosure provides a computing device configured to detect at least one abundance of at least one cell identification information in a biological sample, wherein the sample comprises DNA, the computing device comprises at least one processor and a non-volatile computer-readable medium, the non-volatile computer-readable medium comprises instructions executable on at least one processor, and provides a CpG library comprising a plurality of entries for receiving a plurality of reads, wherein each read comprises a sequence of DNA and associated methylation states, wherein each entry comprises a CpG site and corresponding cell identification information, each CpG site comprises a co-associated CpG site, and each corresponding cell identification information comprises a cell type or cell state, and / or provides a computing device for converting a plurality of reads into a plurality of read assignments according to at least one assignment rule, wherein each read assignment comprises one of cell identification information, cell-associated identification information, and irrelevant identification information, and / or converts the plurality of read assignments into at least one abundance, wherein each abundance corresponds to one cell identification information, and each abundance comprises the total number of read assignments comprising one cell identification information. In some embodiments, at least one assignment rule includes, if the read contains one or fewer CpG sites from multiple entries in a CpG library, using a computing device to convert the read into cell-related identification information; if the read contains at least two CpG sites from multiple entries in a CpG library having the same corresponding cell identification information, using a computing device to convert the read into cell identification information; and / or if the read does not contain any CpG sites from multiple entries in a CpG library, using a computing device to convert the read into irrelevant identification information.In some embodiments, the non-volatile computer-readable medium further includes instructions executable on at least one processor to convert each abundance into at least one of relative abundance and absolute abundance, where each relative abundance includes an abundance of one cell identification information normalized by the sum of all abundances of all cell identification information, and / or each absolute abundance includes an abundance of one cell identification information normalized by the sum of the abundance and the total number of read assignments. In some embodiments, each CpG site is differentially methylated within the cell of one cell identification information, and each co-associated CpG site includes a sequence position proximal to at least one additional CpG site having the same corresponding cell identification information. In some embodiments, the biological sample includes body fluid. In some embodiments, the body fluid is selected from whole blood, plasma, urine, saliva, or feces. In some embodiments, the biological sample does not include solid tissue biopsy material. In some embodiments, the DNA is cell-free DNA. In some embodiments, instead of DNA, the device detects RNA.

[0009] A further aspect of the present disclosure provides a computer-aided method for detecting at least one abundance of at least one cell identification information in a biological sample, wherein the sample comprises DNA, and the method comprises: providing a plurality of reads, each read comprising a DNA sequence and associated methylation state; providing a methylated haplotype block (MHB) library comprising a plurality of entries, each entry comprising an MHB and corresponding cell identification information, each MHB comprising at least two co-associated CpG sites, and each corresponding cell identification information comprising a cell type or cell state; using a computing device to convert the plurality of reads into a plurality of read assignments according to at least one assignment rule, each read assignment comprising one of cell identification information, cell-associated identification information, and irrelevant identification information; and / or using a computing device to convert the plurality of read assignments into at least one abundance, each abundance corresponding to one cell identification information, and each abundance comprising the total number of read assignments comprising one cell identification information. In some embodiments, if at least one assignment rule includes converting a read to cell identification information using a computing device, the read contains at least one MHB from multiple entries of an MHB library having corresponding cell identification information. In some embodiments, the method includes converting each abundance to a relative abundance using a computing device, where each relative abundance contains the abundance of one cell identification information normalized by the sum of all abundances of all cell identification information. In some embodiments, providing multiple reads further includes performing bisulfite sequencing or microarray methylation profiling on the DNA. In some embodiments, each MHB site contains at least two differentially methylated CpG sites that are in close proximity to each other within the cell of one cell identification information.In some embodiments, providing an MHB library further includes providing a plurality of isolated DNAs corresponding to one cell identification information; performing bisulfite sequencing or microarray methylation profiling on the plurality of isolated DNAs to obtain a plurality of isolated reads, wherein each isolated read includes the isolated sequence and associated methylation state of the isolated DNA; performing differential methylation region analysis on the plurality of isolated reads to identify a plurality of candidate CpG sites; and / or assigning each sequence containing at least two neighboring candidate CpG sites as an MHB corresponding to one cell identification information in the MHB library for one cell identification information. In some embodiments, the biological sample includes body fluids. In some embodiments, the body fluids are selected from whole blood, plasma, urine, saliva, or feces. In some embodiments, the biological sample does not include solid tissue biopsy material. In some embodiments, the DNA is cell-free DNA. In some embodiments, the method uses RNA instead of DNA.

[0010] A further aspect of the present disclosure is a computing device configured to detect at least one abundance of at least one cell identification information in a biological sample, wherein the sample comprises DNA, the computing device comprises at least one processor and a non-volatile computer-readable medium, the non-volatile computer-readable medium comprises instructions executable on at least one processor, and for receiving a plurality of reads, wherein each read comprises a DNA sequence and associated methylation state, and for receiving a methylated haplotype block (MHB) library comprising a plurality of entries, wherein each entry comprises the MHB and the corresponding cell identification information The computing device provides a computing device for converting multiple reads into multiple read assignments according to at least one assignment rule, where each read assignment includes one of cell identification information, cell-associated identification information, and irrelevant identification information, and / or converting multiple read assignments into at least one abundance, where each abundance corresponds to one cell identification information, and each abundance includes the total number of read assignments containing one cell identification information. In some embodiments, the at least one assignment rule includes converting reads into cell identification information using the computing device if the reads include at least one MHB from multiple entries of an MHB library having corresponding cell identification information. In some embodiments, the non-volatile computer-readable medium further includes instructions executable on at least one processor for converting each abundance into a relative abundance, where each relative abundance includes an abundance of one cell identification information normalized by the sum of all abundances of all cell identification information. In some embodiments, each MHB site comprises at least two differentially methylated CpG sites located in close proximity to each other within a cell of one cell identification information. In some embodiments, the biological sample comprises a body fluid. In some embodiments, the body fluid is selected from whole blood, plasma, urine, saliva, or feces.In some embodiments, the biological sample does not include solid tissue biopsy material. In some embodiments, the DNA is cell-free DNA. In some embodiments, the device detects RNA instead of DNA.

[0011] A further aspect of the present disclosure provides a computer-aided method for detecting the abundance of at least one of at least two cell identification information in a biological sample, wherein the sample comprises DNA, and the method comprises providing a plurality of reads, each read comprising a sequence of DNA and associated methylation states, providing a signature matrix comprising at least two plurality of differentially methylated CpG sites, each portion corresponding to each of the at least two cell identification information, and / or deconvolving the plurality of reads to at least two relative abundances, each relative abundance comprising a portion of one cell identification information in the biological sample. In some embodiments, the DNA is cell-free DNA. In some embodiments, instead of DNA, the method uses RNA.

[0012] A further aspect of the present disclosure provides a computing device configured to detect at least one abundance of at least two cell identification information in a biological sample, wherein the sample comprises DNA, the computing device comprises at least one processor and a non-volatile computer-readable medium, the non-volatile computer-readable medium comprising instructions executable on at least one processor, for receiving a plurality of reads, wherein each read comprises a sequence of DNA and associated methylation states, a signature matrix comprising at least two plurality of differentially methylated CpG sites, wherein each portion corresponds to each cell identification information of at least two cell identification information, and for deconvoluting the plurality of reads into at least two relative abundances, wherein each relative abundance comprises a portion of one cell identification information in the biological sample. In some embodiments, the DNA is cell-free DNA. In some embodiments, instead of DNA, the method uses RNA.

[0013] Other purposes and features are partially apparent and are partially noted below. Those skilled in the art will understand that the drawings shown below are for illustrative purposes only. The drawings are not intended to limit the scope of this instruction. [Brief explanation of the drawing]

[0014] [Figure 1] Methylation profiling reveals consistent TIL-specific signatures among colorectal cancer patients, but distinct signatures from peripheral blood leukocytes and tumor epithelial cells derived from colorectal cancer. The heatmap shows whole-genome bisulfite (WGBS) data from selected tumor (tum), tumor-infiltrating leukocyte (TIL), and peripheral blood leukocyte (PBL) populations from different colorectal cancer patients (columns), followed by differential methylation region (DMR) analysis. The 70 most distinctive CpG locations (rows) based on methylation status are shown here (blue vs. yellow = hypomethylation vs. hypermethylation), showing stereotypic similarities within each of the three populations, but distinct methylation signatures between them. [Figure 2] LiquidTME detects TIL signals in colorectal cancer plasma. Whole-genome bisulfite sequencing (WGBS) was applied to cell-free plasma DNA from 13 colorectal cancer (CRC) patients. The sequencing results were deconvoluted using CIBERSORTx with methylation signatures derived from the analysis in Figure 1. This analytical method is called LiquidTME. (a) Percentage of cell-free plasma DNA, consisting of DNA derived from tumor-infiltrating leukocytes (TILs) (red), tumor cells (blue), and normal peripheral blood leukocytes (gray), in patient (left) and healthy donor samples (right). (b) Comparison of plasma TIL DNA levels (left) and plasma-derived tumor DNA levels (right) determined by LiquidTME between CRC and healthy donors. The mean is represented by the horizontal gray bar. The P-value is calculated by a t-test with Welch's correction. [Figure 3] LiquidTME validation of TIL detection from plasma in colorectal cancer. Figure 2 shows plasma cfDNA (LiquidTME) results compared to tumor ground truth for nine colorectal cancer (CRC) patients with detectable plasma TIL signals. The X-axis represents the fraction of cell-free DNA derived from a specific population (tumor cells vs. TILs vs. PBLs), and the Y-axis represents the proportion of ground truth from tumor measurements and sequencing (CIBERSORTx deconvolution results multiplied by the sum of longest tumor diameters (SLD)). Data were analyzed in both rank space (indicated here by Spearman ρ) and non-rank space (indicated by Pearson r). Significance for both Spearman and Pearson correlations is indicated by P<0.05. There is a strong correlation between the levels of tumor signal in plasma compared to ground truth in the tumor (ρ=0.75, r=0.81). Surprisingly, there is a strong correlation between TIL DNA in plasma and ground truth in tumors (ρ=0.71, r=0.70). However, when the groups are cross-compared to each other, the positive correlation is not evident, which indicates specificity. [Figure 4]LiquidTME measurement of TIL signals from plasma strongly correlates with immunotherapy response in melanoma. Plasma cell-free DNA obtained within 4 weeks of immunotherapy initiation from 12 patients was analyzed by whole-genome bisulfite sequencing (WGBS) followed by CIBERSORTx deconvolution using our custom methylation signature matrix (see Figure 1). Eight out of 12 samples (67%) were detectable and are shown here. ctilDNA refers to the percentage of cell-free DNA resulting from TILs calculated by LiquidTME. (a) Melanoma patients are classified as immunotherapy responders (R) versus non-responders (NR), and the percentage of ctilDNA is shown in red. (b) Receiver operating characteristic (ROC) analysis of response status based on ctilDNA yields an area under the curve (AUC) of 0.94 with a P-value of 0.04, indicating that ctilDNA levels function as a strong classifier of response. (c) Kaplan-Meier analysis of progression-free survival stratified by the optimal cut point from the ROC analysis of panel b (12%) nearly completely stratifies persistent responders from rapidly progressing early disease patients with a hazard ratio of 9.3 and a p-value of 0.03. [Figure 5] Differential methylation CpG sites in a purified leukocyte subset after methylation sequencing. The heatmap shows whole-genome bisulfite (WGBS) data and subsequent differential methylation region (DMR) analysis of a selected leukocyte subset (shown above). Identifiable CpG locations (rows) based on methylation status are shown here (blue vs. yellow = hypomethylation vs. hypermethylation). [Figure 6] Ultra-high-resolution digital cytometry was performed by detecting co-associated CpGs within methylated sequencing read pairs and using these to assign each read to a matching reference cell type / state. A bulk leukocyte mixture was sequenced by whole-genome bisulfite sequencing (WGBS). Ultra-high-resolution digital cytometry was performed using a different number of co-associated CpGs per read pair and correlated with flow cytometry grand truth. Pearson r and associated P values ​​are shown to quantify the strength of the correlation. [Figure 7]Ultra-high resolution digital cytometry in relative and absolute modes. A bulk leukocyte mixture was sequenced by whole-genome bisulfite sequencing (WGBS). Ultra-high resolution digital cytometry was performed to detect co-associated CpGs for each read pair, and each read pair was subsequently assigned to its corresponding reference cell type / state. Results are shown in relative mode (left), where reads assigned to reference are quantified relative to each other, and in absolute mode (right), where fragments assigned to reference are normalized relative to the total number of unique reads with overlapping CpG locations. In both cases, the results of ultra-high resolution digital cytometry correlated with flow cytometry grand truth. Pearson r and associated P values ​​are shown to quantify the strength of the correlation. [Figure 8] This diagram shows how tumors release cells and genetic material into the bloodstream (circulation). While ctDNA has been described previously, it was discovered here that ctilDNA is also present in peripheral blood. [Figure 9] This map shows clonally related CD8 T cells and T cell exhaustion signatures between tissue compartments. RNA-seq reveals TIL-specific cellular states that are different from normal. Left: Single-cell RNA sequencing identifies CD8TIL gene expression profiles that are different from normal. Clones are distinguished by color. Right: Gene set enrichment analysis showing that exhaustion genes are upregulated in CD8TIL compared to normal CD8 T cells. [Figure 10]This is a flowchart and a series of graphs illustrating the modeling of ctilDNA detection. Theoretical detection limit modeling of LiquidTME. A) Top: Typical cell-free DNA yield and sequencing depth from 10 mL of blood. Bottom: After target capture and bisulfite sequencing, the inventors conservatively estimate an 80% loss of input molecules and estimate TIL content considering the median level of ctDNA (approximately 1%) in patients with advanced solid tumors. Assuming equal proportions of DNA release from cancer cells and TILs, and an average TIL content of 30%, the inventors estimate that approximately 0.4% of cell-free DNA is ctilDNA. B) Left: Average estimated percentage of major TIL subsets in advanced cancer. Right: Corresponding percentage of TIL subsets in cell-free DNA based on the assumptions in panel a. C) Binary probability for detecting at least one TIL subset based on the number of reporters (DMRs) targeted by the LiquidTME assay, considering the above assumptions (and a 2,000x deduplication sequencing depth). D) Same as panel C, but shows the expected number of reporters (DMRs) detected as a function of the number of targets. DMR: Differential methylation region. Duplicate removal: After removal of duplicate sequencing reads. [Figure 11] Strategies used to develop and validate the LiquidTME assay and its clinical application. [Figure 12] Liquid biopsy reveals TME signals in plasma. A) TIL and tumor cell signatures were detected in cell-free DNA and tumors from three CRC patients, but not in peripheral blood mononuclear cells (PBMCs). B) Presumptive TIL and tumor cell levels based on plasma cell-free DNA whole-genome bisulfite (WGBS) analysis correlated strongly with flow cytometry and imaging. [Figure 13] TIL signals measured by LiquidTME correlate with melanoma immunotherapy response. A) ctilDNA levels measured by LiquidTME and stratified by response (DCB = permanent clinical benefit, NDB = no permanent benefit). [Figure 14]This figure shows the development of assays for non-invasive TME profiling and measurement of technical and in vivo performance. [Figure 15] Cryopreservation does not introduce epigenetic artifacts. Left: Genomic regions with over 75% methylation in fresh cells versus cryopreserved cells from the same healthy donor. The Jacquard coefficient indicates the similarity between the two datasets. Right: Heatmap shows inter-genomic methylation in three fresh samples, three frozen cell samples, and three frozen DNA samples from the same donor. [Figure 16] Visualization of differential methylation of the PDCD1 gene in CD8 T cells. Top 3: Three CD8T TIL samples purified from independent CRC patient tumors. Bottom 7: The three top PBL CD8T samples are from these same CRC patients, and the four bottom samples are from a BLUEPRINT healthy donor. [Figure 17] Strategies, technical optimization, and testing for the development of the LiquidTME assay; validation of the inventors' technology; and clinical application of LiquidTME. [Figure 18] Enumeration of leukocyte subsets by CIBERSORT deconvolution of whole blood methylation profiles. (a, b) Scatter plots showing deconvolution performance for flow cytometry in two publicly available datasets, namely Chakravarthy et al. (a) and Accomando et al. (b). [Figure 19] LiquidMIDOS will be an all-in-one liquid biopsy technology prepared to revolutionize the diagnosis, monitoring, management, and ultimately survival of sepsis patients. [Figure 20] A fatal hyperimmune response typically prevails during the first few days of sepsis (A). This is followed by a potentially self-limiting (B) or fatal (C) hypoimmune phase resulting from T-cell dysfunction / exhaustion, which increases the risk of secondary infections and may be responsive to immunotherapy (D). (Cited from Boomer et al., 2014.) [Figure 21]Plasma cfDNA source in sepsis. Modified from Crowley et al., 2013. [Figure 22] In hospitalized patients, liver-derived plasma cell-free DNA levels (Y-axis) are significantly correlated with serum ALT (X-axis), a gold-standard biomarker for liver injury. (Moss et al., 2018) [Figure 23] Left: Yield of cell-free DNA from 10 mL of blood undergoing whole-genome bisulfite sequencing after bisulfite conversion and library preparation. Center: Estimated sources and their relative proportions of plasma cell-free DNA in sepsis patients, based on Moss et al. and Grumaz et al. Right: Binomial probabilities of detecting each queried cell-free DNA compartment as a function of the number of specific reporters. [Figure 24] A FACS sorting scheme for exhausted T cells from tissue using standard surface marker staining. [Figure 25] Plasma cell-free DNA versus tumor ground truth from nine colorectal cancer patients with detected cfDNA in epithelial signals (left) and tissue lymphocyte signals (right). Data analyzed in rank space (illustrated; Spearman ρ) and non-rank space (Pearson r). [Figure 26] Whole-genome sequencing detected spike-in shear microbial DNA (diluted in human plasma at 32–1,000 molecules / microliter) from Staphylococcus aureus (S. aureus), Staphylococcus epidermidis (S. epidermidis), and adenovirus B with high sensitivity and specificity, as evaluated by sequencing of four independent healthy donor plasma cell-free DNA samples. [Figure 27] This is a block diagram illustrating a system according to one aspect of this disclosure. [Figure 28] This is a block diagram illustrating a computing device according to one aspect of the present disclosure. [Figure 29] This is a schematic block diagram illustrating a remote computing device or user computing device according to one aspect of the present disclosure. [Figure 30]This is a block diagram illustrating a server system according to one aspect of this disclosure. [Modes for carrying out the invention]

[0015] This disclosure is at least in part based on the finding that cellular status can be measured in tissue or body fluids. It should be noted that the scope of this method is not limited to DNA methylation or plasma-derived cell-free DNA. This method can be applied to any sequenced nucleic acid mixture (i.e., DNA or RNA) from any cell or cell-free DNA source (i.e., any body fluid or tissue source). The examples disclosed herein use bisulfite / methylation sequencing, but this method can be used in conjunction with any type of next-generation sequencing or microarray technology known in the art (see, for example, Rajesh et al., 2017 - Next-Generation Sequencing Methods; Current Developments in Biotechnology and Bioengineering: Functional Genomics and Metabolic Engineering 2017, pp. 143-158; Moss et al., 2018 Comprehensive human cell-type methylation atlas reveals origins of circulating cell-free DNA in health and disease. Nat Commun 9, 5068; Bumgarner, 2013, Overview of DNA Microarrays: Types, Applications, and Their Future, Vol. 101, No. 1, pp. 22.1.1-22.1.11).

[0016] As described herein, the methods of this disclosure enable the detection and profiling of the tumor microenvironment (including tumor-infiltrating leukocytes and tumor cell status) using a blood-based liquid biopsy approach. This is done by methylation sequencing of plasma-derived cell-free DNA (see, for example, Figures 8 and 21 showing genetic material released from cells such as cancer cells, microbial cells, and infected cells that can be detected by this method). Individual single-cell statuses are profiled from the bulk using either genome-wide or targeted bisulfite sequencing (e.g., leukocyte and tumor cell status by counting plasma methylation sequencing data or, optionally, by deconvolution).

[0017] This method is not deconvolution, but rather single-molecule counting, allowing molecules (DNA or RNA) to be enumerated at the molecular level and classified into a reference bin. Therefore, this method involves counting, not deconvolution. By starting with individual molecules and enumerating and classifying them one by one, it learns how the complete system is structured molecularly. This gives the method very high resolution.

[0018] In some embodiments, a machine learning model may be used to enumerate DNA or RNA molecules and classify them into a reference bin. In these embodiments, the machine learning model may be trained using DNA or RNA molecules obtained from isolated cell types or cellular states described herein.

[0019] On the other hand, deconvolution begins by viewing the entire bulk sequenced mixture as a whole, and then attempts to weigh and add cell type-specific signatures together to achieve a mixture representation matrix. Therefore, the deconvolution method has a fundamentally lower resolution and is fundamentally different from the disclosed method. The specific technological advancement that has been achieved is error suppression based on methylation haplotype blocking ("pseudo-UMI") (described in Example 1).

[0020] This method allows for the enumeration and differentiation of cell types and / or cellular states without requiring solid tissue biopsy material. “Cellular state” can be defined as a context-dependent version of a given cell type (e.g., normal versus tumor-associated CD8 T cells). This unique capability makes the non-invasive approach of this disclosure possible in measuring non-malignant cells within a tumor and differentiating them from their normal tissue counterparts. This is believed to be the first time this has been achieved. Previous studies have focused solely on cell types, histological types, and differentiating cancer versus normal cells, all of which are less granular than cellular states.

[0021] The disclosed method relies on prior knowledge of cell state-specific signatures (e.g., from known cells). These signatures enable this approach to enumerate specific cell types and cell states directly from methylation signals in cell-free DNA. Such signatures can be derived by physically isolating the cell states of interest by FACS or by inferring them by single-cell bisulfite sequencing. However, these methods have significant drawbacks, such as variable loss of specific cell types due to tissue dissociation, the sensitivity and specificity of antibody panels (required for FACS), and the small amount of tissue typically obtained from tumor biopsy material. Therefore, we have developed a novel alternative to complement these techniques. Our approach is based on inferring cell state signatures directly from bulk tumor methylation profiles. The inventors can achieve this by statistical deconvolution in a process that is essentially the reverse of measuring cell composition from bulk methylation profiles (e.g., CIBERSORTx; Newman et al. (2019) Nature Biotechnology (37) pp. 773-782). This novel approach can be used to flexibly generate signatures for virtually any desired cell state without the need for antibodies, live cells, or physical cell isolation.

[0022] It should be noted that the scope of this method is not limited to DNA methylation or plasma-derived cell-free DNA. This method can be applied to any sequenced nucleic acid mixture from any cell source or cell-free DNA or RNA source (i.e., any bodily fluid or tissue source). Method and system for non-invasively measuring the cellular state in body fluids

[0023] This disclosure provides a non-invasive measurement method for measuring the cellular state in body fluids or biological fluids. More specifically, it enumerates specific cell types and cellular states directly from methylation signals present in cell-free DNA.

[0024] As described herein, this technique can identify cell types and cellular states in single cells or bulk mixtures of cells. Cellular states can be defined as cellular phenotypes. Cellular phenotypes may be “homeostatic phenotypes” that imply plasticity arising from dynamically changing but characteristic patterns of gene / protein expression.

[0025] The methods described herein can be applied to a wide range of commercial / biomedic problems, including immunotherapy response assessment, immunotherapy toxicity assessment, response of any tumor to any drug, non-invasive tracking of the tumor microenvironment in research, clinical, or commercial applications, and enabling true fluid biopsies of tumors, including both cancer and tumor microenvironment profiling.

[0026] This technology can be used for a wide variety of applications by utilizing any type of epigenetic data (i.e., whole-genome bisulfite sequencing, reduced-expression bisulfite sequencing, methylation microarrays, etc.) from any bodily fluid (e.g., urine, saliva, plasma, stool, etc.).

[0027] This method enables the detection and profiling of the tumor microenvironment (including tumor-infiltrating leukocytes and tumor cell status) using a liquid biopsy approach. The inventors perform this by methylation sequencing of plasma-derived cell-free DNA, followed by digital cytometry (deconvolution). The inventors profiled individual single-cell statuses from bulk (e.g., leukocyte and tumor cell status by deconvolution of plasma methylation sequencing data) using either genome-wide or targeted bisulfite sequencing.

[0028] This method, while presented herein for detecting cellular status and cell type in cell-free DNA, may also be a useful method for use in conjunction with nucleic acid sequencing of any length. The nucleic acids may be full-length DNA, DNA fragments, cell-free DNA, RNA, or cell-free nucleic acid fragments assigned to cell types derived from microorganisms such as tumor cells, infected cells, injured cells, normal cells, bacterial cells, organ or tissue cells, tissue cells secreting cfDNA, bacteria, viruses (DNA or RNA), fungi, or eukaryotic parasites. In some embodiments, the DNA fragments may be approximately 300 base pairs or less. It should also be noted that the scope of this method is not limited to DNA methylation or plasma-derived cell-free DNA. This method can be applied to any sequenced or microarray-profiled nucleic acid mixture from any cell source or cell-free DNA source (i.e., any bodily fluid or tissue source).

[0029] As described herein, one or more CpG methylation sites are detected. CpG methylation sites can co-associate (e.g., proximal or near each other) between any number of base pairs along the length of the DNA molecule. In some embodiments, the amount of base pairs between co-associated CpGs may be about 1 base pair (bp) to about 1000 bp (proximal or near each other), 1 bp to about 500 bp, or about 1 bp to about 300 bp. For example, nearby or proximal CpGs are approximately 1 bp; 2 bp; 3 bp; 4 bp; 5 bp; 6 bp; 7 bp; 8 bp; 9 bp; 10 bp; 11 bp; 12 bp; 13 bp; 14 bp; 15 bp; 16 bp; 17 bp; 18 bp; 19 bp; 20 bp; 21 bp; 22 bp; 23 bp; 24 bp; 25 bp; 26 bp; 27 bp; 28 bp; 29 bp; 30 bp; 31 bp; 32 bp; 33 bp; about 34bp; about 35bp; about 36bp; about 37bp; about 38bp; about 39bp; about 40bp; about 41bp; about 42bp; about 43bp; about 44bp; about 45bp; about 46bp; about 47bp; about 48bp; about 49bp; about 5 0bp; approximately 51bp; approximately 52bp; approximately 53bp; approximately 54bp; approximately 55bp; approximately 56bp; approximately 57bp; approximately 58bp; approximately 59bp; approximately 60bp; approximately 61bp; approximately 62bp; approximately 63bp; approximately 64bp; approximately 65bp; approximately 66bp; 7bp; approximately 68bp; approximately 69bp; approximately 70bp; approximately 71bp; approximately 72bp; approximately 73bp; approximately 74bp; approximately 75bp; approximately 76bp; approximately 77bp; approximately 78bp; approximately 79bp; approximately 80bp; approximately 81bp; approximately 82bp; 84bp; about 85bp; about 86bp; about 87bp; about 88bp; about 89bp; about 90bp; about 91bp; about 92bp; about 93bp; about 94bp; about 95bp; about 96bp; about 97bp; about 98bp; about 99bp; about 100bp; About 101bp; About 102bp; About 103bp; About 104bp; About 105bp; About 106bp; About 107bp; About 108bp; About 109bp; About 110bp; About 111bp; About 112bp; About 113bp; About 114bp; About 11 5bp; about 116bp; about 117bp; about 118bp; about 119bp; about 120bp; about 121bp; about 122bp; about 123bp; about 124bp; about 125bp; about 126bp; about 127bp; about 128bp; about 129bp;Approximately 130bp; Approximately 131bp; Approximately 132bp; Approximately 133bp; Approximately 134bp; Approximately 135bp; Approximately 136bp; Approximately 137bp; Approximately 138bp; Approximately 139bp; Approximately 140bp; Approximately 141bp; Approximately 142bp; Approximately 143bp; Approximately 144bp; Approximately 145bp; Approximately 146bp; Approximately 147bp; Approximately 148bp; Approximately 149bp; Approximately 150bp; Approximately 151bp; Approximately 152bp; Approximately 153bp; Approximately 154bp; Approximately 155bp; Approximately 156bp; Approximately 157bp; Approximately 158bp; Approximately 159bp; Approximately 160bp; Approximately 161bp; Approximately 162bp; Approximately 163bp; Approximately 164bp; Approximately 165bp p; approx. 166bp; approx. 167bp; approx. 168bp; approx. 169bp; approx. 170bp; approx. 171bp; approx. 172bp; approx. 173bp; approx. 174bp; approx. 175bp; approx. 176bp; approx. 177bp; approx. 178bp; approx. 179bp; approx. 180bp; approx. 181bp; approx. 182bp; approx. 183bp; approx. 184bp; approx. 185bp; approx. 186bp; approx. 187bp; approx. 188bp; approx. 189bp; approx. 190bp; approx. 191bp; approx. 192bp; approx. 193bp; approx. 194bp; approx. 195bp; approx. 196bp; approx. 197bp; approx. 198bp; approx. 199bp; approx. 200bp; approx. 20 1bp; approximately 102bp; approximately 203bp; approximately 204bp; approximately 205bp; approximately 206bp; approximately 207bp; approximately 208bp; approximately 209bp; approximately 210bp; approximately 211bp; approximately 212bp; approximately 213bp; approximately 214bp; approximately 215bp; approximately 216bp; approximately 217bp; approximately 218bp; approximately 219bp; approximately 220bp; approximately 221bp; approximately 222bp; approximately 223bp; approximately 224bp; approximately 225bp; approximately 226bp; approximately 227bp; approximately 228bp; approximately 229bp; approximately 230bp; approximately 231bp; approximately 232bp; approximately 233bp; approximately 234bp; approximately 235bp; approximately 236bp; approximately 237bp; approx. 238bp; approx. 239bp; approx. 240bp; approx. 241bp; approx. 242bp; approx. 243bp; approx. 244bp; approx. 245bp; approx. 246bp; approx. 247bp; approx. 248bp; approx. 249bp; approx. 250bp; approx. 251bp; approx. 252bp; approx. 253bp; approx. 254bp; approx. 255bp; approx. 256bp; approx. 257bp; approx. 258bp; approx. 259bp; approx. 260bp; approx. 261bp; approx. 262bp; approx. 263bp; approx. 264bp; approx. 265bp; approx. 266bp; approx. 267bp; approx. 268bp; approx. 269bp; approx. 270bp; approx. 271bp; approx. 272bp;They can be separated by approximately 273bp; 274bp; 275bp; 276bp; 277bp; 278bp; 279bp; 280bp; 281bp; 282bp; 283bp; 284bp; 285bp; 286bp; 287bp; 288bp; 289bp; 290bp; 291bp; 292bp; 293bp; 294bp; 295bp; 296bp; 297bp; 298bp; 299bp; or approximately 300bp.

[0030] The control or reference samples described herein may be samples from healthy subjects. Reference values ​​may be used in place of control or reference samples previously obtained from healthy subjects or groups of healthy subjects. Control or reference samples may also be samples with known cellular or tumor compositions. Computing systems and devices

[0031] In various embodiments, the methods described herein are carried out using computing devices and systems. Figure 27 shows a simplified block diagram of a system 800 for carrying out the methods described herein. As shown in Figure 27, system 800 may be configured to implement at least some of the tasks related to the disclosed methods. System 800 may include a computing device 802. In one embodiment, computing device 802 is part of a server system 804, which also includes a database server 806. Computing device 802 communicates with database 808 through database server 806 via a network. Network 850 may be any network that enables local area communication or wide area communication between devices. For example, network 850 may enable communication coupling to the Internet via at least one of many interfaces, including but not limited to at least one of networks such as the Internet, Local Area Network (LAN), Wide Area Network (WAN), Integrated Services Digital Network (ISDN), Dial-up Connection, Digital Subscriber Line (DSL), Mobile Phone Connection, and Cable Modem. The user computing device 830 may be any device capable of accessing the Internet, including but not limited to a desktop computer, laptop computer, personal digital assistant (PDA), mobile phone, smartphone, tablet, phablet, wearable electronic device, smartwatch, or other web-based connectable device or mobile device.

[0032] In other embodiments, computing device 802 is configured to perform a number of tasks related to the method for detecting the abundance of cellular states and / or cell types described herein. Figure 28 shows a component configuration 400 of computing device 402, including database 410 along with other related computing components. In some embodiments, computing device 402 is similar to computing device 802 (shown in Figure 27). User 404 may access the components of computing device 402. In some embodiments, database 420 is similar to database 808 (shown in Figure 27).

[0033] In one embodiment, database 410 includes library data 418, algorithm data 412, ML model data 416, and sample data 420. In one embodiment, library data 418 includes library entries defining the characteristics of different cell types or cell states whose abundances are detected as described herein. Non-limiting examples of library data 418 include entries for a CpG library, entries for a methylated haplotype block (MHB) library, and signature matrices. As used herein, a CpG library is defined as a plurality of entries, each entry containing differentially methylated CpG sites representing one of the cell types or cell states. In some embodiments, differentially methylated CpG sites are further co-associated CpG sites. As used herein, co-associated CpG sites refer to differentially methylated CpG sites that characterize one of the cell types or cell states located at a distance of about 200 bp or less from additional differentially methylated CpG sites characterizing the same cell type or cell state. As used herein, an MHB library is defined as a plurality of entries, each entry containing at least two co-associated CpG sites representing one of the cell types or cellular states. As used herein, a signature matrix contains a plurality of differentially methylated CpG sites characterizing all of at least one cell type or cellular state. The signature matrix is ​​used as part of a digital deconvolution method described herein. A non-limiting example of a preferred digital deconvolution method is CIBERSORTx.

[0034] In various embodiments, algorithm data 412 includes any parameters used to carry out the method described herein. Non-limiting examples of preferred algorithm data 412 include any values ​​of parameters that define the calculation of abundance counts, relative abundances, absolute abundances, and any other relevant parameters. Non-limiting examples of ML model data 416 include any values ​​of parameters that define a machine learning model used to optimize a CpG library to perform digital deconvolution and any other transformation, classification, or other task according to the method described herein. Non-limiting examples of sample data 420 include any multiple reads related to the analysis of a biological sample by the method described herein, including DNA sequences, RNA sequences, DNA methylation sequences, and any other suitable nucleic acid sequences.

[0035] The computing device 402 also includes several components for performing specific tasks. In an exemplary embodiment, the computing device 402 includes a data storage device 430, an abundance component 440, an analysis component 450, an ML component 470, and a communication component 460. The data storage device 430 is configured to store data received or generated by the computing device 402, such as any of the data stored in the database 410 or any output of a process implemented by any component of the computing device 402. The abundance component 450 is configured to convert a plurality of reads associated with a sample for each of the at least one cell type or cellular state detected according to the method described herein into at least one abundance, at least one relative abundance, at least any absolute abundance, or any combination thereof. The analysis component 450 is configured to perform any additional analysis of any of the abundances generated in connection with the method described herein. Non-limiting examples of additional analyses performed using the analysis component 450 include the diagnosis of a disease or disorder such as cancer or sepsis, classification of patients into categories such as responders or non-responders to treatment, determination of treatment effectiveness, and any other suitable analysis. In various embodiments, the ML component 470 is configured to implement any of the machine learning model-based transformations and analyses described herein. Non-limiting examples of transformations or analyses implemented using the ML component 470 include digital deconvolution of cell types or cellular states based on multiple reads in a mixed sample. Optimization of CpG libraries or MHB libraries, or any other suitable transformations or analyses, follows the methods described herein.

[0036] The communication component 460 is configured to enable communication of the computing device 402 over a network such as network 850 (shown in Figure 27), or over multiple network connections using a predetermined network protocol such as TCP / IP (Transmission Control Protocol / Internet Protocol).

[0037] Figure 29 shows the configuration of a remote computing device or user computing device 502, such as a user computing device 830 (shown in Figure 27). The computing device 502 may include a processor 505 for executing instructions. In some embodiments, executable instructions may be stored in a memory area 510. The processor 505 may include one or more processing units (e.g., in a multi-core configuration). The memory area 510 may be any device that enables the storage and retrieval of information such as executable instructions and / or other data. The memory area 510 may include one or more computer-readable media.

[0038] The computing device 502 may also include at least one media output component 515 for presenting information to the user 501. The media output component 515 may be any component capable of conveying information to the user 501. In some embodiments, the media output component 515 may include output adapters such as a video adapter and / or an audio adapter. The output adapter may be operably coupled to the processor 505 and may be operably coupled to an output device such as a display device (e.g., a liquid crystal display (LCD), an organic light-emitting diode (OLED) display, a cathode ray tube (CRT), or an "electronic ink" display) or an audio output device (e.g., a speaker or headphones). In some embodiments, the media output component 515 may be configured to present an interactive user interface (e.g., a web browser or a client application) to the user 501.

[0039] In some embodiments, the computing device 502 may include an input device 520 for receiving input from the user 501. The input device 520 may include, for example, a keyboard, a pointing device, a mouse, a stylus, a touch-sensitive panel (e.g., a touchpad or touchscreen), a camera, a gyroscope, an accelerometer, a position detector, and / or an audio input device. A single component, such as a touchscreen, may function as both an output device and an input device 520 for the media output component 515.

[0040] The computing device 502 may also include a communication interface 525 which may be communicatively coupled to a remote device. The communication interface 525 may include, for example, a wired or wireless network adapter or a wireless data transceiver for use with a cellular network (e.g., Global Mobile Communication System (GSM), 3G, 4G, or Bluetooth) or another mobile data network (e.g., Worldwide Interoperability for Microwave Access (WiMAX)).

[0041] The memory area 510 stores, for example, computer-readable instructions for providing a user interface to user 501 via media output component 515, and optionally for receiving and processing input from input device 520. The user interface may also include, among other possibilities, a web browser and a client application. The web browser allows user 501 to view and interact with media and other information embedded on web pages or websites, typically from a web server. The client application allows user 501 to interact with server applications associated with, for example, a vendor or company.

[0042] Figure 30 shows an example configuration of server system 602. Server system 602 may include, but is not limited to, a database server 806 and a computing device 802 (both shown in Figure 27). In some embodiments, server system 602 is similar to server system 804 (shown in Figure 27). Server system 602 may include a processor 605 for executing instructions. Instructions may be stored, for example, in a memory area 625. Processor 605 may include one or more processing units (for example, in a multi-core configuration).

[0043] The processor 605 may be operably coupled to a communication interface 615 so that the server system 602 may communicate with a user computing device 830 (shown in Figure 27) or a remote device such as another server system 602. For example, the communication interface 615 may receive requests from the user computing device 830 via a network 850 (shown in Figure 27).

[0044] The processor 605 may also be operably coupled to the storage device 625. The storage device 625 may be any computer operating hardware suitable for storing and / or retrieving data. In some embodiments, the storage device 625 may be integrated into a server system 602. For example, the server system 602 may include one or more hard disk drives as the storage device 625. In other embodiments, the storage device 625 may be external to the server system 602 and may be accessed by multiple server systems 602. For example, the storage device 625 may include multiple storage units such as hard disks or solid-state disks in a redundant array of inexpensive disks (RAID) configuration. The storage device 625 may include a storage area network (SAN) and / or network-attached storage (NAS) system.

[0045] In some embodiments, the processor 605 may be operably coupled to a storage device 625 via a storage interface 620. The storage interface 620 may be any component capable of providing the processor 605 with access to the storage device 625. The storage interface 620 may include, for example, an Advanced Technology Attachment (ATA) adapter, a Serial ATA (SATA) adapter, a Small Computer System Interface (SCSI) adapter, a RAID controller, a SAN adapter, a network adapter, and / or any component that provides the processor 605 with access to the storage device 625.

[0046] Memory regions 510 (shown in Figure 29) and 610 may include, but are not limited to, random-access memory (RAM) such as dynamic RAM (DRAM) or static RAM (SRAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), and non-volatile RAM (NVRAM). The above memory types are merely examples and therefore do not limit the types of memory that can be used to store computer programs.

[0047] The computer systems and computer implementations described herein may include additional, fewer, or alternative operations and / or functions, including those described elsewhere herein. The computer systems may include, or be implemented through, computer-executable instructions stored in a non-temporary computer-readable medium. The methods may be implemented via one or more local, remote, or cloud-based processors, transceivers, servers, and / or sensors (e.g., processors, transceivers, servers, and / or sensors mounted in a vehicle or mobile device, or associated with smart infrastructure or a remote server), and / or via computer-executable instructions stored in a non-temporary computer-readable medium or medium.

[0048] In some embodiments, a computing device is configured to implement machine learning so that the computing device “learns” to analyze, organize, and / or process data without being explicitly programmed. Machine learning may be implemented by machine learning (ML) methods and algorithms. In one embodiment, a machine learning (ML) module is configured to implement ML methods and algorithms. In some embodiments, ML methods and algorithms are applied to a data input to produce a machine learning (ML) output. The data input may include, but is not limited to, images or frames from a video, object characteristics, and object classifications. The data input may further include sensor data, image data, video data, telematics data, authentication data, authorization data, security data, mobile device data, geolocation data, transaction data, personally identifiable data, financial data, usage data, weather pattern data, “big data” sets, and / or user preference data. ML outputs may include, but are not limited to, tracked shape outputs, object classification, motion type classification, object motion-based diagnostics, and object motion analysis. Trained model parameter ML outputs may further include speech recognition, image or video recognition, functional connectivity data, medical diagnostics, statistical or financial models, autonomous vehicle decision-making models, robot behavior modeling, fraud detection analysis, user recommendations and personalization, game AI, skill acquisition, targeted marketing, big data visualization, weather forecasting, and / or information extracted about computer devices, users, homes, vehicles, or parts of transactions. In some embodiments, data inputs may include specific ML outputs.

[0049] In some embodiments, at least one of several ML methods and algorithms may be applied, which may include, but are not limited to, linear or logistic regression, instance-based algorithms, regularization algorithms, decision trees, Bayesian networks, cluster analysis, association rule learning, artificial neural networks, deep learning, dimensionality reduction, and support vector machines. In various embodiments, the implemented ML methods and algorithms target at least one of several classifications of machine learning, such as supervised learning, unsupervised learning, and reinforcement learning.

[0050] In one embodiment, the ML methods and algorithms are intended for supervised learning, which involves identifying patterns in existing data and making predictions about subsequently received data. Specifically, the ML methods and algorithms intended for supervised learning are “trained” through training data, which includes exemplary inputs and associated exemplary outputs. Based on the training data, the ML methods and algorithms may generate a prediction function that maps outputs to inputs and use the prediction function to generate ML outputs based on data inputs. The exemplary inputs and outputs of the training data may include either the data inputs or ML outputs described above. For example, an ML module may receive training data including customer identification and geographic information and associated customer categories, generate a model that maps customer categories to customer identification and geographic information, and generate an ML output that includes customer categories for subsequently received data inputs including customer identification and geographic information.

[0051] In another embodiment, ML methods and algorithms address unsupervised learning, which involves finding meaningful relationships within unorganized data. Unlike supervised learning, unsupervised learning does not involve user-initiated training based on exemplary inputs with relevant outputs. Rather, in unsupervised learning, unlabeled data, which may be any combination of the above data inputs and / or ML outputs, is organized according to relationships determined by the algorithm. In one embodiment, an ML module receives unlabeled data, including customer purchase information, customer mobile device information, and customer geolocation information, and the ML module uses an unsupervised learning method, such as "clustering," to identify patterns and organize the unlabeled data into meaningful groups. The newly organized data may be used, for example, to extract further information about customer consumption habits.

[0052] In yet another embodiment, the ML methods and algorithms address reinforcement learning, which involves optimizing the output based on feedback from a reward signal. Specifically, the ML methods and algorithms addressing reinforcement learning may receive a user-defined reward signal definition, receive a data input, use a decision model to generate an ML output based on the data input, receive a reward signal based on the reward signal definition and the ML output, and subsequently modify the decision model to receive a stronger reward signal for the generated ML output. The reward signal definition may be based on either the data input or the ML output described above. In one embodiment, the ML module implements reinforcement learning in a user recommendation application. The ML module may use a decision model to generate a ranked list of options based on user information received from the user, and may further receive selection data based on the user's selection of one of the ranked options. The reward signal may be generated based on comparing the selection data with the ranking of the selected option. The ML module may update the decision model so that the subsequently generated rankings more accurately predict the user's selection.

[0053] As understood in accordance with the preceding specification, the above aspects of the Disclosure may be implemented using computer programming or engineering techniques, including computer software, firmware, hardware, or any combination or subset thereof. Any resulting program having computer-readable code means may be embodied or provided in one or more computer-readable media, thereby creating a computer program product, i.e., a product, in the manner described in the Disclosure. The computer-readable media may be, but are not limited to, fixed (hard) drives, diskettes, optical disks, magnetic tapes, semiconductor memory such as read-only memory (ROM), and / or any transmitting / receiving medium such as the Internet or other communication networks or links. A product containing computer code may be created and / or used by executing the code directly from one medium, by copying the code from one medium to another, or by transmitting the code over a network.

[0054] These computer programs (also known as programs, software, software applications, “apps,” or code) contain machine instructions for programmable processors and can be implemented in high-level procedural and / or object-oriented programming languages ​​and / or assembly / machine languages. As used herein, the terms “machine-readable medium” and “computer-readable medium” refer to any computer program product, apparatus, and / or device (e.g., magnetic disks, optical disks, memory, programmable logic devices (PLDs)) used to provide machine instructions and / or data to a programmable processor, and include machine-readable mediums that receive machine instructions as machine-readable signals. However, “machine-readable medium” and “computer-readable medium” do not include transient signals. The term “machine-readable signal” refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0055] As used herein, a processor may include any programmable system that uses a microcontroller, a reduced instruction set circuit (RISC), an application-specific integrated circuit (ASIC), a logic circuit, and any other circuit or processor capable of performing the functions described herein. The above examples are merely illustrative and therefore do not in any way limit the definition and / or meaning of the term “processor”.

[0056] As used herein, the terms “software” and “firmware” are interchangeable and include any computer program stored in memory for execution by a processor, including RAM memory, ROM memory, EPROM memory, EEPROM memory, and non-volatile RAM (NVRAM) memory. The memory types described above are merely examples and are therefore not limited to the types of memory that can be used to store computer programs.

[0057] In one embodiment, a computer program is provided, and the program is implemented on a computer-readable medium. In one embodiment, the system runs on a single computer system without requiring connection to a server computer. In a further embodiment, the system runs in a Windows® environment (Windows is a registered trademark of Microsoft Corporation, Redmond, Washington). In yet another embodiment, the system runs on a mainframe environment and a UNIX® server environment (UNIX is a registered trademark of X / Open Company Limited, Reading, Berkshire, UK). The application is flexible and designed to run in a variety of different environments without compromising its core functionality.

[0058] In some embodiments, the system includes multiple components distributed across multiple computing devices. One or more components may be in the form of computer executable instructions embodied in a computer-readable medium. The system and processes are not limited to the specific embodiments described herein. Furthermore, each system and each process component can be implemented separately and independently of the other components and processes described herein. Each component and process can also be used in combination with other assembly packages and processes. These embodiments may enhance the functionality and capabilities of computers and / or computer systems.

[0059] The methods and algorithms of the present invention may be included in a controller or processor. Furthermore, the methods and algorithms of the present invention can be embodied as one or more computer implementations for carrying out one or more such computer implementations, and can also be embodied in the form of a tangible or non-temporary computer-readable storage medium containing a computer program or other machine-readable instructions (hereinafter referred to as "computer program"), and when the computer program is loaded into a computer or other processor (hereinafter referred to as "computer") and / or executed by the computer, the computer becomes a device for carrying out one or more methods. Storage mediums for containing such computer programs include, for example, floppy disks and diskettes, compact disk (CD)-ROMs (writable or not), DVD digital disks, RAM and ROM memory, computer hard drives and backup drives, external hard drives, "thumb" drives, and any other storage mediums readable by a computer. One or more methods may be embodied in the form of a computer program, whether stored in a storage medium, transmitted through a transmission medium such as an electrical conductor, optical fiber, or other optical conductor, or transmitted by electromagnetic radiation, and when a computer program is loaded into a computer and / or executed by the computer, the computer becomes a device for carrying out one or more methods. One or more methods may be carried out on a general-purpose microprocessor or on a digital processor specifically configured to carry out one or more processes. When a general-purpose microprocessor is used, the computer program code constitutes the circuitry of the microprocessor to create a particular logic circuit configuration. Computer-readable storage mediums include mediums readable by the computer itself or by another machine that reads computer instructions to provide the computer with those instructions to control its operation. Such a machine may include, for example, a machine for reading the above-mentioned storage mediums.

[0060] The compositions and methods described herein, utilizing molecular biology protocols, can follow various standard techniques known in the art (e.g., Sambrook and Russel (2006) Condensed Protocols from Molecular Cloning: A Laboratory Manual, Cold Spring Harbor Laboratory Press, ISBN-10: 0879697717; Ausubel et al. (2002) Short Protocols in Molecular Biology, 5th edition, Current Protocols, ISBN-10: 0471250929; Sambrook and Russel (2001) Molecular Cloning: A Laboratory Manual, 3rd edition, Cold Spring Harbor Laboratory Press, ISBN-10: 0879695773; Elhai, J. and Wolk, CP1988. Methods in Enzymology 167, pp. 747-754; Studier (2005) Protein Expr See Purif.41(1), pp. 207-234; Gellissen (ed.) (2005) Production of Recombinant Proteins: Novel Microbial and Eukaryotic Expression Systems, Wiley-VCH, ISBN-10:3527310363; Baneyx (2004) Protein Expression Technologies, Taylor & Francis, ISBN-10:0954523253).

[0061] The definitions and methods described herein are provided to better define this disclosure and to guide those skilled in the art in the practice of this disclosure. Unless otherwise specified, terms should be understood in accordance with the conventional usage of those skilled in the art.

[0062] In some embodiments, numbers representing quantities of components, properties such as molecular weight, reaction conditions, etc., used to describe and claim specific embodiments of this disclosure should be understood to be modified in some cases by the term “approximately.” In some embodiments, the term “approximately” is used to indicate that the value includes the mean standard deviation of the device or method used to determine the value. In some embodiments, the numerical parameters described in the specification and the appended claims are approximations that may vary depending on the desired properties to be obtained by a particular embodiment. In some embodiments, the numerical parameters should be interpreted by applying common rounding techniques in light of the reported number of significant figures. Although the numerical ranges and parameters representing a wide range of some embodiments of this disclosure are approximations, the numerical values ​​shown in specific examples are reported as accurately as possible. The numerical values ​​presented in some embodiments of this disclosure may include certain errors that inevitably arise from the standard deviation observed in each test measurement. The enumeration of value ranges in this specification is intended simply to serve as a convenient way to refer individually to each distinct value that falls within that range. Unless otherwise indicated herein, each individual value is incorporated herein as if it were individually enumerated. An enumeration of discrete values ​​is understood to include the range between each value.

[0063] In some embodiments, the terms “a,” “an,” and “the,” and similar references used in the context describing a particular embodiment (particularly in the specific context of the claims below), can be interpreted as encompassing both singular and plural unless otherwise specified. In some embodiments, the term “or” as used herein, including in the claims, is used to mean “and / or” unless it is explicitly stated that it refers only to substitutes or that the substitutes are mutually exclusive. The terms “comprise,” “have,” and “include” are open-ended linking verbs. Any form or tense of one or more of these verbs, such as “comprises,” “comprising,” “has,” “having,” “includes,” and “including,” are also open-ended. For example, any method that “comprises,” “has,” or “includes” one or more steps is not limited to having only one or more of those steps, but may also include other steps not listed. Similarly, any composition or device that “comprises,” “has,” or “includes” one or more features is not limited to having only one or more of those features, but may also include other features not listed.

[0064] All methods described herein may be carried out in any preferred order, unless otherwise indicated herein or unless it is clearly inconsistent with the context. The use of any examples or exemplary language (e.g., "etc.") provided in reference to specific embodiments herein is intended solely to better illustrate the disclosure and, unless otherwise claimed, does not limit the scope of the disclosure. No language herein should be construed as indicating any unclaimed element essential to the practice of the disclosure.

[0065] The grouping of alternative elements or embodiments of the present disclosure disclosed herein should not be construed as limitation. Each group member may be referenced and claimed individually or in any combination with other members of the group or other elements found herein. One or more members of a group may be included in or removed from a group for convenience or patentability reasons. In the event of such inclusion or removal, this specification shall be deemed to include the modified group and thus satisfy the description of all Markush groups used in the appended claims.

[0066] All publications, patents, patent applications, and other references cited herein are incorporated herein by reference in the same way that each individual publication, patent, patent application, or other reference is specifically and individually indicated so as to be incorporated by reference in whole for all purposes. No reference herein should be construed as an acknowledgment that it is prior art to this disclosure.

[0067] While this disclosure has been described in detail, it will be apparent that modifications, variations, and equivalent embodiments are possible without departing from the scope of this disclosure as defined in the attached claims. Furthermore, it should be understood that all examples in this disclosure are provided as non-limiting examples. [Examples]

[0068] The following non-limiting examples are provided to further illustrate the present disclosure. Those skilled in the art should understand that the techniques disclosed in the following embodiments represent approaches that the inventors have found to function well in implementing the present disclosure and can therefore be considered examples of modes for implementation. However, those skilled in the art should understand that many modifications can be made in light of the present disclosure without departing from the spirit and scope of the present disclosure, and similar or comparable results can still be obtained. Example 1: LiquidTME: Liquid biopsy for predicting immune checkpoint inhibitor (ICI) response

[0069] This example describes a fluid biopsy of the tumor microenvironment for early immunotherapy response evaluation. Immunotherapy has transformed modern cancer treatment and improved cancer survival rates. Immunotherapy “extracts” tumor immune cells (TILs) to improve cancer cell death. TILs in the tumor microenvironment (TME) play a crucial role in the response to therapy. Many patients do not respond to immunotherapy. There are five classes of leukocytes (white blood cells) that work together to provide defense against infection (e.g., neutrophils, eosinophils, basophils, monocytes, or lymphocytes). Several subsets may include naive and memory CD8 T cells and CD4 T cells, NK cells, naive and memory B cells, monocytes / macrophages, and granulocytes.

[0070] The following examples and disclosures provide solutions to the problem of early assessment of treatment response. Early imaging assessment is difficult and is confounded by factors such as pseudo-progression. Other major predictive measures, such as tumor PDL1, TMB, and oncogene expression profiling, lack sufficient sensitivity or specificity. Currently, there is no reliable method for predicting early immunotherapy response.

[0071] A solution to this problem, namely Liquid TME (LiquidTME), is disclosed. This solution involves measuring the level / activity of tumor immune cells themselves. Conventional, repeated invasive biopsies are impractical, and biopsies are subject to sampling bias. A liquid biopsy approach called LiquidTME is described to address this issue. Co-associated CpG methylation patterns

[0072] Adjacent CpGs have been shown to share similar methylation patterns due to the locally coordinated activity of methyltransferases, and CpGs function at the block level within promoters to regulate gene transcription. We utilized this concept in our ultra-sensitive method for internal error correction and confirmed the methylation state of CpG sites in a single sequenced DNA molecule by also examining its adjacent CpGs. Counting of co-associated CpG methylation patterns at the single-molecule level 1. Identify differentially methylated CpGs in purified reference cell types / states after methylation sequencing. 2. Assign sequencing reads from the bulk mixture to each reference cell type / state by tracking co-associated CpGs (based on the detection of cell type / state-specific co-associated CpGs at the individual read / molecular level). 3. Count the number of reads per cell type / state to determine their abundance in the bulk mixture. This is ultra-high-resolution digital cytometry at the single-molecule level, enabling the high performance of LiquidTME. background

[0073] Cancer is the second leading cause of death in the United States. 1 Immunotherapy is a powerful method for treating the advanced stages of disease. 2,3 However, only a small fraction of patients respond initially. 4 In many cases, the initial response is not sustained. 5 CT imaging is the standard treatment for evaluating immunotherapy response. 6,7 Early imaging evaluation is not reliable. 8,9 Currently, there is no reliable method to predict the early response to immunotherapy. Tumors release cellular and genetic material into circulation (see, for example, Figure 8). Previously, liquid biopsy was applied to ctilDNA, or cytoplasmic tumor cells (CTCs), but not to tumor-infiltrating leukocytes (TILs). "ctilDNA" is cell-free DNA derived from TILs. This platform, LiquidTME, profiles and measures ctilDNA. The results predict an early immunotherapy response.

[0074] Tumor-infiltrating leukocytes (TILs) in the tumor microenvironment (TME) determine the patient's response to immunotherapy. 10-22 , if enhanced, enables the killing of tumor cells 3,23Some groups have shown that early assessment of TILs by invasive biopsy in melanoma patients against immune checkpoint blockade is beneficial for treatment response. 16,20-22,24 TILs can be evaluated by invasive biopsy, but it is difficult and potentially dangerous to monitor TILs during treatment via repeated invasive biopsies. 25,26 Furthermore, unlike non-invasive liquid biopsies, invasive solid tumor biopsies are subject to sampling bias and can confound results. 27-30 There is no method available to measure the overall TIL content in the non-invasive liquid biopsy format.

[0075] The inventors hypothesized that liquid biopsy analysis of methylation signatures in plasma cell-free DNA enables accurate quantification of TILs and reliably predicts immunotherapy response. To support that TILs have a different epigenomic profile from normal white blood cells, Philip et al. showed that tumor-infiltrating CD8 T cells have a different chromatin profile compared to normal CD8 T cells. 31 Both myeloid and lymphoid TILs have also been shown to have a different gene expression profile from normal white blood cells by single-cell RNA sequencing. 32-34 The inventors' novel data also show that TILs have a different methylation profile compared to normal white blood cells and tumor cells, enabling their quantification via cell-free DNA liquid biopsy.

[0076] In addition to data support, the inventors have published their expertise in cell-free DNA analysis, the ability to detect solid tumor molecular residual disease, and detect ultra-low levels of circulating tumor DNA low enough to infer tumor mutation burden. 35-37 The inventors can also infer the relative abundance of individual cell states from bulk sequencing data 38 based on the most widely validated deconvolution model in the art. 39We have developed the deconvolution technology CIBERSORTx. Our experience in ultra-sensitive cell-free DNA analysis, state-of-the-art sequencing deconvolution, and translational research applying these technologies will facilitate the development of a novel liquid biopsy method called LiquidTME for non-invasive TIL analysis and improved prediction of immunotherapy response.

[0077] The inventors have developed LiquidTME for any cancer or disease state, and present it here for pretreatment of colorectal cancer and melanoma to non-invasively detect cellular status and predict response to various types of therapies, including immune checkpoint blockade. The inventors hypothesized that LiquidTME enables highly sensitive TIL quantification and predicts treatment response better than leading technologies. Furthermore, LiquidTME is in line with current efforts toward early cancer detection using cell-free DNA. 40 This will complement existing methods. Our research makes it possible for researchers to specifically evaluate TILs without requiring invasive tumor biopsy for the first time. Furthermore, the principle established here should be generalized to the etiology and type of therapy of almost all diseases, opening the door to routine non-invasive TIL assessment in research and clinical practice. data

[0078] Methylation profiles accurately distinguish TILs from PBLs and tumor cells. The inventors have used recent scRNA-seq and ATAC-seq data 32-34As suggested by [reference], we began by asking whether there are clear stereotypical epigenomic differences between tumor-infiltrating leukocytes (TILs) and normal peripheral blood leukocytes (PBLs). Therefore, we performed flow cytometry to isolate Epcam+ tumor cells, CD45+ TILs, and CD45+ PBLs from 10 patients with metastatic colorectal cancer (CRC). We performed whole-genome bisulfite sequencing (WGBS) followed by differential methylation region (DMR) analysis on each sample to identify 70 of the most differentially methylated CpG locations (Figure 1). This revealed that TILs have a different methylation profile compared to normal PBLs and tumor cells, suggesting that we can quantify TILs using methylation sequencing.

[0079] Thus, TILs were shown to have different methylation profiles based on methylation profiling of selected cells (e.g., see Figure 1) (whole-genome bisulfite sequencing (WGBS) on selected colorectal cancer samples), and differential methylation region (DMR) analysis revealed TIL-specific methylation patterns.

[0080] The TIL signature is detected in cell-free plasma DNA derived from CRC patients. Next, we asked whether TIL signals could be detected in cell-free DNA using a liquid biopsy technique that we call LiquidTME. To do this, we isolated plasma cell-free DNA (cfDNA) from 13 patients with metastatic CRC and performed WGBS in an Illumina NovaSeq S4 flow cell targeting 65 genome-wide coverage. We deconvolved this data by querying specific TIL vs. PBL vs. tumor cell signatures shown in Figure 1 using CIBERSORTx. Even at this low sequencing depth, using this approach (which we call LiquidTME), we were able to detect TIL signals from plasma in 9 of the 13 patients (Figure 2A). As an indication of specificity, four healthy donor plasma samples processed and analyzed in the same manner showed no evidence of TIL or tumor DNA signals. Therefore, plasma TIL and tumor DNA levels using our LiquidTME approach were significantly higher in CRC patients than in healthy donor controls (Figure 2B). Our data demonstrate the superior sensitivity and specificity of the method. Thus, LiquidTME in CRCs was shown to detect ctilDNA in CRC plasma (see, for example, Figure 2A), as evidenced by the detection of ctilDNA and tumor signals in cell-free plasma DNA from colorectal cancer patients, the absence of ctilDNA or tumor signals in cell-free plasma DNA from 12 healthy donor samples, and elevated ctilDNA levels in patients compared to healthy controls (see, for example, Figure 2B). TIL levels detected by LiquidTME in plasma cell-free DNA correlate with tumor ground truth.

[0081] Next, the inventors asked whether the level of TIL signaling detected by LiquidTME correlated with tumor ground truth. To answer this, the inventors correlated LiquidTME results from the nine detectable CRC patients described above with tumor ground truth. Surprisingly, TIL DNA levels in plasma cfDNA strongly and significantly correlated with tumor ground truth (Spearman ρ=0.71, Pearson r=0.70, P<0.05) (Figure 3). As an indicator of specificity, LiquidTME-derived TIL levels did not correlate with ground truth PBL or tumor cell fractions. Thus, it was shown that ctilDNA in plasma correlates with tumor ground truth (see, for example, Figure 3).

[0082] TIL signatures in plasma predict immunotherapy response in melanoma. Next, the inventors applied their LiquidTME assay in a pilot setting to melanoma patients treated with immune checkpoint blockade. To do this, the inventors analyzed banked pre-treatment and early in-treatment plasma samples from 12 patients with metastatic melanoma, from whom treatment samples were collected within one month of the initiation of immune checkpoint blockade. The response rate for this pilot cohort was 58%. When the above LiquidTME was applied to cfDNA extracted from each of these samples, an assay sensitivity of approximately 70% was achieved. Interestingly, when plasma TIL DNA was quantified as a percentage of total cfDNA, it was revealed that responders had higher plasma TIL DNA levels than non-responders (Figure 4A). In fact, ROC analysis showed remarkable results with an area under the curve (AUC) of 0.94 (Figure 4B), and the optimal plasma TIL DNA cut point was 12%. This cutpoint was applied to eight assay-detected patients, and Kaplan-Meier progression-free survival analysis (HR=9.3, P=0.03) nearly completely stratified long-term survivors from those with short-term progression (Figure 4C). Our data demonstrate that quantification of plasma TIL DNA in melanoma patients can accurately predict immunotherapy response.

[0083] Therefore, it has been shown that LiquidTME can also be applied to the melanoma immunotherapy response, as demonstrated by applying LiquidTME to melanoma plasma samples collected before or during early immunotherapy (see, for example, Figure 4). The levels of ctilDNA detected in 8 out of 13 samples (62%), or in tumor signals and cell-free DNA, were shown to correlate strongly with a persistent response among these 8 detectable patients.

[0084] Ultra-high resolution digital cytometry The inventors have developed a completely novel technique for ultra-high-resolution digital cytometry to achieve the sensitivity necessary for LiquidTME to function robustly. Specifically, the inventors track differentially methylated CpGs at the single-molecule level while utilizing the methylation state of adjacent CpGs ("co-associated CpGs") for internal error correction. The process of the present inventors' technology is as follows: 1. Identify differentially methylated CpGs in purified reference cell type / state methylation sequencing or microarray data (Figure 5). Figure 5 shows whole-genome bisulfite sequencing (WGBS) methylation data indicating differentially methylated CpGs in a selected leukocyte subset. 2. Methylation profiling of the bulk cell mixture (i.e., WGBS). After confirming that adjacent CpGs within the same read / read pair have the same methylation state ("co-associated CpG"), individual sequencing reads (or read pairs if paired-end sequencing is performed) from this bulk mixture are assigned to each of the reference cell types / states from step 1 by identifying differentially methylated CpGs at the single-molecule level (examined per read or per read pair). We examined this with different numbers of co-associated CpGs required (i.e., 2, 3, 4 per read pair) and demonstrated comparable high performance to ground truth flow cytometry regardless of parameter values ​​(Figure 6). Our method correlates well with flow cytometry ground truth across a set of co-associated CpGs (Figure 6). 3. After assigning individual DNA molecules (sequencing reads / read pairs) in the bulk mixture according to step 2, it is possible to quantify how the bulk mixture is composed of individual reference cell types / states (ultra-high-resolution digital cytometry). This is done either by counting the binned / assigned reads relative to each other (relative mode) or by normalizing the number of fragments assigned to the reference relative to the total number of unique reads with overlapping CpG positions (absolute mode). When the inventors applied this method to a bulk leukocyte mixture, they quantified the individual leukocyte subsets constituting these mixtures, and it correlated strongly with flow cytometry ground truth (Figure 7). The inventors' method correlates well with flow cytometry ground truth in both relative and absolute read count modes (Figure 7). The inventors' single-molecule read counting method is an ultra-high-resolution digital cytometry method for tracking cell types / states.

[0085] Overall, our ultra-high-resolution digital cytometry technology for quantifying and tracking cell types / states demonstrates high performance, ultra-high sensitivity, applicability to cell-free DNA, and enables non-invasive detection of rare cellular states, such as those arising from the tumor microenvironment, which are crucial for predicting immunotherapy responses by our LiquidTME method. Innovation: High-resolution digital cytometry at the single-molecule level

[0086] 1. Differential methylation-coasted CpGs are identified by DMR analysis of purified reference cell type / state. 2. Assign sequencing reads from the bulk mixture to each cell type / state (based on the detection of cell type / state-specific co-associated CpGs at the individual read level). 3. Count the number of reads per cell type / state to determine their relative abundance in the bulk mixture. Innovation: An alternative approach using methylation haplotype blocks 1. The chain disequilibrium principle ("methylation haplotype block") for identifying tightly bound CpG sites. 2. The epigenome is divided into approximately 150,000 methylated haplotype blocks (MHBs) of tightly bound CpG sites. 3. Reference profiles of purified cell types / states after sequencing, obtained by examining differential methylation MHB. 4. Assign sequencing reads from the bulk mixture to each cell type / state-specific MHB (based on the MHB identified at the individual read level). 5. Count the number of reads per cell type / state to determine their relative abundance in the bulk mixture. significance

[0087] This is the first method to profile TILs by liquid biopsy (see, for example, Figure 11). LiquidTME enables early immunotherapy response prediction, achieves serial profiling of TILs, and improves clinical decision-making and patient survival. summary

[0088] The described technology enables robust, ultra-high-resolution digital cytometry for measuring cellular status from methylation sequencing data. Given its ultrasensitivity, it can be applied to cell-free DNA, enabling non-invasive detection of rare cellular statuses, such as cellular status in the tumor microenvironment. This approach, called LiquidTME, serves as a robust early predictor of immunotherapy response in cancer patients through ultra-sensitive detection of tumor-infiltrating leukocytes. References [Table 1] TIFF0007874822000002.tif229162 TIFF0007874822000003.tif229162 TIFF0007874822000004.tif230162 TIFF0007874822000005.tif230162 TIFF0007874822000006.tif195160 Example 2: Fluid biopsy of tumor microenvironment for evaluation of immunotherapy response and toxicity

[0089] significance Cancer is the second leading cause of death in the United States. 3 Immunotherapy checkpoint inhibitors are now a powerful way to treat the advanced stages of disease. 4,5 Most advanced cancers alter their tumor microenvironment (TME) by activating cell surface receptors on immune cells, such as PD-1 and CTLA4, which inhibit the anti-tumor immune response. 6-8 Immune checkpoint inhibitors (ICIs) block these receptors, converting a subset of tumor-infiltrating leukocytes (TILs) in TMEs into cancer-killing cells, a phenomenon that has revolutionized the field of oncology. 4,5Unfortunately, however, most patients do not respond to immunotherapy, and as a result, the majority experience poor outcomes, largely due to the cellular composition of their TMEs. 6-8,10-19 This is because TME may also contain cells that promote resistance to immune checkpoint blockade, or it may lack cells that have cancer-killing properties. 4-8,10-21 Standard clinical practice does not monitor TME, making it difficult to reliably identify early on which patients will respond to immunotherapy. 22 While the tumor microenvironment directly underlies the treatment response, TME analysis requires invasive biopsy. 11 This is impractical to do continuously and could be dangerous for the patient. 23,24 To overcome this, the inventors have developed a liquid biopsy approach called LiquidTME, based on digital cytometry analysis of bisulfite-treated cell-free DNA (cfDNA) next-generation sequencing (NGS).

[0090] Development of LiquidTME A developed liquid biopsy approach called LiquidTME can distinguish TILs from tumor cells and normal leukocytes using methylation signatures (see, for example, Figure 14).

[0091] Digital cytometry of bisulfite-treated cfDNA was hypothesized to robustly detect TILs, tumor cells, and peripheral blood leukocytes. The inventors demonstrated that cell type abundances can be accurately deconvoluted from bulk tissue NGS data using CIBERSORTx. 20,25-28 Here, the inventors developed a similar approach to enable "digital cytometry" of bisulfite-treated cfDNA NGS data, identify and profile TILs, and distinguish them from tumor cells and normal peripheral blood leukocytes (PBLs).

[0092] Establishment of the technical performance of LiquidTME Here, we will explain how to establish the technical performance of LiquidTME and determine whether it can accurately capture TIL content from cfDNA obtained from melanoma patients (see, for example, Figure 14).

[0093] Digital cytometry of cfDNA bisulfite NGS was assumed to faithfully capture TIL content. Here, we applied our LiquidTME method to cfDNA isolated from melanoma patients and compared our predictions with the proportion of ground truth cells from deconvolution of tumor flow cytometry and bulk tumor genomic data at matched time points.

[0094] Application of LiquidTME to predict the ICI response in melanoma This section describes the application of LiquidTME to predict the ICI response in melanoma and compares it with other techniques.

[0095] Digital cytometry of cfDNA bisulfite NGS was hypothesized to enable ICI response prediction and allow for more accurate detection of molecular changes than other tumor / blood-based techniques and faster detection than standard imaging. The inventors applied their pre-treatment assay to patients with advanced melanoma treated with ICI to identify response signatures, validate these with a heldout test set, and compare them with clinical / imaging surveillance, peripheral blood TCR sequencing, tumor PDL1 percentage scores, and pre-treatment tumor genomic features.

[0096] background cell-free DNA Physiological cfDNA in the blood is thought to arise from cell death. 29-32 Malignant tumors also release DNA into circulation (ctDNA), which can then be isolated, quantified, and sequenced. 29-35 The mechanism of ctDNA release into the bloodstream is related to tumor cell death. 29-33 The challenge in ctDNA detection is that it is present at low levels in plasma and typically contains only a small number of normal cell-free DNA molecules. 32Therefore, modern NGS-based technologies have been developed that enable the detection of ctDNA at a low level of approximately 0.01% of total cell-free DNA, which is low enough to detect post-treatment molecular residual disease (MRD). 36,37 Just as 50 tumor cells secrete ctDNA, the inventors hypothesized that the tumor microenvironment also releases cell-free DNA that can be effectively measured using a highly sensitive method (Figure 8). The inventors refer to this new type of cell-free DNA as "circulating tumor-infiltrating leukocyte DNA," or "ctilDNA."

[0097] Immunotherapy response ICI is currently transforming cancer care and improving outcomes for a subset of patients with advanced cancer. 4,5,38 Nevertheless, the response to immunotherapy in individual patients is unpredictable, with the overall rate ranging from 1% to 60%, and most cancer types have a response rate of 5-20%. 39 What makes the problem more difficult is that standard therapeutic CT imaging cannot reliably distinguish between true and false progression at an earlier stage, making it impossible to reliably assess the response for approximately three months after the start of treatment. 40-42 This initial scan is still susceptible to false progress. 40-42 Current radiation guidelines recommend ordering a second scan at least one month later (approximately four months after the start of immunotherapy) to confirm if progression is suspected. 41-43 Despite these efforts, delayed false progression occurring after this initial period is still described. 41,42 Previous studies have shown that earlier response assessment can be performed by serial tumor biopsies analyzed by immunohistochemistry and genomics. 11,44,45 While this is a compelling approach, it is clinically impractical. Therefore, it is important to develop a liquid biopsy method for early evaluation of immune checkpoint inhibitor responses that can be easily applied sequentially, and this is the plan of the inventors hereby. Melanoma

[0098] Melanoma is the fifth most common cancer in the United States, a poster child for immunotherapy response, and boasts a high objective response rate of approximately 60% with combination ICIs. 46 Despite this, clinical outcomes remain poor, with a 4-year survival rate of only about 50%. 46 Cell-free DNA and ctDNA concentrations are typically elevated in patients with advanced disease, and several studies have demonstrated the ability of plasma fluid biopsy to assess this compartment. 9,47-52 Given the poor clinical outcomes, high cfDNA content, and the clear role of immunotherapy, it is worthwhile to focus on this cancer subtype for these studies. Bisulfite Sequencing

[0099] Bisulfite sequencing involves treating DNA with bisulfite to identify methylated bases, followed by NGS to identify patterns of DNA methylation. These methylation patterns can be used to identify the tissue of origin. 53,54 Recent publications demonstrate the usefulness of methylation profiling for detecting tumor cell-derived cfDNA. 55-57 However, the composition of TME has not been epigenetically profiled from cell-free DNA. The inventors plan to fill this gap using a novel approach.

[0100] data Molecular profiles distinguish TIL from PBL. Philip et al. demonstrated a distinct epigenetic program in tumor-specific CD8 T cells exhibiting cellular dysfunction using ATAC-seq. 58 Based on these results, we analyzed scRNA-seq data from T cells isolated from hepatocellular carcinoma patients (Zheng et al.). 59 ), we identified stereotypic differences between CD8 T cells of the same chronotype found in more than one tissue compartment (tumor, adjacent normal, and / or peripheral blood) (Figure 9). Markers associated with T cell depletion and dysfunction7 (i.e., ICOS, PD1, and CTLA4) as well as markers related to tumor responsiveness. 60 (i.e., CD103 and CD39) were consistently upregulated in tumor CD8 T cells, but were low or absent in the same chronotype in adjacent normal compartments and PBL compartments. This data suggests that TILs and PBLs can be distinguished using epigenetics. Mathematical modeling of ctilDNA detection using plasma cfDNA analysis Factors underlying the detection limits of cell-free DNA application include (1) the number of cell-free DNA molecules recovered, and (2) the number of independent "reporters" in the patient's tumor being queried. 1 Regarding these factors, previously described and validated binomial models can be used to predict the detection limit of circulating tumor DNA. 1 Using this, the inventors have found that (1) a realistic amount of cell-free DNA input (approximately 32 ng of cell-free DNA in one blood collection tube) 1 ), (2) median of circulating tumor DNA fraction in metastatic melanoma (approximately 1%) 61 ), (3) Estimated TIL content in progressive melanoma tumors 20 (4) Estimated cell-free DNA recovery rate after bisulfite conversion (20-60%) 62 ), and (5) Published recovery rates of cell-free DNA using hybrid capture sequencing (40-60%) 1 Considering the following, we estimated the number of unique cell-type-specific differential methylation regions (DMRs; i.e., "reporters") that would be required to achieve various detection limits. Assuming approximately 10,000 genomic equivalents of cell-free DNA (approximately 32 ng of cell-free DNA), and assuming 80% DNA loss from library preparations, this model suggests that more than 10 DMRs per cell type would be sufficient for TIL detection with 95% confidence (Figure 10). Our model suggests a high probability of success, as the detection limits required to track ctilDNA are within the range that can be reliably achieved with ctDNA. 32,36,37 .

[0101] The TME signature can be detected in cfDNA. Next, we inquired whether tumor microenvironment signals could be detected in cell-free DNA using liquid biopsy techniques. To do this, we FACS-selected CD45+ TILs and EPCAM+ tumor cells from three cryopreserved colorectal cancer (CRC) tumor samples and their corresponding PBLs, and performed whole-genome bisulfite sequencing. We then performed differential methylation region analysis using methylene 63 Using this method, we identified distinct DMRs between each population used as reporters for deconvolution. Next, we performed whole-genome bisulfite sequencing (WGBS) of cell-free DNA from these patients using an Illumina NovaSeq S4 flow cell targeting 4050 genome-wide coverage, and queried these reporters using deconvolution by non-negative least-squares regression. Surprisingly, even at this low sequencing depth, we were able to detect TIL signals from plasma in 2 out of 3 patients using this approach (Figure 12). We also detected tumor signals in all 3 patients. Demonstrating the specificity of our method, TIL signals were not detected in the PBL compartment. Furthermore, as we expected, PBL signals were lower in tumors than in the periphery. Only TIL and tumor signals in cell-free DNA showed positive correlations with flow cytometry and imaging in matched tumors. The inventors can optimize their assay, determine whether multiple TIL subsets can be quantified in plasma, and significantly expand this study to demonstrate its clinical utility.

[0102] Application of LiquidTME to melanoma Next, the inventors applied their assay to melanoma in a pilot setting. To do this, they analyzed banked pre-treatment plasma samples from 12 patients with advanced melanoma, sampled within one month of the initiation of immune checkpoint blockade. The response rate for this pilot cohort was 58%. The above version of LiquidTME was then applied to each of these samples, detecting ctilDNA in 6 samples (50%) and below the detection limit of the assay in the remaining 6. Interestingly, the inventors did not report any lasting clinical benefit (DCB). 64,65 Three patients with detectable ctilDNA who achieved a lasting benefit had significantly elevated ctilDNA levels compared to patients who did not achieve a lasting benefit (NDB) (P=0.02) (Figure 13), with a ctilDNA cutoff of 12% that fully classified patients by lasting response status (Figures 13 and 4B). Kaplan-Meier analysis based on this optimal cutoff point fully stratified long-term survivors from short-term progression patients (HR=15.3, P=0.02) (Figure 4C). Our data support our ability to predict melanoma response to immunotherapy using LiquidTME applied at an early stage.

[0103] Experimental Design and Methods Development of a liquid biopsy platform that uses methylation signatures to distinguish TME cells from tumor cells and normal leukocytes.

[0104] Definition and validation of digital cytometric signatures for TILs, tumor cells, and PBLs. The inventors analyzed banked, viable, preserved tumor and PBMC samples from 10 patients with advanced melanoma and isolated TILs, tumor cells, and PBLs by FACS. Nine major leukocyte subsets were profiled from the tumor and PBL samples: naive and memory CD8 T cells and CD4 T cells, NK cells, naive and memory B cells, monocytes / macrophages, and granulocytes. The inventors also isolated MAGE1+ tumor cells. The inventors extracted at least 10 ng of genomic DNA from each of these samples (approximately 1.5k cells / sample) containing the corresponding bulk tumors and PBLs and performed WGBS. To do this, the inventors utilized the Zymo EZ DNA Methylation-Lightning kit for bisulfite conversion, the Swift Biosciences Accel-NGS Methyl-Seq DNA kit for library preparation, and the Illumina NovaSeq for 4050 coverage WGBS. The inventors analyzed these data and found that methylene 63 We use DMR and random forests, glmnet, and / or previous optimization schemes for feature selection to identify specific signatures for each cell type. 27,28 The inventors identify these signatures. They evaluate the discriminative power of these signatures by applying them to bulk tumor and PBL methylation profiles from 10 additional patients with ground truth percentages determined by flow cytometry and bulk tissue RNA-seq deconvolution. 33 These analyses are used to establish a minimum set of approximately 1,500 DMRs that can distinguish melanoma tumor cells, distinct TME subsets, and PBL subsets.

[0105] Design of DNA capture panels for targeted melanoma TME bisulfite sequencing

[0106] The inventors design a capture panel that targets all DMRs identified above in order to maximize analytical sensitivity and improve error tolerance. 1,66 Other regions are added according to their clinical or biological relevance (e.g., ICI co-inhibitor receptors) until a final size of approximately 2,000 genomic segments is achieved (each approximately 100 bp). The inventors utilize commercially available and published approaches for panel design (e.g., molecular inversion probes). 55 Evaluate both of the following:

[0107] The inventors (1) define TILs, PBLs, and melanoma-specific methylation signatures for deconvolution purposes, and (2) design an optimized sequencing panel with genomic bandwidth for profiling melanoma tumor cells, TILs, and PBLs with high analytical sensitivity.

[0108] Where higher sensitivity is desired to distinguish between distinct TILs, PBLs, and tumor populations, we may perform a deeper WGBS (approximately 65) to reduce coverage dropout rates, expand the capture panel to include more genomic regions, profile additional patients, and / or pool cell types into broader phenotypic classes.

[0109] Establishing the technical performance of LiquidTME and determining whether it can accurately capture TIL content from cfDNA obtained from melanoma patients.

[0110] Evaluation of the technical performance of assays using defined in vitro mixtures. To evaluate the accuracy and lower limit of detection of our method, we prepare a series of defined mixtures in which sonicated DNA from tumor cells, TILs, and PBL subsets (either remaining from those obtained above or selected from further patients) is added in vitro to Horizon synthetic plasma. The simulated TME content in the plasma will be in the range of 5% to less than 0.1% to mimic the TIL content in melanoma tumors adjusted for clinically realistic ctDNA levels. 8,9,11,17,19,44,47-52,67 This panel is used to apply targeted bisulfite sequencing to 10, 20, 30, and 50 ng DNA mixtures, and to evaluate the levels of each TME component using digital cytometry. These analyses allow for establishing performance expectations and tuning the method for maximum sensitivity and specificity.

[0111] Perform TME profiling on cfDNA and bulk PBMCs, and evaluate their consistency with paired tumors.

[0112] The inventors analyze cryopreserved tumor, PBL, and plasma samples taken from 30 patients with melanoma. Patients underwent tumor biopsies, and blood was collected before treatment. A subset of patients with recurrent specimens is also evaluated, allowing for assessment of changes in TME content from baseline. In parallel, the inventors process banked blood samples (plasma and bulk PBL) from 10 age-matched healthy controls. The inventors isolate cfDNA from plasma samples and genomic DNA from tumors and PBLs. The inventors compare cell abundance estimates from their platform with flow cytometry to (1) evaluate the accuracy and precision of the method and (2) determine whether cfDNA or PBL DNA better captures TIL content.

[0113] The inventors (1) profile TIL subsets from genomic DNA and cfDNA, (2) accurately quantify TILs in cfDNA and distinguish them from normal PBLs, (3) extend the inventors' analysis shown in Figure 12 to demonstrate the superiority of cfDNA over PBLs for capturing TME content, and (4) demonstrate high specificity by comparison with healthy cfDNA and PBLs.

[0114] Because it shows high ctDNA levels in melanomas where research has progressed. 9,47-52 Unexpectedly, if the amount of cell-free DNA may be too low to distinguish different TIL subsets, the inventors can improve detection by increasing the input cfDNA amount and sequencing depth and refining the signature. Separately, tumor dissociation can distort flow cytometry. 28 The inventors compare the ctilDNA profile with tumor RNA-seq deconvolution. Application of LiquidTME to predict melanoma ICI response compared to other technologies

[0115] The inventors banked serial blood samples from over 100 patients with advanced melanoma treated as first-line therapy with ICI. In parallel, patients underwent standard treatment CT imaging and were followed for at least one year to determine response rate versus progression rate. Approximately half of these patients achieved lasting clinical benefit, while the rest developed progressive disease. The inventors utilized pre-treatment plasma samples from 50 patients (randomly selected) and evaluated pre-treatment ctilDNA to identify features corresponding to lasting clinical benefit (i.e., increased ctilDNA content). Next, the inventors analyzed the remaining 50 patients from the bank to validate the response profiles learned from the inventors' test set. The inventors evaluated ROC AUC and CT imaging scored by LiquidTME, PDL1 tumor proportion score, peripheral blood TCR sequencing, pre-treatment tumor NGS profile (i.e., "hot" RNA signature vs. "cold" RNA signature, tumor mutagenesis), and RECIST1.1. 68 We compare these factors. To associate these factors with progression-free survival and overall survival, we perform Cox regression. Statistical considerations

[0116] To determine the required sample size for the inventors' training and validation cohorts, the inventors assumed that patients had a 50% response rate. Based on conservative predictions from the inventors' data, the inventors assumed a 25% higher 1-year response rate for patients with a TIL response signature and a 25% lower response rate for patients with a TIL non-responder signature. To achieve 90% power (α=0.05, two-sided) to reject the null hypothesis that there is no difference in PFS between the two groups, data from at least 38 patients needs to be analyzed. To account for attrition, approximately 30% more will be analyzed per cohort.

[0117] The inventors (1) determine the TIL profile from pre-treatment cfDNA (i.e., elevated ctilDNA content as shown in Figure 2B) corresponding to the lasting clinical benefit of ICI treatment, (2) validate it using a held-out cohort, and (3) demonstrate more accurate response and outcome prediction than other techniques tested. If the sensitivity remains below the optimal level, bioinformatics background error correction is performed. 1 Methods can be implemented to improve the analytical limits of detection, such as adding DMRs / reporters to the capture panel, increasing sequencing depth, and optimizing deconvolution using machine learning. The inventors also analyze early intra-treatment samples (approximately 4 weeks of treatment) to enhance the clinical sensitivity / specificity of their approach, as early intra-treatment evaluation remains valuable even if pre-treatment evaluation is difficult. innovation This technique is a highly innovative combination of cfDNA bisulfite sequencing and digital cytometry for the first time to profile TME in solid tumor cancer patients by liquid biopsy. This approach will help address a major unmet need: early prediction of ICI response. References [Table 2] TIFF0007874822000008.tif224160 TIFF0007874822000009.tif230161 TIFF0007874822000010.tif228162 TIFF0007874822000011.tif228162 TIFF0007874822000012.tif228162 TIFF0007874822000013.tif133162 Example 3: Development of a liquid biopsy technique for tumor microenvironment profiling

[0118] problem Immune checkpoint inhibitors have transformed modern cancer treatment, becoming the sole treatment for many years due to their ability to provide sustained remission and significant survival improvements in many cancer types. Despite their success, most patients do not respond to these drugs, and there is a significant risk of immuno-related toxicity, which inventors cannot reliably predict early on. The key to maximizing the potential of immune checkpoint inhibitors lies in understanding the tumor microenvironment (TME). However, the only way to analyze the TME is through invasive biopsies, which are impractical to perform sequentially and can be harmful to patients.

[0119] solution In this specification, the inventors disclose the development and testing of a liquid biopsy method for tumor microenvironment profiling based on next-generation methylation sequencing of cell-free DNA. This method, which the inventors call liquid TME, is developed in the context of colorectal cancer and lung cancer (two of the most common cancers worldwide), but can be directly extended to most malignancies. If successful, the inventors' method will enable tumor microenvironment analysis through a simple blood test, which should have a direct clinical impact by allowing for faster and more accurate assessment of thousands of cancer patients being treated with immunotherapy.

[0120] Cancer is the second most common cause of death in the United States. 1 Immune checkpoint inhibitors are now a powerful way to treat advanced disease stages. 2、3 Cancers in the most advanced stages alter their tumor microenvironment (TME) by activating cell surface receptors on immune cells, such as PD-1 and CTLA4, which inhibit the anti-tumor immune response. 4~6Immune checkpoint inhibitors block these receptors, transforming a subset of tumor-infiltrating leukocytes (TILs) in TMEs—a phenomenon that has revolutionized the field of oncology—into cancer-killing cells. 2、3 .

[0121] Unfortunately, however, most patients do not respond to immunotherapy, and as a result, the majority experience a poor prognosis due to the cellular composition of their TME. 4~16 This is because TME may include cells that promote resistance to immune checkpoint blockade, or it may lack cells that possess cancer-killing properties. 2~18 In standard clinical practice, the inventors do not monitor TME, and therefore cannot reliably identify early on which patients will respond to immunotherapy. 19 There is also a serious risk of immune-related side effects. 20 , literature 21、22 Cases of fatality have been reported. The tumor microenvironment may directly receive the therapeutic response and similarly play an important role in toxicity. 23 TME analysis requires an invasive biopsy. 7 This is impractical to perform continuously and could be dangerous to the inventors' patients. 24、25 In this specification, the inventors describe a non-invasive liquid biopsy technique called liquid TME to overcome this problem.

[0122] Our method for TME liquid biopsy allows tumors to continuously release DNA in circulation, which can be isolated as cell-free circulating tumor DNA (ctDNA). 26~30 This fact is utilized. The mechanism of ctDNA release into the bloodstream is related to tumor cell death. 26~30 A challenge with ctDNA detection is that plasma levels are low, typically containing less than 1% of normal cell-free DNA molecules. 26 Modern NGS-based technologies have been developed that enable the detection of ctDNA at levels as low as approximately 0.01% of total cell-free DNA, which is low enough to detect molecular residual disease (MRD) after treatment. 31、32The inventors hypothesized that tumor cells not only secrete ctDNA but also release cell-free DNA that can be effectively measured using sensitive methods by the tumor microenvironment (Figure 8). The inventors refer to this new type of cell-free DNA as "circulating tumor infiltrating leukocyte DNA" or "ctilDNA".

[0123] Disclosed herein is an ultrasensitive method for detecting ctilDNA by tracking highly specific epigenomic markers on DNA rather than tumor mutations. The epigenome is composed of compounds bound to DNA molecules that instruct which parts of the genome are turned on or off. 33 Each cell type has a unique epigenomic signature that can be profiled by analyzing methylation patterns on DNA using a method called bisulfite sequencing. 34、35 The inventors used these epigenomic signatures to distinguish cell types by machine learning-based cell deconvolution that is conceptually similar to CIBERSORT but applicable to very low levels of ctilDNA present in plasma. To assist with this, the inventors performed a mathematical modeling exercise using this method (Figure 10). The inventors' model suggests a high likelihood of success as the detection limits required to track ctilDNA are within the limits that the inventors have reliably achieved with ctDNA. 33 The inventors used these epigenomic signatures to distinguish cell types by machine learning-based cell deconvolution that is conceptually similar to CIBERSORT but applicable to very low levels of ctilDNA present in plasma. 36、37 To assist with this, the inventors performed a mathematical modeling exercise using this method (Figure 10). The inventors' model suggests a high likelihood of success as the detection limits required to track ctilDNA are within the limits that the inventors have reliably achieved with ctDNA. 26、31、32 .

[0124] Importantly, tumor infiltrating leukocytes (TILs) are different from their normal peripheral blood leukocyte (PBL) counterparts as shown by recent single cell RNA sequencing studies of lung and breast tumors. 11、41、42 Since this difference has been demonstrated to be also seen in the epigenome, Philips et al. utilized ATAC-seq to demonstrate a distinct epigenomic program in tumor-specific CD8 T cells showing cell dysfunction. 43. To significantly expand on these results, we re-analyzed published single cell RNA sequencing (scRNA-seq) data (Zheng et al 44 ) from T cells isolated from hepatocellular carcinoma patients to clearly observe the consistent differences between tumor infiltrating CD8 T cells and their normal counterparts (from both adjacent normal tissue and PBL) (Figure 10). We also performed this analysis at the clonal type level and strikingly, CD8 T cells with the same T cell receptor (derived from the same precursor) still showed significant epigenetic differences between tumor and normal, and the final tissue / blood site of tumor vs normal was shown to be a major determinant of their expression signature, regardless of their clonal genomic identity. Markers 5 associated with T cell exhaustion and dysfunction (i.e., ICOS, PD-1 and CTLA4), as well as markers 45 associated with tumor reactivity (i.e., CD103 and CD39) were consistently upregulated in tumor CD8 T cells but low or absent within the same clonal types from other compartments. This data is compelling and suggests that these differences between TILs and normal PBLs can be exploited to identify TIL signatures from cell-free DNA, even if the majority of cell-free DNA originated from normal PBLs.

[0125] This technology is based on the premise that ultrasensitive detection and profiling of ctilDNA from the TME enables early and accurate cancer treatment response and toxicity assessment. Our approach uses innovative technical methods for highly sensitive and specific detection of individual TME cell subsets (i.e., CD8 T cells, CD4 T cells, NK cells, B cells, monocytes / macrophages, cancer associated fibroblasts) from cell-free DNA, data from methylation sequencing studies (e.g., ENCODE 46 , BLUEPRINT 47 , NIH Roadmap Epigenomics Project 33The inventors utilize machine learning to combine this with their own data, which they generate through methylation sequencing of patient samples. This technique is a non-invasive TME profiling assay that the inventors apply to cancers such as lung and colorectal cancer and should be readily expandable to all common cancer types. The potential impact of the inventors' research is enormous, and if successful, the inventors' assay could become a routine laboratory test directed for thousands of patients annually. Continuous ctilDNA monitoring ultimately provides clinicians with a real-time window into the workings within the tumor microenvironment, thus enabling them to switch treatments (i.e., redirect to alternative therapies early if the patient is unlikely to respond or is likely to experience severe toxicity).

[0126] Clinical relevance The inventors wish to re-emphasize the potential clinical significance of this research. Immune checkpoint inhibitors have transformed cancer care and improved the prognosis of a large number of patients with advanced cancer. 2、3 In the inventor's field of practice (lung cancer), immunotherapy has dramatically improved survival rates in patients with both locally advanced and advanced disease. 49~52 Many patients have been able to live longer than previously thought. Furthermore, the response to immunotherapy in individual patients is unpredictable, with the overall rate ranging from 1% to 50%, and most cancer types have a response rate of 5-20%. 53 What makes the problem even more difficult is that standard therapeutic CT imaging cannot distinguish between true progression and pseudo-progression in the early stages, and it is not possible to reliably assess the response for about three months after the start of treatment. 54~56 This first scan may still be subject to pseudo-progression. 54~56 Current X-ray guidelines recommend that in cases of suspected progression, a second scan should be ordered at least one month later (approximately four months after the start of immunotherapy) to provide confirmation. 55~57 Despite these efforts, delays in pseudoprogression occurring after this initial period are still described.55、56 Recent research has shown that immunohistochemistry and genomics are compelling but clinically impractical methods. 7、58、59 The analysis of serial tumor biopsies demonstrated that early response assessments could be performed. Therefore, it is important to develop a liquid biopsy method for early assessment of immune checkpoint inhibitor responses that can be easily applied sequentially, as currently disclosed herein. Considering the broad importance of the tumor microenvironment, the technology developed by the inventors is also applicable to other clinical and research situations.

[0127] Toxicity can be a factor behind immunotherapy responses. 20 The proportion of severe toxicity requiring hospitalization was approximately 60% in patients treated with combination immune checkpoint inhibitors (anti-CTLA4 and anti-PD1) and approximately 25% in patients treated with monotherapy. 60、61 Unfortunately, numerous deaths attributable to immune checkpoint blockade have also been documented. 21、22 In a large meta-analysis of 613 patients who experienced fatal immune checkpoint blockade-related toxicity, the median time to death after initiation of treatment was only 14.5 days in patients treated with a combination of immune checkpoint inhibitors, compared to 40 days in patients treated with either anti-PD1 or anti-CTLA4 monotherapy. 21 It was emphasized that biomarkers must be developed to predict these as early as possible. Higher toxicity rates are associated with certain mechanisms of action (i.e., anti-CTLA4 versus anti-PD1). 60、61 The precise pathophysiology underlying these serious immune-related adverse events remains unknown, and parallel studies suggest that multiple immune pathways may be involved. 20 It has been suggested to some extent that B cells play an important role in toxin production. 62 A recent report in Nature Medicine links oligoclonal proliferation of CD4 T cells targeting EBV-specific and EBV-like domains to cases of fatal encephalitis. 22Using liquid TME, the inventors were able to profile cell-free DNA from TILs and circulating leukocytes in a single assay that allows tracking a diverse repertoire of immune cell dynamics before and during treatment. Thus, the inventors hypothesized that, based on the results of their tests, they would gain new insights into the biology of toxicity, meaning that clinicians would consider alternative treatments for patients who appear to be at high risk of toxicity. Therefore, using the inventors' method, it was possible to identify and track immuno-related toxicity from immunotherapy and, as well as potentially other forms of treatment.

[0128] Disclosed herein is a novel method for detecting tumor microenvironment-derived DNA in cell-free DNA, called liquid TME. Liquid TME requires the purification of pre-determined genomic regions highly enriched with DMR that identify tumor microenvironment cell subsets and distinguish them from their normal counterparts. Liquid TME is ultrasensitive and directly applicable to cancer patients, with its most direct clinical role being the early prediction of immunotherapy response and toxicity. In describing the experimental design for developing Liquid TME, the inventors first describe in detail the technical development of the method, and then describe experiments to evaluate its clinical utility. Thus, this technology can deliver an optimized method for non-invasively profiling the tumor microenvironment, having passed initial clinical validation applicable to immunotherapy patients. Hereinafter, Liquid TME is developed in the context of CRC and NSCLC.

[0129] The inventors have chosen to focus on colorectal cancer (CRC) and non-small cell lung cancer (NSCLC) because they are the most common causes of cancer and cancer death worldwide. Furthermore, the inventors are radiation oncologists specializing in the treatment of lung and gastrointestinal cancers, thus possessing clinical expertise in this field and easy access to specimens. The inventors believe their liquid TME test is applicable to other cancer types, likely requiring only minor optimizations. Focusing on NSCLC and CRC allows for the development and testing of this method in the initially defined clinical context, as well as in the inventors' optimal clinical expertise and access to specimens.

[0130] Proof-of-Concept Experiment of Liquid TME The inventors began with mathematical modeling experiments in Figure 10 and TIL vs. normal scRNA-seq analysis in Figure 9. Next, they addressed the practical question of developing a method: Does freezing affect the cellular epigenetic methylation profile? To answer this, the inventors performed whole-genome bisulfite sequencing (WGBS) on nine healthy peripheral blood leukocyte samples from healthy donors. All sample preparations were performed with fresh (unfrozen) DNA for three samples, frozen DNA for three samples, and cryopreserved cells for the remaining three samples before any further processing. Following WGBS on all nine samples, the inventors observed no significant differences in global methylation patterns (Figure 15), and prior literature... 64 This suggests that cold-banked cells or DNA do not introduce epigenetic artifacts.

[0131] The inventors generated proof-of-concept data showing that methylation signatures differ between individual TIL subsets and their normal counterparts. To do this, the inventors isolated CD8 T cell subsets selected from tumors of three cryopreserved CRC patients and peripheral blood CD8 T cells from these same patients, followed by whole-genome bisulfite sequencing, sequence alignment, and methylation analysis. 65Differential methylation region analysis was performed using BLUEPRINT, and the methylation levels in these samples were compared with them. 47 We compared these CD8 T cells with those from healthy donor CD8 T cells publicly available through the project. Reduced methylation levels were observed in genes associated with T cell exhaustion / dysfunction, including ICOS, PDCD1, and CTLA4, in CD8 TILs (supporting the inventors' scRNA-seq in Figure 9). This is shown for the PDCD1 locus in Figure 16.

[0132] Next, we investigated whether tumor microenvironment signals could be detected in cell-free DNA using liquid biopsy techniques. To do this, we performed whole-genome bisulfite sequencing of CD45+ TIL and EPCAM+ tumor cells from three cryopreserved CRC tumor samples and their corresponding peripheral blood leukocytes, starting with FACS sorting. Metilene was used for differential methylation region analysis. 65 Using this method, we identified distinct DMRs between each population and then examined them in cell-free DNA using deconvolution by non-negative least squares regression. We performed whole-genome bisulfite sequencing of cell-free DNA using Illumina NovaSeq S4 flow cells targeting a 4050 genome-wide coverage, and notably, we were able to detect TIL signals in the plasma of 2 out of 3 patients even at this low sequencing depth (Figure 12). We also detected tumor signals in all 3 patients. As indicated by the specificity of our method, this TIL signal was not detected in the peripheral blood cell compartment. Furthermore, as we expected, peripheral blood cell signals were lower in the tumor than in the surrounding tissue. Only TIL and tumor signals in cell-free DNA positively correlated with flow cytometry and imaging in the fitted tumors. The inventors plan to optimize their assay, determine whether multiple TME subsets can be quantified in plasma, and significantly expand upon this initial effort to demonstrate its clinical utility in the context of immunotherapy.

[0133] To develop a liquid TME for non-invasive TME profiling, the inventors followed the plan outlined in Figure 17. A robustly functioning liquid TME requires separate input signatures derived from the inventors' cell types of interest. Therefore, we began by FACS purification of viable-conserved tumor and peripheral blood leukocyte samples from 10 patients with progressive CRC or NSCLC, and isolated major leukocyte subsets including naive and memory CD8 T cells and CD4 T cells, NK cells and NK T cells, naive and memory B cells, myeloid-derived suppressor cells (MDSCs), monocytes / macrophages, and granulocytes. The inventors also examined the immunosuppressive tumor microenvironment. 9、10 Cancer-associated fibroblasts (CAFs) and EPCAM+ tumor cells, which have been reported to promote bisulfite synthesis, are isolated. The inventors extract at least 10 ng of genomic DNA from each of these samples (approximately 1.5k cells / sample), including the corresponding bulk tumor and plasma-free whole blood. To prepare samples for bisulfite sequencing, the inventors utilize the Zymo EZ DNA Methylation-Lightning Kit for bisulfite conversion, followed by the Swift Biosciences Accel-NGS Methyl-Seq DNA Kit for library preparation, and then sequence their samples using Illumina NovaSeq with an S4 flow cell targeting a 4050 genome coverage. Sequence alignment and BISCUIT are performed. 66 After determining methylation sites using a software suite and performing quality control using laboratory scripts, differential methylation region (DMR) analysis is performed using Metilene. 65 The present inventors apply this. In this way, they identify specific methylation signatures corresponding to each cell type that enable differentiation between each TME subset and differentiation from normal peripheral blood leukocytes.

[0134] By leveraging machine learning feature selection techniques, including random force and elastic net, we identify DMRs that are most likely to enable clear distinction between cell types (Figure 17). These distinguished DMRs are incorporated into a sequencing panel (e.g., utilizing molecular inversion probes) that can distinguish tumor cells, TME subsets, and PBL subsets, while achieving a sequencing depth far greater than that of WGBS (typically ≤40×) (aiming for a 2,000× deduplication-excluded depth as shown in Figure 10). This 2,000× depth is achieved by a targeted hybrid capture-based ctDNA detection method. 26、38、67、68 This is a typical example, and since the sequencing space is limited to a small portion of the genome, the cost is not exorbitant. 26、29、30 Therefore, setting it as a goal is reasonable.

[0135] Next, the inventors optimize their method and validate it in plasma (Figure 17). To do this, the inventors use a predefined DNA mixture derived from a TME subset and peripheral blood cells (sheared to simulate the size of cell-free DNA). 69、70Liquid TME is applied to the following mixtures. To simulate ctilDNA in plasma, these mixtures contain TME content ranging from 4% to 0.04% to emulate clinically realistic levels within a range of 10 times the inventors' estimates in Figure 10. A more sophisticated machine learning-based deconvolution strategy is considered to estimate the relative percentage of each cell type in the inventors' simulated TME mixtures. The inventors expect the ctilDNA signal to be low and mixed with normal leukocyte DNA with a high background, but anticipate a high probability of success. Furthermore, Figure 12 shows that the TME signal in cell-free DNA could be detected without the major technological innovations described herein. Liquid TME can be clinically validated by applying the assay to plasma samples from patients with advanced-stage CRC and NSCLC (Figure 12). The inventors also have cryopreserved and registered tumor samples from these patients at the same point in time. The inventors evaluate the accuracy and precision of their method in these clinical samples compared to flow cytometry in dissociated tumors. In addition to flow cytometry, tissue dissociation can lead to the loss of various fragile cell types, which can distort the results of flow cytometry by supporting cell types that fit through the pores of filters and instruments. 37 , CIBERSORT in the tumor sample 36、37 The researchers will also conduct the following: To clinically validate the performance, they will compare the agreement between their method applied to cell-free DNA and gold-standard tumor analysis.

[0136] Before proceeding to the clinical implementation evaluation of liquid TME, the inventors investigate several physical properties of ctilDNA. CtilDNA has not been investigated and is defined herein for the first time. Having established the inventors' method, the inventors take this opportunity to analyze the biophysical properties that may make ctilDNA unique compared to its ctDNA and normal cell-free DNA counterpart. First, the inventors investigate whether ctilDNA has a unique size distribution, as has been observed for ctDNA. 69、70The unique size distribution allows for prior enrichment of ctilDNA using bead-based cell-free DNA size selection, as the group is currently doing for ctDNA. Secondly, the inventors investigate whether ctilDNA is enriched in exosomes. Exosomes are tiny vesicles present in plasma that can contain nucleic acids. 72 To test whether cell-free DNA derived from TME is enriched inside or outside the exosome, the inventors have used the previously described method. 73 Plasma fractionation was performed using [a specific method], and the exosome-rich fraction and the exosome-removed fraction were arranged side by side. Finally, the inventors confirmed that their understanding of ctilDNA and the data shown in Figure 18 suggest significant enrichment of ctilDNA in a cell-free state relative to the cellular compartments of blood, and compared their results with the gold standard tumor assessment. These studies have enhanced the inventors' biophysical understanding of ctilDNA and provided a method for enriching it.

[0137] To establish the clinical utility of liquid TME, the inventors will test it in a cohort of patients treated with immune checkpoint blockade, for whom they have response and toxicity data (Figure 12). Since last year, the inventors have been collecting samples from CRC and NSCLC patients treated at the University of Washington, and as a result of their efforts, the inventors' laboratory has a 700-box -80°C freezer. Samples are processed immediately after collection using a standardized protocol and divided into aliquots for cryopreservation.

[0138] To test the efficacy of liquid TME, it was applied to patients with advanced-stage NSCLC and CRC being treated with immunotherapy (Figure 12). Immune checkpoint blockade-based approaches have become standard care for patients with advanced-stage NSCLC and MSI-induced high CRC, with response rates of approximately 30–50% overall. 51、52、74、75 Unfortunately, some patients are unresponsive. 53For most cases, it can take several months to confirm the lack of response with CT imaging (approximately 3 months for the initial scan and approximately 1 month for the confirmation scan). 19、54、55、57、61 To begin addressing this issue, the inventors applied liquid TME to approximately 50 patients treated with immune checkpoint blockade. The inventors applied their assay before treatment and again 2-3 weeks after the start of treatment (two blood draws during the chemotherapy cycle). The results of the inventors' assay were correlated with the final clinical response to the treatment. The inventors expected to demonstrate increases in CD8 T, NK, and NK T TILs, as well as decreases in immunosuppressive macrophages, MDSCs, and CAFs in responders and patients. The inventors used pre-treated biopsy samples (and biopsies taken during treatment, if available) for flow cytometry and CIBERSORT. 36、37 This confirms the "responsive TME profile". Furthermore, the inventors have used their technology to determine peripheral blood T cell receptor sequencing. 76 "Hot" vs. "Cold" Tumor RNA Expression Signatures 77 PD-L1 tumor ratio score 78 , and tumor mutation burden 53、79 We will compare our method with other recent and novel methods. Our hope is that our technique demonstrates robustness and superiority over these other methods. Naturally, it is important to confirm our results in an independent external cohort, which we plan to do with the help of our clinical collaborators.

[0139] Finally, we determine whether severe toxicity can be predicted from immune checkpoint blockade using our liquid TME method (Figure 12). Unfortunately, there are no biomarkers for immune-related adverse events in clinical use. 20 This is a significant issue in patients with malignant tumors of the chest and gastrointestinal tract, because pneumonia and colitis can occur simultaneously when treatment is delivered in other forms. 80、81 Among the most common adverse events 20、21Therefore, to initiate this approach, the same cohort of approximately 50 patients as above will be used at the same time (during pretreatment and 2-3 weeks of treatment). Published literature 60 Based on this, we expect approximately 25% of these patients to experience severe toxicity. We classify these adverse events and correlate the type and incidence of toxicity with TIL and circulating leukocyte dynamics from our liquid TME assay. B cells and CD4 T cells have recently been suggested to be associated with immune-related adverse events. 22、62 The inventors determine whether this is a case where their assay should be used. As described above, the inventors' TME prediction is performed using flow cytometry and CIBERSORT. 36、37 Compare the results with pre-treated tumor biopsies (and, where available, those taken during treatment) analyzed by [method / tool ​​name]. External validation of our results is important and is planned. If successfully validated, these results will allow for safer delivery of immunotherapy by predicting severe toxicity before its clinical manifestation.

[0140] The inability to accurately predict immunotherapy response or toxicity early on is one of the most challenging problems in clinical cancer research. This problem can be solved by developing a highly innovative technique called liquid TME. Liquid TME has been able to revolutionize immunotherapy response and toxicity assessment in two ways. First, it can function as a primary assessment method that provides clinicians with accurate data when imaging and clinical assessments have proven inadequate. Second, it can help track patients sequentially, supplement ambiguous assessments from the inventors' standard clinical model, distinguish borderline responses from progression, and predict the severity of potential symptomatic toxicity. Since non-invasive TME assessment can be useful in multiple research and clinical settings, the inventors' research can be further generalized more broadly.

[0141] This technology tracks previously undescribed entities (ctilDNA) in a robust and comprehensive manner, and applies the inventors' technology to the most important clinical challenges in the field of oncology.

[0142] technological innovation The techniques described herein are exceptionally innovative because they concern a novel and previously undescribed component of cell-free DNA arising from the tumor microenvironment, and the inventors disclose a new technical method for profiling and tracking it in blood. The inventors' method offers a potential solution to one of the most important problems occurring in modern oncology: predicting which patients will respond to immunotherapy and which will be affected by serious toxicity from immunotherapy. If successful, liquid TME represents a groundbreaking advance in immunotherapy response and toxicity assessment and has clear clinical implications. This could revolutionize oncological practice by enabling the inventors to more accurately select and monitor patients, potentially impacting the lives of thousands of individuals each year. Furthermore, by non-invasively and robustly profiling the tumor microenvironment, the inventors' research should be generalized to almost all cancer types and anticancer therapies, opening the door to routine, non-invasive tumor microenvironment assessment in both research and clinical settings.

[0143] Various methods can be used to increase sensitivity. First, the inventors can utilize a sequencing panel targeted to include more differentially methylated regions. The inventors use ctilDNA 26、32 To detect them with greater sensitivity, sequencing can be performed to a deeper level. The main drawback of these optimizations is the increased sequencing cost. However, sequencing costs have been plummeting and are expected to continue decreasing. 82To further enhance sensitivity, the inventors may reduce the number of TME cell subsets being tracked, for example, limiting it to simply B cells, CD8 T cells, CD4 T cells, NK cells, and monocytes / macrophages rather than all 12 TME cell types mentioned above. If successful, this simplified method is expected to be clinically significant as it includes a broad category typically evaluated by standard flow cytometry. 83 . References [Table 3] TIFF0007874822000015.tif226162 TIFF0007874822000016.tif227162 TIFF0007874822000017.tif234162

[0144] Example 4: Non-invasive TIL quantification The following examples describe the development of an ultra-sensitive framework for profiling tumor-infiltrating leukocytes using cell-free DNA methylation profiles and evaluate the technical performance of non-invasive digital cytometry for profiling TILs from patients with metastatic melanoma in vitro.

[0145] Tumor-infiltrating leukocytes (TILs) play a crucial role in tumor growth, cancer progression, and patient prognosis. While techniques for characterizing TIL composition (e.g., flow cytometry, immunohistochemistry) have yielded deep insights into cancer biology and medicine, they generally require invasive, morbidity-related tumor biopsy or resection procedures that do not account for geographical tumor heterogeneity. Currently, there is no reliable method for non-invasively assessing TIL composition.

[0146] Liquid biopsy is a novel class of techniques for non-invasive tumor profiling based on cell-free DNA continuously flowing in circulation from normal and malignant cells. Despite the potential of cell-free DNA to enable safe and non-invasive assessment of diverse physiological states across sequential time points, no liquid biopsy method currently exists available for monitoring TIL composition. A genomics platform applied to cell-free DNA can enable non-invasive profiling of TIL subsets, allowing for accurate profiling of the tumor microenvironment. This can be achieved by deconvolution of cellular composition from methylated signatures, following next-generation sequencing after bisulfite treatment of plasma-derived cell-free DNA, which we apply to metastatic melanoma as a proof of principle. We hypothesized that our “non-invasive digital cytometry” method would enable accurate, biopsy-free monitoring of the tumor microenvironment, not limited to (1) small combinations of pre-selected marker genes (by flow cytometry), (2) T / B cell receptor variable regions (by VDJ profiling), or (3) viable single cells (by single-cell RNA sequencing). Importantly, the dynamics of cell-free DNA release from TILs are unknown, and it has not yet been established whether methylation signatures can quantitatively capture specific non-malignant tumor cell types from cell-free DNA. The following experiments were designed to address these technical issues and novel assays for safe, high-resolution profiling of TIL dynamics in cancer patients.

[0147] Development of an ultra-sensitive framework for profiling tumor-infiltrating leukocytes using cell-free DNA methylation profiles. It was hypothesized that DNA methylation signatures could robustly distinguish TILs from other cell types and enable highly sensitive quantification from small amounts of DNA.

[0148] A. Definition of a cell-type specific methylation signature that distinguishes normal peripheral blood leukocytes and non-hematopoietic cells from major TIL subsets. Herein, to define TIL-specific methylation sites, whole-genome bisulfite sequencing is applied to sorted melanoma TIL subsets, malignant melanocytes, stromal cells, and normal peripheral blood leukocytes. Next, a computational framework is developed and validated to estimate the proportion of each cell type from a mixture of methylated DNA.

[0149] B. Design and optimization of the performance of a targeted bisulfite sequencing panel for profiling TILs from clinically realistic DNA input amounts. The inventors devise an analytical method for designing a cost-effective capture sequencing panel that targets multiple TIL-specific genomic reporters while maximizing sensitivity from small amounts of DNA (e.g., the amount of cfDNA obtained in a single blood draw).

[0150] Technical performance evaluation of non-invasive digital cytometry for profiling TILs from patients with metastatic melanoma in vitro Non-invasive digital cytometry was hypothesized to faithfully capture the TIL content in defined in vitro mixtures and cell-free DNA from melanoma patients.

[0151] A. Technical performance evaluation of non-invasive digital cytometry using defined in vitro mixtures. To evaluate the detection accuracy and lower limit of the inventors' method, the inventors create a series of defined mixtures by in vitro adding sonicated DNA from tumor leukocyte subsets to cell-free DNA from healthy donors. The total leukocyte content emulates the immune levels in melanoma tumors adjusted for clinically realistic amounts of circulating tumor DNA. Using the above panel, targeted bisulfite sequencing is applied to these DNA mixtures at a range of input amounts, and non-invasive digital cytometry is used to evaluate the TIL content. Therefore, the inventors establish performance expectations and adjust their method to maximize sensitivity and specificity.

[0152] B. Non-invasive TIL profiling is performed in melanoma patients, and the concordance with paired tumors is evaluated. For in vivo validation, the inventors analyze registered and viably preserved tumor, plasma, and peripheral blood mononuclear cell (PBMC) samples (from the time of concordance) from 30 patients with metastatic melanoma. In parallel, the inventors process registered blood samples (plasma and PBMCs) from 10 age-matched healthy controls (where TILs should not be present). The inventors compare TIL predictions by the inventors' method with orthogonal measurements of TIL content in paired tumors (e.g., by flow cytometry), and compare methylation signatures from cell-free DNA with cellular DNA (PBMCs) to determine which compartments better capture the known TIL composition.

[0153] Research method significance Tumor-infiltrating leukocytes (TILs) play a crucial role in tumor growth, cancer progression, and patient prognosis (1-8). Recent advances in immuno-oncology have revolutionized cancer treatment, but patient responses to existing and new immunotherapies are often heterogeneous, and effective predictive biomarkers are lacking (9-12). For example, there are currently no biomarkers with high sensitivity / specificity to predict early on which patients may and may not benefit from immune checkpoint inhibitors (ICIs) (11-13). While several powerful techniques are available to characterize TIL composition (e.g., flow cytometry, immunohistochemistry, CyTOF, single-cell RNA sequencing), they generally require tumor biopsy or resection procedures that are invasive (14), morbidity-related (15), and do not take into account geographical tumor heterogeneity (16, 17). Consequently, most analyses of human TIL composition are limited to a single snapshot of tumor heterogeneity obtained from a single point in time, due to the limited availability of tumors.

[0154] This obstacle leaves a significant gap in the inventors' understanding of TIL dynamics and hinders their ability to utilize these cells for the development of more effective biomarkers and therapies.

[0155] The techniques described herein may be novel for non-invasive TIL quantification. The ability to non-invasively monitor TIL composition offers an attractive solution to the above-mentioned problems in both research and clinical settings. However, currently there is no reliable method for TIL assessment without biopsy. Previous studies of peripheral blood leukocytes (PBLs) in cancer patients have identified subpopulations that are similar to those found in tumors and have prognostic / predictive properties (18, 19); however, the cell type marker profiles used in these studies are unlikely to be TIL-specific, and the extent to which these cells truly capture the tumor immune composition is unclear (20). Separately, highly specific T cell receptor (TCR) clonal types from tumors can be discovered and tracked in peripheral blood (21, 22), but this method (1) provides a limited understanding of TIL heterogeneity and (2) cannot distinguish tumor-derived T cells from normal T cells without a highly biased clonal representation of tumor-specific TCRs or prior knowledge.

[0156] Over the past few years, many groups, including the inventors, have developed and validated techniques for non-invasively detecting tumor burden and tumor genotype using plasma-derived circulating tumor DNA (PIL) isolated, quantified, and sequenced in the form of cell-free DNA released into peripheral blood (23-26). Physiological cell-free DNA in blood is thought to originate mostly from non-malignant cells and arise from cell death due to necrosis, apoptosis, phagocytosis, and possibly active secretion (24-26). This increases the likelihood that TIL-derived cell-free DNA is detectable in plasma and could serve as non-invasive reading data for TIL heterogeneity. While several studies have demonstrated high sensitivity (24, 27-30) in profiling and tracking circulating tumor DNA using PCR and next-generation sequencing (NGS)-based methods, the extent to which cell-free DNA can penetrate TIL biology in solid tumors remains unclear. This specification describes a novel method demonstrating that TIL DNA can be detected and quantified in the plasma of cancer patients. This technology has implications for non-invasive TIL diagnosis.

[0157] The development of assays for non-invasive TIL profiling has revolutionized our understanding of tumor immunology by applying the discovery of improved biomarkers for diverse anticancer therapies. For example, ICIs are now transforming cancer care, improving the prognosis of a subset of patients with advanced cancer, giving patients a superior therapeutic response, and enabling a subset of these responders to achieve long-term survival (9, 31-33). ICI response rates for different cancers range from 1% to 50% (34), and response rates are influenced by multiple factors, including tumor PDL1 expression, tumor mutational burden, neoantigen burden, and tumor histological diagnosis (34-37). Standard care for evaluating ICI response is serial CT imaging initiated 2-3 months after the start of immunotherapy (38) and assessed by RECIST 1.1 (39) or iRECIST (40) criteria. CT imaging is typically performed within 2-3 months after the start of treatment due to concerns about delayed radiographic response and early pseudo-progression (13, 38, 41). This approach allows researchers to explore ways to assess immunotherapy responses more quickly, enabling a rapid shift towards more effective treatment modes for patients with advanced disease, including the majority of patients.

[0158] To this end, we evaluate the technical performance of our assay in patients with advanced melanoma, a “poster child” (42) of solid tumor immunotherapy. While some melanoma patients exhibit a persistent antitumor T-cell response to ICI, many are unable to respond, and treatment is often associated with immune-related adverse events such as colitis, pneumonia, hepatitis, and endocrine disorders (43, 44). Cell-free DNA and circulating tumor DNA concentrations are typically elevated in patients with metastatic melanoma (29, 45), indicating the presence of sufficient material to non-invasively assess this compartment. Given the heterogeneous clinical prognosis, high cell-free DNA content, and established role of immunotherapy, we believe it is worthwhile to focus on melanoma for this technical study.

[0159] technological innovation This technology provides a platform for the following innovations:

[0160] Firstly, cell-free DNA has informational value regarding the origin tissue, including methylated cytosines in CpG dinucleotides, possessing distinct lineage-specific patterns and epigenetic signatures that can be profiled using bisulfite sequencing (46). The Lo group demonstrated that genome-wide bisulfite sequencing enables tissue identification of plasma-derived cell-free DNA in pregnant women, organ transplant patients, and hepatocellular carcinoma patients (47). Zhang and collaborators applied whole-genome bisulfite sequencing using the linkage disequilibrium principle to identify tightly bound CpG sites called methylated haplotype blocks (48). Methylated haplotype blocks are more accurate than conventional methylation metrics in distinguishing between tissue-specific methylation patterns and enable tissue identification of cancer origin from cell-free DNA in patients with different malignancies (48). Despite these results, the composition of the tumor immune microenvironment has not been profiled by methylation signatures in cell-free DNA. This technology can generate a novel framework to address this gap using targeted bisulfite sequencing.

[0161] Secondly, flow cytometry and immunohistochemistry are commonly used to analyze tissue cell composition. However, both methods generally rely on small combinations of pre-selected marker genes, limiting the number of cell types that can be examined simultaneously. Single-cell RNA sequencing has emerged as a powerful technique for defining novel cell subsets (49), but is not currently practical for large-scale analysis. To complement these methods and facilitate cell profiling of large patient cohorts, we previously developed CIBERSORT, an "in silico flow cytometry" method for enumerating cell composition from bulk tissue gene expression profiles (50). When evaluated with fresh, frozen, and fixed specimens, CIBERSORT outperforms previous calculation methods and is favorably comparable to flow cytometry and immunohistochemistry (3, 50). Furthermore, in a pan-cancer analysis of tumors from approximately 6,000 patients, CIBERSORT revealed important new correlations between TILs and clinical outcomes (3). This method can be adapted to the deconvolution of cell-free DNA bisulfite sequencing data, allowing for the determination of the proportion of separate TIL subsets from cell-type specific methylation profiles identified in cell-free DNA.

[0162] Thirdly, this method can help address a major unmet need: high-resolution monitoring of TIL dynamics across continuous time points, which will advance biomarker discovery and precision cancer medicine.

[0163] method The experiments described herein aim to develop and experimentally evaluate a novel platform for non-invasive profiling of TILs from melanoma patients. This study may include an innovative combination of experimental and computational techniques, including tools developed by the research team, to construct a novel genomic platform for profiling and decoding TIL-derived methylation signatures identified from plasma-derived cell-free DNA molecules. The research plan is schematically shown in Figure 14. This specification describes the application of whole-genome bisulfite sequencing to define cell-type-specific methylation signatures of major TIL subsets from primary patient tumors. The design and optimization of next-generation sequencing panels and corresponding computational frameworks for profiling TIL-specific methylation sites from clinically viable amounts of plasma-derived cell-free DNA are also described. The inventors describe testing their “non-invasive digital cytometry” assay by evaluating its performance with both a defined DNA mixture and DNA isolated from tumor, peripheral blood, and plasma obtained from metastatic melanoma patients, along with ground truth data (ground truth) of the proportion of TILs in paired tumors determined by flow cytometry. In both experimental sets, registered and despecified melanoma biological samples available from Yale SPORE (YSPORE) for skin cancer can be utilized. These samples, including melanoma biopsies, plasma samples, and peripheral blood leukocyte samples, are collected with notified and signed consent from participants in accordance with the Human Investigation Committee Protocol and Health Insurance Portability and Accountability Act (HIPAA) rules.

[0164] Development of an ultra-sensitive framework for profiling tumor-infiltrating leukocytes using cell-free DNA methylation profiles. Definition of cell type-specific methylation signatures to distinguish major TIL subsets from normal peripheral blood leukocytes and non-hematopoietic cells. Logical basis

[0165] High-throughput methylation profiling has revealed exceptional insights into the epigenetic landscape of separate histological types and cell lineages, including normal immune subsets (53). However, to our knowledge, comparative analyses of genome-wide methylation signatures in major melanoma TIL subsets versus their normal peripheral blood counterparts remain undescribed. To successfully identify and quantify TIL subsets using methylation profiles identified by bisulfite sequencing, it is crucial to first characterize genome-wide patterns of differentially methylated CpG dinucleotides in melanoma TILs, melanoma and healthy PBL subsets, and non-hematopoietic cells.

[0166] Methods and data The inventors analyzed registered, viable-preserved tumor and peripheral blood mononuclear cell (PBMC) samples from five patients with metastatic melanoma and isolated TILs, tumor cells, stromal elements, and PBLs by fluorescence-activated cell sorting (FACS). PBLs were also evaluated from a cohort of five age-matched, healthy, non-pregnant individuals (where TILs should not be present) (obtained as described above). Six major leukocyte subsets were profiled from the PBLs and tumor samples: CD8 T cells, CD4 T cells, NK cells, B cells, monocytes / macrophages, and granulocytes / bone marrow-derived suppressor cells (MDSCs). The inventors extracted at least 100 ng of genomic DNA from each of these samples (approximately 10k cells / sample), including corresponding bulk tumors and PBLs, and performed methylation profiling by whole-genome bisulfite sequencing (WGBS) targeting a coverage of 4050 per sample, along with 225M 150bp×2 read data on Illumina NovaSeg. Importantly, WGBS has been shown to achieve better CpG coverage than reduced-resolution bisulfite sequencing, an alternative technique that uses restriction enzymes to enrich CpG sites (54). WGBS allows for the examination of CpG sites across the entire genome at single-nucleotide resolution, maximizing the number of detectable identification markers. As a quality control step, methylation profiles from three cancer cell lines were profiled and compared with publicly available WGBS data (55). As described above, we plan to evaluate two commercially available kits for WGBS. The read data is mapped to the genome and processed to identify methylation sites as previously described (56, 57). Samples obtained from the same human donor are validated by evaluating germline SNP agreement (58).

[0167] To identify differentially methylated regions (DMRs) that improve TIL-specific quantification and error tolerance, we apply a previously described linkage equilibrium-based method (48) to identify methylated haplotype blocks (regions with multiple methylated CpGs within approximately 200 consecutive bases; cell-free DNA molecules are highly stereotypic in length, approximately 170 bp (27, 28)). To improve marker specificity, we omit further consideration of any genomic regions corresponding to haplotype blocks that are significantly differentially methylated / expressed in non-hematopoietic tissues, cell types, and melanoma using data from the NIH Roadmap Epigenomics Project, ENCODE, BLUEPRINT, and WGBS data generated in this study. Next, we analyze the remaining haplotype blocks to identify highly specific signatures for each cell type using our previously described method (50), adapted for methylation data. The discriminative power of these signatures is evaluated by applying CIBERSORT(50) to bulk tissue methylation profiles with ground truth data ratios determined by FACS. To assess the generalizability of leukocyte signatures, an artificial mixture including publicly available DNA methylation profiles (59-63) from a normal leukocyte population is also evaluated. These analyses are used to establish a minimum set of DMRs that best distinguish melanoma tumors and leukocyte subsets, including TILs and PBL populations.

[0168] As proof of principle, the CIBERSORT signature matrix was trained to distinguish the major PBL subset profiled on the Infinium HumanMethylation 450K BeadChip array (64). When applied to whole blood methylation profiles generated by the two groups (65, 66), a very significant agreement with the proportions determined by flow cytometry was observed (Figure 18). Furthermore, given the strong agreement between the methyl chip array and WGBS (67), bisulfite sequencing should also be performed if it is not good due to the higher resolution applicability of CpG.

[0169] Finally, the inventors compare the deconvolution performance between hypermethylated and hypomethylated regions by in-silico simulation, if any, to determine which of the two events should be prioritized in the inventors' panel design.

[0170] Considering numerous reports (20, 68 - 71) of differences in the phenotypic states of TIL, normal adjacent tissues, and normal peripheral blood leukocytes, a number of prominent TIL subset-specific methylation blocks are identified. Furthermore, the identified WGBS methylation profiles will be made available as a community resource to facilitate further research into TIL-specific epigenetics. Separately, considering promising data (Figure 18), robust TIL deconvolution from mixed methylation profiles is expected.

[0171] The 4050 application range may be insufficient to robustly identify single and / or allelic methylation events. If so, additional sequencing is performed to target the 65 application range. Since specific TIL subsets should not be distinguishable from normal leukocytes, consider excluding them from further analysis or pooling them into a broader lineage.

[0172] Design and optimization of a targeted bisulfite sequencing panel for profiling TIL from clinically realistic DNA input amounts Rationale

[0173] Some commercially available bisulfite sequencing kits are compatible with low input DNA amounts (e.g., the amount of cell-free DNA obtained in one blood collection tube (28)). Nevertheless, achieving low-cost and high-sensitivity TIL cell-free DNA profiling requires the design of a custom capture panel. The design of a targeted sequencing panel that includes multiple TIL-specific genomic reporters to maximize analytical sensitivity and improve error tolerance is described herein (27, 28, 51).

[0174] Methods and data To develop this assay, we evaluated both commercially available and published methods for panel design (e.g., NimbleGen SeqCap Epi Choice probe S vs. molecular inversion probe (72)) and bisulfite sequencing (e.g., Zymo EZ DNA Methylation-Lightning kit, Swift Biosciences Accel-NGS Methyl-Seq DNA) to determine the trade-offs between cost, DNA recovery, and bisulfite conversion efficiency.

[0175] Three key factors—(1) the number of cell-free DNA molecules recovered, (2) the number of independent “reporters” in the patient tumors being examined, and (3) the technical background—form the basis for the detection limits of cell-free DNA application (27, 28). For the first two factors, using the previously described validated binomial models (27, 28) to predict the detection limits of circulating tumor DNA, we estimated the number of specific cell-type-specific differential methylation regions (DMRs; i.e., "reporters") required to achieve various detection limits, taking into account (1) realistic cell-free DNA input (approximately 32 ng of cell-free DNA per blood collection tube (28)), (2) median circulating tumor DNA fraction in metastatic melanoma (approximately 1% (45)), (3) estimated TIL content in advanced melanoma tumors (3, 72, 73), (4) estimated cell-free DNA recovery rate after bisulfite conversion (20–60% (74)), and (5) cell-free DNA recovery rate published using hybrid capture sequencing (40–60% (28)). Considering approximately 10,000 genomic equivalents of cell-free DNA (assuming approximately 32 ng of cell-free DNA), the library Assuming an 80% DNA loss from preparation, modeling suggests that >10 DMRs per cell type are sufficient for TIL detection with 95% confidence (Figure 10). Furthermore, only 12 DMRs per cell type are required to achieve a detection limit of approximately 0.01% with a probability of 0.9. Since it is expected that tens to hundreds of DMRs will be covered per cell type, this means that the theoretical recovery of multiple DMRs per cell type with respect to the detection limit must be as low as 0.01%. This is well within the range of ultra-deep targeted sequencing, as demonstrated in our previous studies (27, 28). Moreover, while the set of specific DMRs per cell type is unique, deconvolution is used to resolve DMRs shared by >1 cell type. Regarding a third factor, bisulfite conversion efficiency and the inherent error rate of NGS can degrade analytical sensitivity.

[0176] The former has been reported to be high (>99% (74)) for many kits, but this needs to be verified. We previously showed that capture-based NGS can detect circulating tumor DNA down to as low as 0.02% without using a specific molecular identifier (UMI) (27). To correct errors in a manner similar to error-tolerant DNA barcode sequences, we utilize methylated haplotype blocks with multiple expected CpGs (75).

[0177] To construct the panel, first, cell type-specific DMRs within haplotype blocks that optimize deconvolution performance are identified, as described herein. Next, 147,888 methylated haplotype blocks published by Guo et al. (48) are reviewed to identify any further methylated haplotype blocks to be co-isolated with the obtained signatures for inclusion in the panel. Other regions are added according to their clinical or biological relevance (e.g., ICI co-inhibitory receptors) until a final size of approximately 200 kb (approximately 2,000 genomes per 100 bp) is achieved.

[0178] (1) Define TIL-specific, PBL-specific, and melanoma tumor-specific methylation signatures for the purpose of deconvolution, and (2) design an optimized targeted hybrid capture panel with genomic bandwidth for profiling TIL and PBL subsets.

[0179] The inventors' capture panel may have insufficient sensitivity to distinguish between separate leukocyte and tumor populations. If this is the case, the panel can be redesigned to relax the inventors' criteria for methylated haplotype blocking. This would allow for the consideration of DMRs with lower-density clustered CpGs, which could identify additional identification markers that improve performance.

[0180] Separately, if the error rate of bisulfite sequencing proves too high for profiling cell-free DNA derived from TILs in minute abundances of less than 0.1%, the inventors consider designing a custom sequencing adapter with a bisulfite-tolerant UMI.

[0181] Technical performance evaluation of non-invasive digital cytometry for profiling TILs in patients with metastatic melanoma in vitro. Technical performance evaluation of non-invasive digital cytometry using a specified in vitro mixture. This specification describes an evaluation of the technical performance of non-invasive digital cytometry using a specified in vitro mixture.

[0182] Theoretical basis To evaluate the accuracy and lower limit of detection of our method, it may be important to establish initial performance predictions in a controlled in vitro titration series. This will help to adjust our method to maximize sensitivity and specificity.

[0183] Experimental method A series of defined mixtures are prepared by adding sonicated DNA from tumor leukocyte subsets (remaining from the above or selected from two additional patients) to cell-free DNA (obtained as described herein) from healthy control subjects in vitro. Total leukocyte content ranges from 5% to less than 0.01% to emulate typical immune levels in metastatic melanoma tumors (3, 72, 73) adjusted for clinically realistic circulating tumor DNA levels (27–30, 45, 51). Using the above panel, targeted bisulfite sequencing is applied to 10, 20, 30, and 50 ng of DNA mixtures, and TIL content is assessed by deconvolution.

[0184] We anticipate that by performing bisulfite sequencing of a defined genomic and cell-free DNA mixture, we can non-invasively profile leukocyte populations and distinguish between TILs and non-TILs.

[0185] Non-invasive TIL profiling in melanoma patients and evaluation of concordance with paired tumors Theoretical basis To evaluate the usefulness of non-invasive TIL profiling in vivo, it is important to compare the estimated TIL composition in the plasma of melanoma patients with orthogonal measurements of TIL content in paired tumors (e.g., by flow cytometry). In addition, comparing methylation signatures from cell-free DNA with those from cellular DNA (PBMCs) will determine which compartments better capture known TIL composition. These data may be useful in establishing baseline values ​​for power calculations and dedicated biomarker studies.

[0186] Experimental method Registered and viably preserved tumor, plasma, and PBL samples from 30 patients with advanced melanoma will be analyzed. Patients will be demographically matched to local populations and will not be intended to exclude any particular sex / gender or minority groups. Patients will undergo tumor biopsies and blood will be collected prior to pretreatment. Evaluating a subset of patients with recurrent samples will allow for assessment of changes in TIL content from baseline. In parallel, registered whole blood samples (plasma and PBL) will be processed from 10 age-matched healthy non-pregnant donors (TILs should not be present) obtained from local blood banks, regardless of demographic characteristics or any particular sex / gender. High background DMR in healthy cell-free DNA will be omitted from further analysis based on our previous research (28).

[0187] Cell-free DNA from plasma samples and genomic DNA from tumor and PBL samples are isolated, subjected to bisulfite conversion, and targeted sequencing is performed using the panel described herein. Subsequently, NGS and deconvolution are applied using the techniques described herein. Cell-free DNA is extracted from approximately 5 ml of plasma using the QiaAmp Circulating Nucleic Acid Kit according to the manufacturer's instructions and stored at -80°C. Following isolation, DNA is quantified using the Qubit dsDNA High Sensitivity Kit (Life Technologies) and a Bioanalyzer (Agilent) to examine the expected fragment length distribution and yield. For input, a median of 32 ng of cell-free DNA and 100 ng of tumor or PBL DNA per sample is targeted for library preparation using the KAPA LTP Library Prep Kit (Kapa Biosystems). High-throughput sequencing is performed using Illumina HiSeq 4000 or NovaSeq 6000, targeting a median of approximately 10,000× non-overlapping depths. Samples obtained from the same human donor are confirmed by evaluating the matching of germline SNPs (58).

[0188] In parallel, flow cytometry is performed on tumor and PBL samples to evaluate the relative fractions of each leukocyte population. The inventors' deconvolution results from each compartment are compared with flow cytometry results from the tumor to (1) evaluate the accuracy and precision of the method and (2) determine whether cell-free DNA or PBL genomic DNA better captures the composition of the tumor immune microenvironment.

[0189] (1) Non-invasive profiling of the leukocyte population by bisulfite sequencing of cell-free DNA; (2) accurate quantification of TILs and differentiation of TILs from the normal leukocyte population in the cell-free DNA compartment; (3) demonstrating the superiority of cell-free DNA over PBL in capturing TIL content; and (4) demonstrating high methodological specificity of TIL detection by comparing it with cell-free DNA and PBL from healthy donors.

[0190] Unexpectedly, cell-free DNA concentrations may be too low for deconvolution of different TILs and tumor populations. This is not expected to be a major problem, as studies have shown high circulating tumor DNA concentrations in metastatic melanoma patients that are sufficient for NGS-based methylation profiling. However, the biology and dynamics of cell-free DNA from TILs are unknown. Where necessary, we will attempt to purify the signatures obtained from the above to improve detection, including increasing the number and amount of input cell-free DNA genomic equivalents and sequencing, expanding the use of our sequencing panel to include more methylation reporters, and possibly expanding its use to whole-genome bisulfite sequencing. If these methods are not yet successful, we can focus on peripheral blood cell compartments (rather than cell-free DNA) to profile TILs present in circulation. References [Table 4] TIFF0007874822000019.tif230162 TIFF0007874822000020.tif230162 TIFF0007874822000021.tif230162 TIFF0007874822000022.tif230162 TIFF0007874822000023.tif230162 TIFF0007874822000024.tif230162 TIFF0007874822000025.tif230162 TIFF0007874822000026.tif230162 TIFF0007874822000027.tif230162 TIFF0007874822000028.tif230162 TIFF0007874822000029.tif230162 TIFF0007874822000030.tif140161

[0191] Example 5: Development of a liquid biopsy technique for diagnosing and monitoring sepsis problem Sepsis is the most common cause of death in US hospitals and the leading cause of death worldwide, with 11 million sepsis-related deaths reported in 2017. Sepsis is difficult to diagnose and monitor in its early stages because it is challenging to determine whether a patient is infected (bacterial cultures take time to grow), the site of infection (requiring imaging and microbial cultures), and the location and extent of peripheral organ damage (often clinically determined, i.e., changes in mental state as a marker of brain damage). Unfortunately, if not detected early, patients miss crucial early interventions, and sepsis progresses rapidly, leading to life-threatening multi-organ failure, septic shock, and immunosuppression resulting in fatal secondary infections. There are no reliable biomarkers clinically used for the early diagnosis and monitoring of sepsis.

[0192] solution This specification discloses the development and testing of a liquid biopsy technique called LiquidMIDOS for organ damage in microbial infection, immunodeficiency, and sepsis, which, via whole-genome bisulfite sequencing of plasma cell-free DNA (Figure 19), enables: 1) detection of the microbial etiology of sepsis; 2) identification of sepsis tissue sites; 3) determination of organs at risk of failure due to damage if not actively managed; 4) determination of whether the adaptive immune response to sepsis has become dysfunctional and exhausted, which can be precisely managed with immunotherapy in the future; and 5) detection of fatal secondary infections at the prognosis. We compare our method with clinical and laboratory studies conducted in hospitals using standard treatments. If successful, our method should enable early diagnosis and monitoring of sepsis and associated peripheral organ damage, immunodeficiency, and secondary infections, which should have a direct clinical impact by potentially saving thousands of lives in the United States and millions worldwide.

[0193] Sepsis is the most common cause of death in US hospitals and accounts for one in five deaths worldwide. 4 Sepsis is defined as life-threatening organ failure caused by a dysregulated immune response to infection. In 2017, 11 million deaths were reported as being associated with sepsis. 4 The mortality rate associated with sepsis is unacceptably high, at 15–25%, and significantly higher in patients diagnosed with associated multiple organ failure. 6~8 Unfortunately, the problem has become even more dire, with a record number of sepsis cases and related deaths witnessed in ICUs in 2020. 9,10 The most important prognostic factor in sepsis is early intervention, which is hindered by diagnostic challenges. Early diagnosis and intervention are crucial for maximizing survival in this high-risk patient population.

[0194] The diagnosis of sepsis depends on a confirmed diagnosis of microbial infection. Infection is typically determined by bacterial cultures, which take time to grow, usually 24–72 hours, with some organisms taking more than 5 days to grow in culture. Bacterial cultures are also not the primary cause of other sources of sepsis, such as viral infections, which account for the recent increase in the proportion of sepsis patients. 10 Biomarkers suggestive of systemic inflammation, such as C-reactive protein, white blood cell count, and procalcitonin, have also been tested, but their sensitivity and specificity are limited, especially in the early stages and in immunosuppressed states. 11~14 Early diagnosis of infection is crucial to prevent delays in treatment and improve patient survival.

[0195] The source of infection can also be difficult to determine early during sepsis, requiring extensive work including chest X-rays, fecal cultures, urine cultures, wound cultures, and blood cultures, which can lead to further delays and confusion in diagnosis. Identifying the site of infection is a crucial determinant of management and prognosis, with unknown sites of infection and lung sites having the highest mortality rates. 15、16 Prioritize early determination of the infection site and source using LiquidMIDOS.

[0196] Even when clinicians suspect sepsis and promptly initiate treatment, there are no reliable biomarkers to track the treatment response. Accurate monitoring of the sepsis response to treatment is crucial for patient survival.

[0197] Another important diagnostic factor in sepsis is organ damage. Unfortunately, single-organ failure can progress to multiple organ failure syndrome (MODS) in sepsis patients who do not receive adequate pre-emptive care in the acute setting. When this occurs, homeostasis is no longer maintained, and the patient's prognosis becomes dire. The more organ systems affected, the higher the mortality rate, reaching approximately 100% when more than five organ systems are affected. 7 Early identification of organ damage is crucial to prevent MODS and the associated high mortality rates.

[0198] Cases of sepsis that were not diagnosed early also incurred a significantly higher financial burden. For patients diagnosed early (at the time of hospitalization), the cost was $18,023 per patient, but when diagnosis was delayed, it jumped to a staggering $51,022. 17 The total cost of hospitalization for sepsis management in U.S. hospitals is the highest of all disease conditions, accounting for $24 billion in 2013, representing 13% of total hospital costs in the U.S. 17 These numbers could swell even further due to the Covid-19 pandemic. 10 The main reason is the length of stay and intensive care required for these patients. Accurately diagnosing and monitoring sepsis with an integrated assay should help reduce the financial burden on patients, in addition to improving their prognosis.

[0199] Sepsis is also an immunological challenge, with the early acute phase typically involving hyperimmunity due to a dysregulated immune "cytokine storm," which requires intensive care and can lead to death from septic shock or multiple organ failure. 5、18、19 If the patient recovers, a hypoimmune phase will follow a few days after this hyperimmune phase, characterized by the depletion and dysfunction of T cells, which are crucial cells in the adaptive immune system, putting the patient at risk of fatal secondary infections (Figure 20). 5、18~21 The majority of these dysfunctional / wasted T cells are located within tissues. 20 Therefore, methods for sensitively and proactively detecting them need to allow for the investigation of their tissue sources.

[0200] Interestingly, more patients survive the initial acute hyperimmune phase of sepsis than the subsequent immunodepletion phase. 5 13–30% of sepsis patients develop fatal secondary infections from opportunistic microorganisms that are usually less likely to affect individuals with a functionally adaptive immune system. 5、22、23 Flow cytometry and gene expression analysis of peripheral blood cells showed no differences at the initial stage. 23、24 Therefore, it is necessary to investigate the tissue source of the depleted immune cells. 20However, biopsies are risky, impractical, and rarely performed in acute care settings. It is crucial to non-invasively and accurately identify the T-cell dysfunction / exhaustion phase of sepsis to reduce the risk of fatal secondary infections.

[0201] The inventors address these key challenges with a non-invasive plasma cell-free DNA liquid biopsy technique called LiquidMIDOS. Specifically, LiquidMIDOS aids in the early diagnosis and monitoring of sepsis by (1) detecting the microbial pathogenesis of sepsis, (2) identifying the septic tissue site, (3) determining which organs are damaged, (4) determining whether the T cell response is dysfunctional, and (5) detecting secondary infections (Figure 19). LiquidMIDOS achieves these objectives through a single assay from a single blood sample that can be performed early and sequentially to improve patient survival.

[0202] The inventors' method for developing LiquidMIDOS utilizes the fact that tissues throughout the body continuously release DNA into circulation, which can then be isolated as cell-free DNA (cfDNA). 1、25、26 Cell-free DNA is released into the bloodstream through cell turnover and death. 27 Therefore, a new next-generation sequencing (NGS)-based technology has been developed that enables the detection of tissue-specific cfDNA at a low level of approximately 0.01% of the total cell-free DNA extracted from a single blood tube. It has been shown that not only tissue cells secrete cfDNA, but also microorganisms, including bacteria, DNA viruses, fungi, and eukaryotic parasites, secrete cfDNA that can be measured via NGS. 29 Furthermore, it was hypothesized that dysfunctional / depleted T cells release cell-free DNA that can be accurately measured by NGS via advanced analytical methods and distinguished from the much larger amount of cfDNA originating from peripheral blood leukocytes (Figure 21). Quantification of cell-free DNA originating from microorganisms, organ-specific tissues, and depleted T cells is described herein to proactively determine the state and etiology of infection, the organ involved and damaged, and immunosuppression.

[0203] The inventors' method relies on both cell-free DNA genomics and epigenomics. The epigenome consists of compounds bound to DNA molecules that indicate which parts of the genome are on or off. 30 Each cell type and tissue type is determined by a method called bisulfite sequencing. 31、32 It possesses a unique epigenomic signature that can be profiled by analyzing the methylation pattern on its DNA. 30 These epigenomic signatures can be used to detect the involved / damaged tissue types and cfDNA released by exhausted T cells via machine learning-based deconvolution.

[0204] Recent published data demonstrates the ability of methylation-based plasma cell-free DNA analysis to detect cancer origin tissue (from an excess of different human histological types) with high sensitivity. 1、28、33 Furthermore, the inventors will achieve a wide dynamic range necessary to measure different levels of organ damage, as recently demonstrated for liver injury (Figure 22). 1 Recent literature has also shown that, although genome-wide sequencing methods offer less depth, they can achieve excellent detection sensitivity comparable to targeted ultradeep sequencing by enabling the tracking of a far greater number of specific reporters through genome-wide sequencing. 34 Furthermore, as previously shown 29 Using whole-genome sequencing of plasma cell-free DNA, multiple infectious microbial species can be detected with high sensitivity. Therefore, the implementation of genome-wide sequencing of cell-free DNA for the highly sensitive detection of sepsis-related / damaged tissues and microbial sources is supported by recently published literature. However, the ability of cell-free DNA analysis to detect immune exhaustion has not been demonstrated. Furthermore, LiquidMIDOS is the first integrated method that combines microbiological analysis, immune exhaustion, and organ tissue analysis from a single blood tube using a single assay.

[0205] LiquidMIDOS in sepsis However, the principle should be applicable to different disease etiologies in this specification. Sepsis is a leading cause of hospital death in the United States and is the most common cause of death worldwide. 4 Therefore, we chose to focus on sepsis. By focusing on sepsis, we could test LiquidMIDOS in a situation that would produce the greatest impact, and thus we have plasma samples along with available paired clinical data.

[0206] To explore our all-in-one liquid biopsy method, we first conducted a mathematical modeling exercise (Figure 23). Factors underlying the detection limit of cell-free DNA analysis include the number of independent "reporters" being examined. 34、35 A previously described validated binomial model for predicting the detection limit of circulating tumor DNA. 35 Using this method, (1) a realistic amount of cell-free DNA input (approximately 50 ng of cell-free DNA in one blood collection tube), and (2) 10% each of relevant / damaged organ-specific cfDNA, exhausted / dysfunctional T cell-specific cfDNA, and microorganism-specific cfDNA. 1 , 1% and 0.4% 2 Considering these factors, the probability of cell-free DNA detection was estimated based on the number of specific compartment-specific reporters (i.e., organ-tissue-specific differential methylation regions, microorganism-specific genome sequences). Our mathematical modeling suggests that two or more reporters per organ-tissue type, more than 15 specific to exhausted T cells, and more than 40 microorganism-specific genome motifs enable highly sensitive specific detection with a probability of over 90%. Our model suggests that LiquidMIDOS enables highly sensitive cell-free DNA detection, particularly considering that millions of potential compartment-specific reporters are available genome-wide. 34 .

[0207] Next, the inventors sought to determine if high-quality sequencing results could be achieved using available, registered plasma samples. First, DNA was input into library preparations ranging from 30 ng to 120 ng using the Accel-NGS Methyl-Seq workflow (Swift Biosciences), and when targeted by multiplex sequencing on an Illumina NovaSeq S4 flow cell, a sequencing depth of 4050 was reliably achieved. Next, another practical question was sought: does freezing affect the ability to reliably measure methylation patterns? To answer this, whole-genome bisulfite sequencing (WGBS) was performed on nine peripheral blood leukocyte samples from healthy donors, with three samples being fresh (unfrozen), three samples having frozen DNA, and the remaining three samples having cells cryopreserved before further processing, performing all preparations for use. Following sequencing analysis, the inventors observed no major differences in global methylation patterns (Figure 15), indicating that cryopreserved cells or DNA did not affect the prior art. 36 This suggests that it does not introduce epigenomic artifacts, consistent with the findings.

[0208] Next, we investigated whether separate methylation reporters could be identified in tissue-derived epithelial cells, tissue lymphocytes enriched with exhausted T cells, and normal peripheral blood leukocytes (PBLs). It was crucial to establish that the epigenomic signatures were distinctly different among these three cell classes. Therefore, flow cytometry was performed to isolate epithelial cells, PBLs, and tissue lymphocytes from 10 patients with minor metastatic colorectal cancer. To focus on exhausted T cells, a flow cytometry method was developed to specifically select these cells from pre-sequencing tissue (Figure 24). Subsequently, after WGBS for each sample, differential methylation region (DMR) analysis was performed to identify 70 of the most differentially methylated CpG locations (Figure 1). This revealed that epithelial cells, tissue lymphocytes (enriched with dysfunctional / exhausted T cells), and PBLs possess distinct methylation profiles, suggesting that they can be epigenomically distinguished using WGBS.

[0209] Next, we investigated whether exhausted T cells could detect signals from enriched epithelial tissue and tissue lymphocytes in cell-free DNA. To do this, we isolated plasma cell-free DNA from 13 patients with minor metastatic colorectal cancer and performed WGBS in an Illumina NovaSeq S4 flow cell targeting 4050 genome-wide coverage. Specific epithelial tissue vs. tissue lymphocyte vs. PBL reporters shown in Figure 1 were analyzed using CIBERSORTx. 37 The data was deconvolved by examining it using [method name]. Using this method, we were able to detect leukocyte-derived cfDNA from all patients, epithelial tissue-derived cfDNA from 9 out of 13 patients, and tissue lymphocyte-derived cfDNA from 9 out of 13 patients (Figure 2A). Furthermore, the levels of epithelial and tissue lymphocyte-derived cfDNA using our methylated cell-free DNA deconvolution method correlated significantly with the sum of the ground truth data and the longest tumor diameter determined by tumor flow cytometry (Figure 25). As methodological specificity indicates, the same analysis performed on plasma cfDNA samples from 12 healthy donors showed only PBL-specific signals, with no evidence of epithelial tissue-derived cfDNA or tissue lymphocyte-specific cfDNA (Figure 2A). Our data demonstrate that LiquidMIDOS has the potential to detect both tissue-derived cell-free DNA and cell-free DNA from exhausted tissue lymphocytes, and can accurately distinguish them from the more dominant PBL signal in plasma.

[0210] Next, as part of our sequencing workflow, we investigated whether microbial DNA could be detected in plasma cell-free DNA. To do this, we focused on Staphylococcus aureus, among the most common toxic types of bacteria that cause sepsis. We also focused on Staphylococcus epidermidis, a non-pathogenic pathogen that normally colonizes human skin but can become pathogenic during the immunosuppressive phase of sepsis. Another pathogen that we focused on was adenovirus B, which normally causes the common cold but can be fatal in immunosuppressive situations. Focusing our analysis on these three key causes of primary and secondary sepsis, we analyzed publicly available whole-genome sequencing of human plasma cell-free DNA with spiked and sheared microbial DNA at low concentrations ranging from 32 to 1,000 molecules per microliter of plasma (https: / / www.ncbi.nlm.nih.gov / bioproject / PRJNA507824). The samples were sequenced using NextSeq 500 with an average of 750,000 reads per sample. Next, megaBLAST sequencing was performed. 38 Using this method, sequencing reads were aligned with cell-free DNA from four healthy donors against microbial genomes in the NCBI microbial genome resource. As expected, this revealed that all human plasma samples with low levels of sheared microbial DNA had detectable reads that mapped to those organisms with over 90% identity (Figure 26). In contrast, the four healthy donors without spike-in microbial DNA showed high methodological specificity, with the exception of significantly low levels of Staphylococcus epidermidis (normal skin commensal bacteria) in two of the four healthy donors, and therefore lacked evidence of genomic motifs specific to these microorganisms (Figure 26). These data suggest that our genome-wide method for cell-free DNA analysis can sensitively detect DNA from underlying microorganisms in sepsis, including secondary infections occurring in immunosuppressive situations.

[0211] The inventors can significantly expand upon this initial work to develop a blood-based, integrated sepsis detection and monitoring assay called LiquidMIDOS, which provides clinicians with data on microbial and tissue sources, site of peripheral organ damage, and the degree and timing of T cell dysfunction / fatigue. LiquidMIDOS is clinically useful and serves as a clinician's "Swiss Army knife" for data-driven diagnosis, monitoring, and management of sepsis (Table 1).

[0212] Table 1. How LiquidMIDOS results can be used to answer clinically important questions in sepsis. [Table 5]

[0213] Robustly functioning LiquidMIDOS requires separate input signatures derived from the cell types of interest to the inventors. Therefore, the inventors encode 39 Blueprint 40 , and the NIH Roadmap Epigenomics Project 30 The process begins by analyzing tissue and lymphocyte sources profiled by WGBS in a database. These represent virtually all normal human tissues and leukocyte cell types. The inventors further isolate exhausted T cells from infection-related tissues cryopreserved immediately after death from sepsis patients using fluorescence-activated cell sorting (FACS) (using a schema similar to that in Figure 24). These sorted exhausted T cells are sequenced by WGBS. Using these data (WGBS from multiple tissue sources, normal peripheral blood leukocytes, and T cells present in exhausted tissue from sepsis patients), Metilene is used for DMR analysis. 41 Apply this. Subsequently, cell type-specific methylation reporter profiles are purified using machine learning feature selection methods including random forests and elastic networks, and then CIBERSORTx 37A signature matrix (conceptually similar to Figure 1) is obtained that can deconvolve patient-derived plasma cell-free DNA WGBS data. This identifies promoter regions that are specifically hypomethylated or hypermethylated in each cell / tissue type of interest, namely PDCD1, CTLA4, TIGIT, LAG3, and TIM3 in exhausted T cells. 42 To confirm the biological relevance of cell / tissue-specific reporters identified by machine learning of our signature matrix, we conducted a literature search and used ToppGene Suite. 43 Gene set enrichment analysis will be performed using these specific methylation reporters. These reporters will enable the differentiation and quantification of sepsis-related cell / tissue types from cell-free DNA.

[0214] To determine the sample size required to derive a signature matrix that can effectively distinguish between different categories of cells / tissues, effect sizes had to be estimated, which was done by examining the inventors' data profiling tissue lymphocytes versus epithelial cells versus PBLs in colon cancer patients (Figure 1). This suggests a large effect size with clear distinction between groups based on the methylation status of the most characteristic reporter locations. However, considering the conservative nature and the attempt to distinguish between multiple types of organ tissues (not just common epithelial cells), the inventors estimate a moderate Cohen d effect size of 0.5. 44 This results in n=18 per group to achieve a power of 0.90 at α=0.05. WGBS data from n=18 per cell / tissue type are analyzed to derive a signature matrix for LiquidMIDOS. The large effect size observed in Figure 1 and other studies that searched for far fewer CpG sites (via targeted sequencing or microarrays) than planned by the inventors using WGBS are shown. 1、28、33 Considering the robust ability to distinguish between different human tissue types via methylation-based cell-free DNA analysis, we expect that greater power may be achieved.

[0215] A cohort of two registered blood samples from sepsis patients for training and validation of the LiquidMIDOS method. The inventors collected these samples at the University of Washington over the past five years. Using a standardized protocol, plasma and PBLs were separated from each other, processed, and cryopreserved immediately after collection. To date, the inventors have registered samples from approximately 100 sepsis patients. Almost all sepsis patients in the inventors' bank have a series of plasma and peripheral blood leukocytes collected daily in the ICU, starting from day 1 of hospitalization, along with fully annotated paired clinical and survival data. The inventors have also registered samples from approximately 100 trend-matched non-sepsis controls. Furthermore, the inventors have access to separate, similarly sized, annotated cohorts from Yale Medical Center, which will be used for methodological validation. From both cohorts, the inventors have access to registered autopsy samples from a subset of sepsis patients, which can be used to confirm the microbial etiology of infection, organs involved and damaged by sepsis, and dysfunctional / exhausted T-cell status. Overall, the inventors possess the necessary ground truth data for training and testing LiquidMIDOS (Table 2).

[0216] Table 2. Clinical parameters and details of their ground truth data in training and validation datasets. [Table 6]

[0217] To train LiquidMIDOS, the inventors applied it to cell-free plasma DNA from approximately 100 sepsis patients from the University of Washington, collected daily from day 1 of admission to the ICU. A WGBS was performed on each of these samples, followed by 1) BLAST analysis of the microbial pathogenesis of the infection (human off-target read data against the NCBI microbial database). 38(1) By applying LiquidMIDOS; (2) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (3) By determining the state of immune system dysfunction, LiquidMIDOS analysis is performed to determine the following: (4) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (5) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (6) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (7) By determining which organ tissue is involved in / damaged; (8) By determining which organ tissue is involved in / damaged; (9) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (10) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (11) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (12) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (13) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (14) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (15) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (16) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (17) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (18) By determining which organ tissue source is primarily contributing to plasma cell-free DNA; (19

[0218] Specifically, cell-free DNA is extracted from plasma samples using the QIAamp Circulating Nucleic Acid Kit (Qiagen), and then library preparation is performed using the Accel-NGS Methyl-Seq DNA Library Kit (Swift Biosciences). The samples are barcoded so that they can be sequenced in a multiplex manner using a NovaSeq S4 flow cell (Illumina) targeting a depth of 4050 (approximately 40 samples per flow cell). The inventors apply a standard NGS quality control (QC) filter and then map the sequenced read data to the human genome. Subsequently, the unmapped human read data that has passed QC is processed using BLAST. 38 The data is sorted into the NCBI microbial database (https: / / www.ncbi.nlm.nih.gov / genome / microbes) using this method. The percentage of plasma cfDNA produced by the microorganism is then quantified by dividing the number of read data sorted with the microbial genome by the total number of QC-passed sequenced read data for the sample. 45 Therefore, the microbial content is determined by analyzing the plasma cfDNA of the inventors' training cohort.

[0219] Next, we examine the methylation patterns in the mapped sequencing readout data of humans that have passed QC. Given the case-control nature of our study, it is important to prevent batch effects that could confuse our results. Therefore, we use samtools mpileup. 46 The inventors compare the sequencing depth and fragment size distribution in sepsis patients (cases) versus non-sepsis patients (controls) using the following method. If these systematically differ, filtration and normalization techniques are applied before further analysis, for example, by removing read data with a size greater than 300 base pairs and / or downsampling the mapped read data to the least common denominator. Furthermore, methylation levels of housekeeping genes are systematically compared between case and control samples, and the methylation levels and variances of their promoters are compared. If a persistent batch effect is observed, COMBAT 47 We utilize bioinformatics batch correction strategies such as those described above. This is important to ensure that the differences observed in our case-control study design are not the result of batch effects.

[0220] Next, our LiquidMIDOS-specific signature matrix and CIBERSORTx 37 Using this method, we deconvolve the QC-passed human mapping readouts from cell-free DNA WGBS. To determine the relative abundance of dysfunctional / exhausted T cells for each searched organ / tissue type, we normalize the signal from the dominant PBL and then use CIBERSORTx 37 The relative abundances of these quantities, as output by this process, are quantified.

[0221] Next, machine learning is applied to the cell-free DNA results of our case-versus-control studies to develop a LiquidMIDOS classifier for predicting sepsis from non-sepsis, along with relevant predictive / prognostic metrics. The observed difference is sepsis specificity, as the cohort tends to score-match in a different way. An optimized classifier is developed by applying different machine learning techniques, including Bayesian classification, generalized linear models, k-means classification, logistic regression, support vector machines, random forests, and principal component analysis, with close attention paid to clearly classifying clinically important parameters such as sepsis state, microbial source of infection, organ / tissue site of involvement / injury, and immunosuppressive state (see Table 2). 48 A goodness-of-fit test is also performed to evaluate prognostic accuracy using the Hosmer-Lemeshow test on binary results such as those mentioned above. Following the evaluation of methodological accuracy, the inventors determine which machine learning technique best classifies their training data and utilize this in their final LiquidMIDOS method. The inventors also use sepsis as a primary criterion for comparison, which is the ability to distinguish between sepsis patients and non-sepsis controls. 14 The obtained LiquidMIDOS scores are compared to laboratory tests that clinicians typically use to diagnose and monitor (C-reactive protein levels, white blood cell count, procalcitonin levels, and lactate levels). This is evaluated by testing whether the AUC / C-index is statistically significant and greater than 0.5. The inventors identify the optimal LiquidMIDOS classification results using Ioden's index (and report the relevant sensitivity and specificity). To determine whether the LiquidMIDOS classification score changes over time as expected in Table 1, this is done time-dependently (using the inventors' sequential samples) for each criterion shown in Table 1. Using this training cohort, a high-performance, blood-based, integrated LiquidMIDOS classification and monitoring tool for sepsis is developed.

[0222] While we expect cell-free DNA to be the optimal blood-based analyte for our LiquidMIDOS assay, some embodiments may function better in the PBL compartment. We have shown that cell-free DNA represents the turnover of human cells / tissues from throughout the body. 1、26、27 Exhausted T cells are far more dominant in tissues than in circulation in sepsis-mediated immunosuppression. 20、23 Therefore, this seems unlikely. Furthermore, if some aspects of LiquidMIDOS from the PBL compartment are more sensitive, LiquidMIDOS can be performed from a single blood draw since plasma and PBL are isolated from the same blood tube, although some parts of the workflow need to be repeated (WGBS performed separately on plasma and PBL-derived DNA). In addition, to ensure that our assay is as sensitive as possible, we sequence, deconvolve, and classify the PBL-derived shear DNA using the same workflow as above. Therefore, we investigate whether cell-free DNA is a better analyte for PBLs for LiquidMIDOS and proceed flexibly using the most sensitive analyte depending on the situation.

[0223] Next, the inventors validate LiquidMIDOS by applying it to a provided cohort of approximately 100 septic and 100 non-septic patients from Yale Medical Center. Similar to the training cohort, plasma and PBL were collected daily from septic patients, starting from day 1 of ICU admission. Propensity score matching was performed to ensure that cases and controls were overall matched with respect to clinical and epidemiological covariates other than sepsis-specific factors. Again, a WGBS was performed for each of these samples, focusing on the sample type that showed the best performance in the training exercises above (plasma vs. PBL), and the sequencing deconvolution and LiquidMIDOS-based classification described above were applied to determine 1) the patient's sepsis status, 2) the source of infection, 3) the associated / injured organs, and 4) the suppressed state of the immune system (but using machine learning optimization cutpoints from the inventors' training cohort). The inventors' predictions were correlated over time with ground truth data, including prognosis assessed by 30-day mortality (Table 2). We will also examine whether an increase / decrease in the LiquidMIDOS score correlates with a worse / better prognosis among the different metrics expected in Table 1. Again, using the LiquidMIDOS score cutpoint determined in our training cohort, we will similarly test blood samples from propensity score-matched non-septic patients to verify the specificity of our method. Sepsis 14 This study compares the inventors' ability to predict prognosis and classify sepsis and non-sepsis conditions with laboratory tests routinely ordered by clinicians when diagnosing sepsis and monitoring C-reactive protein levels, white blood cell count, procalcitonin levels, and lactate. LiquidMIDOS can be validated in an independent clinical cohort to demonstrate the inventors' integrated blood-based method for the diagnosis and monitoring of sepsis and to prove its superiority over laboratory tests of standard care.

[0224] As described above, the inventors have access to two well-annotated clinical cohorts (Table 2) and can generate a comprehensive blood-based microbiological and human sequencing repository for sepsis and propensity-matched controls with paired clinical correlation data. Such datasets do not currently exist and would serve as a valuable resource for the scientific community for this and other innovative work.

[0225] Cost-effective methods Based on estimates for library preparation, sequencing, and genomic analysis, the cost of LiquidMIDOS is estimated at $2,000 per assay. As mentioned above, cases of sepsis that were not diagnosed early incurred a significantly higher financial burden of $51,022 per patient compared to $18,023 when sepsis was accurately diagnosed at admission. 17 Delayed diagnosis is associated with increased sepsis severity, longer hospital and ICU stays, and reduced survival rates. Among late-diagnosed sepsis patients (costing $51,022 per patient), LiquidMIDOS continuous monitoring x3 conservatively reduces costs by 25% to a baseline level of $18,023 per patient (+ $6,000 assay costs), saving an average of $2,250 per patient with the use of LiquidMIDOS. As assays become more streamlined in terms of CLIA-accredited and CAP-approved laboratory workflows, NGS costs continue to plummet. 49 Therefore, as the significant cost burden of sepsis in the US healthcare system decreases, further increases in the use of LiquidMIDOS in clinical settings are expected to actually result in cost savings.

[0226] Our method also utilizes the commercially available Karius assay for detecting microorganisms from cell-free DNA. 29 Including the increasing prevalence of genomics-based assays in acute care settings, they can also be used in clinical settings. With the increasing sophistication of molecular pathology laboratories in hospitals, many of which have their own next-generation sequencers, the turnaround time for our assay is 24 hours (Karius).29 (The same next-day turnaround time as the whole-genome sequencing-based microbial cfDNA assay provided by [company name]). This is initially too slow to ensure a point-of-care diagnosis, but should serve as a rapid confirmatory test for diagnosis and an efficient, integrated sepsis monitoring tool. Improved technology increases NGS speed, and as LiquidMIDOS is implemented within a highly streamlined CLIA-accredited and CAP-approved laboratory workflow, we expect even faster turnaround times, with results potentially available within the same timeframe as most other laboratory tests directed in a hospital setting in acute care situations. References [Table 7] TIFF0007874822000034.tif227161 TIFF0007874822000035.tif87161

[0227] Example 6: Cell-free DNA epigenomics to track the dynamics of organ damage and immune exhaustion in sepsis Sepsis is the most common cause of hospital death in the United States and accounts for one-fifth of all deaths worldwide. 2 Sepsis is an immunological challenge. In its initial acute phase, it usually involves an overactive immune response accompanied by an immunomodulatory "cytokine storm," requiring intensive care and potentially leading to death from septic shock or multiple organ failure. 3-5 Even after patients recover from this hyperimmune phase, they enter a weakened immune phase a few days later. This is characterized by exhaustion and dysfunction of T cells, which are crucial cells in the adaptive immune system, and poses a risk of fatal secondary infections for the patient. 3-7 (Figure 20). The majority of these dysfunctional and exhausted T cells are located in visceral tissues. 5 .

[0228] Interestingly, more patients with sepsis survive the initial acute hyperimmune phase than those who survive the later immune exhaustion phase. 3 13% to 30% of sepsis patients develop fatal secondary infections, but these are usually caused by opportunistic microorganisms that would not typically infect someone with a functioning adaptive immune system. 3 ’ 9 ’ 10 Flow cytometry and gene expression analysis of peripheral blood cells showed no differences at the early stage. 10 ’ 11 Therefore, it was necessary to search for the tissue source of exhausted immune cells. 6 In acute care settings, biopsies can be dangerous and difficult to perform, and are rarely performed. To reduce the risk of fatal secondary infection, it is important to non-invasively and accurately determine the T-cell dysfunction / exhaustion phase of sepsis.

[0229] The tissues throughout the body continuously release DNA into the circulating system, and this can be isolated as cell-free DNA (cfDNA). The inventors' approach utilizes this process. 1 ’ 16 ’ 17 Cell turnover and cell death release cell-free DNA into the bloodstream. 18 Technologies based on the latest next-generation sequencing (NGS) methods are being developed, making it possible, for example, to detect low levels of tissue-specific cfDNA, which constitute only about 0.01% of the total cell-free DNA extracted from a single tube of blood. 19 It has been shown that infectious microorganisms release cfDNA that can be measured by NGS, just as tissue cells release cfDNA. 20Furthermore, the inventors hypothesized that cell-free DNA released by dysfunctional / exhausted T cells can be accurately measured by NGS using an improved analytical method, and that it can be distinguished from very common cfDNA derived from peripheral blood leukocytes (Figure 21). In this example, we describe the quantification of cell-free DNA derived from organ-specific tissues and exhausted T cells, and DNA quantification aimed at tracking visceral damage and immune dysfunction / exhaustion, respectively, under sepsis.

[0230] The present inventors' method is based on cell-free DNA epigenomics. The epigenome consists of compounds bound to DNA molecules that instruct which parts of the genome are switched on and off. 21 Each cell type and tissue type has a unique epigenetic signature. 21 These profiles can be profiled by analyzing the methylation patterns on DNA using a method called whole-genome bisulfite sequencing (WGBS). 22 ’ 23 Using these epigenetic signatures, cell-free DNA released by related tissue types / damaged tissue types and dysfunctional / exhausted T cells can be detected through machine learning-based reverse superposition processing.

[0231] Recently published data demonstrates that methylation-based cell-free DNA analysis can be used to detect cancer tissue origins with high sensitivity (from a wide range of different human tissue types). 1 ’ 19 ’ 24 Furthermore, using a simpler methylation microarray approach applied to cfDNA, liver injury has recently been studied. 1 As shown, a wide dynamic range is necessary to determine organ damage of varying degrees, and this must be achieved (Figure 22). Genome-scale cell-free DNA sequencing approaches are used for minimal residual disease. 25 It is possible to achieve a detection sensitivity of 90% for this, and this is the targeted ultra-deep sequencing method.12 ’ 13 ’ 26 Recent literature has shown that this is comparable; the reason is that sequencing depth is lower when using genome-wide approaches, while whole-genome sequencing allows for the achievement of tracking a far greater number of specific reporters. 25 Furthermore, whole-genome sequencing of cell-free DNA can be used to detect infectious microbial species in sepsis with high sensitivity. 20 In other words, genome-scale sequencing of cell-free DNA allows for highly sensitive detection of related / damaged tissues.

[0232] data Specifically, we investigated whether it is possible to distinguish exhausted tissue lymphocytes from tissue-derived epithelial cells and normal peripheral blood leukocytes (PBLs) using a methylation reporter. Therefore, we performed flow cytometry to isolate epithelial cells, PBLs, and tissue lymphocytes from 10 patients with a small number of metastatic colorectal cancers. Focusing on exhausted T cells, we developed a flow cytometry approach to specifically select these cells from tissue before sequencing (Figure 24). Next, after performing WGBS on each sample, we conducted difference in methylation regions (DMR) analysis to identify the 70 most distinct CpG sites (Figure 1). This revealed that epithelial cells, tissue lymphocytes (enriched with dysfunctional / exhausted T cells), and PBLs have different methylation profiles, suggesting that they can be distinguished using WGBS.

[0233] Next, the inventors investigated whether epigenomic signals from epithelial tissue and epigenomic signals from tissue lymphocytes enriched with exhausted T cells could be detected in cell-free DNA. For this purpose, the inventors isolated plasma cell-free DNA from 13 patients with a small number of metastatic colorectal cancers and performed WGBS using Illumina NovaSeq S4 cell flow with a target genome-wide coverage of 4050 genomes. CIBERSORTx 27The above data was reverse-superimposed by searching for epithelial tissue-specific reporters, tissue lymphocyte-specific reporters, and PBL-specific reporters shown in Figure 1. Using this approach, it was possible to detect PBL-derived cfDNA from all patients, and epithelial tissue-derived cfDNA was detected in 9 out of 13 patients, and tissue lymphocyte-derived cfDNA was detected in 9 out of 13 patients (Figure 2A). Furthermore, the levels of epithelial-derived and tissue lymphocyte-derived cfDNA when using our methylated cell-free DNA superimposition approach showed a significant correlation with the levels actually measured by tumor flow cytometry and the sum of the maximum tumor diameters (Figure 25). As an indicator of methodological specificity, the same analysis was performed on cfDNA samples from 12 healthy donors, and only PBL-specific signals were observed; neither epithelial tissue-derived cfDNA nor tissue lymphocyte-specific cfDNA was detected (Figure 2A). Our data show that WGBS can detect both tissue-derived and cell-free DNA derived from exhausted tissue lymphocytes, and that they can be accurately distinguished from the more dominant PBL signal in plasma.

[0234] This research can significantly advance the investigation of the temporal behavior / dynamics of peripheral organ damage, and separately, the investigation of T cell dysfunction / exhaustion in sepsis.

[0235] To ensure the stable functioning of our cell-free DNA-based genome-wide methylation reverse superposition treatment approach, it will be necessary to have an identifiable input signature for the cell species targeted by our inventors that is input to CIBERSORTx. 27 Therefore, WGBS encoding 30 Blueprint 31 and the NIH Roadmap Epigenomics Project 21We begin by analyzing the profiled tissue and lymphocyte sources in the database. These represent almost all human tissue types and leukocyte cell types. Using this data (multiple tissue sources, normal peripheral blood leukocytes, and WGBS of exhausted T cells present in the tissue), we perform methylation region difference analysis using Metilene. 32 Next, we use a machine learning feature selection approach to fine-tune the cell type-specific methylation reporter profile, which includes random forests and elastic networks; this is CIBERSORTx 27 This is to obtain a signature matrix (conceptually similar to Figure 1) that can be used for reverse superimposing WGBS data from patient-derived plasma cell-free DNA. This will allow for the identification of promoter regions that are specifically hypomethylated or hypermethylated in each target cell / tissue type, namely PDCD1, CTLA4, TIGIT, LAG3, and TIM3 in exhausted T cells. 33 These specific methylation reporters make it possible to identify and quantify cell / tissue types associated with sepsis using cell-free DNA.

[0236] To determine the sample size required to derive a signature matrix capable of distinguishing different cell / tissue categories, effect sizes had to be estimated. This was done by examining our data profiling of tissue lymphocytes, epithelial cells, and PBLs in colon cancer patients (Figure 1); this suggests that the methylation status of the 128 most distinctive reporter sites results in a large effect size, allowing for clear differentiation between groups. However, given the need for universality, the assumption that cancer has a fundamentally different etiology from sepsis, and the fact that we are attempting to distinguish between multiple organ tissue types (not just common epithelial cells), we assume a Cohen moderate d-effect size of 0.5. 34This results in a power of 0.90 when n=18 per group and a=0.05. Therefore, the inventors plan to analyze WGBS data with n=18 per cell / tissue type to derive a signature matrix. The effect size shown in Figure 1 is large, and other studies have been conducted that search for fewer CpG sites (by targeted sequencing or microarrays) than what the inventors plan to do with WGBS. 1 ’ 19 ’ 24 Considering that methylation-based cell-free DNA analysis exhibits stable discriminative ability between human tissue types, the inventors anticipate the possibility of achieving higher detection power.

[0237] Next, the inventors used blood samples from sepsis patients stored in two different banks, along with clinical paired data (see Table 2; Example 5). These samples have been collected at the University of Washington over the past five years. Using a standardized protocol, plasma and PBL were separated, processed, and cryopreserved immediately after collection. Barnes Jewish Hospital (University of Washington School of Medicine) is a large medical center with a high volume of cases, allowing for rapid sample collection. The inventors' bank stores daily continuous plasma and peripheral blood leukocytes from almost all sepsis patients, starting from day 1 of hospitalization in the ICU, along with fully annotated clinical and survival paired data. Samples from approximately 100 propensity score-matched non-sepsis controls (IRB#201903142; PI: Aadel Chaudhuri) are also stored in the bank. Overall, this provides the practical data necessary to investigate the dynamics of cell-free DNA from sepsis patients, along with suitable healthy donors.

[0238] The inventors performed WGBS on each of these serial plasma samples collected from sepsis patients and conducted bioinformatics analysis. (1) Determine the relevant / damaged organ by quantitatively examining which visceral tissue is the source of the plasma cell-free DNA; and (2) The state of immune system dysfunction is determined by quantifying cell-free DNA derived from exhausted T cells. To perform this quantification, CIBERSORTX 27 Using this method, human readout data of cell-free DNA mapped by WGBS are desuperimposed with the inventors' proprietary signature matrix to normalize the PBL-derived dominant signal, and then the relative abundance of each visceral tissue type and dysfunctional / exhausted T cell is examined. The inventors' predictions are correlated with actual clinical cohort data (see Table 2; Example 5). Correlation analysis at each time point is planned, which should be possible given the inventors' advanced clinical and laboratory notes, and the temporal trends of tissue-specific and exhausted T cell-specific cell-free DNA that correlate temporal behavior and dynamics with actual clinical practice. Separately from the above, plasma samples obtained from propensity score-matched controls are analyzed to test the specificity of the inventors' approach. K-fold cross-validation is performed to evaluate the generalizability of the inventors' results.

[0239] This analysis models the temporal behavior and dynamics of organ- and tissue-specific cell-free DNA released during the course of sepsis, but the latest literature 1 Since this snapshot is only shown in isolated cases, this modeling represents a significant advance. Furthermore, because the elevation of cell-free DNA derived from dysfunctional / exhausted T cells is expected to progress to secondary infection in a considerable number of patients, we will track the temporal behavior and dynamics of dysfunctional / exhausted T cells. We are confident that our findings will help elucidate the temporal and spatial mechanisms of organ damage and immune exhaustion in sepsis, and will lead to future revitalization of research aimed at improving organ damage and immune exhaustion, which are major causes of sepsis-mediated morbidity and mortality. In addition to advancing scientific understanding, we plan to build a set of clinical pair data and sequencing data, which does not currently exist but will be a valuable resource for the scientific community in the future.

[0240] innovation Below is an outline of a two-pronged attempt to shift the paradigm regarding the dynamics of sepsis through plasma cell-free DNA analysis. Specifically, this involves (1) tracking the dynamics of organ-specific injury and (2) tracking the dynamics of T-cell exhaustion during the course of sepsis. The epigenomic analysis approach of cell-free DNA used by the inventors is novel in both the field of sepsis and, more broadly, in that the deconvolution of cell-free DNA whole-genome bisulfite sequencing data for analyzing organ tissues and exhausted lymphocytes has not been demonstrated to date. The concept of quantifying exhausted lymphocytes from cell-free DNA data is entirely new. Nevertheless, the studies described herein by the inventors are individual snapshots / cases. 1 This approach is supported by existing literature that supports a simpler approach using microarrays, as well as by the inventors' own data.

[0241] The inventors can also create sequencing data of cell-free DNA accompanied by clinical pair correlation data collected sequentially from sepsis patients and propensity score-matched non-sepsis controls. While such data does not currently exist, it would be a valuable resource for the scientific community; it would enable secondary analyses by the inventors' research group and others, further deepening our understanding of the time-series dynamics of cell-free DNA correlated with clinical parameters and outcomes in sepsis. This data resource would represent a significant paradigm shift in the fields of sepsis and cell-free DNA genomics.

[0242] This technology could facilitate the development of non-invasive biomarkers to track patients with sepsis. The results will further clarify how to develop and clinically interpret such biomarkers, leading to a better understanding of when patients with sepsis progress to life-threatening multi-organ failure or when the risk of life-threatening secondary infection increases. Cell-free DNA biomarkers are beginning to be used in the field of sepsis, such as in the Karius assay, and plasma whole-genome sequencing approaches. 20 It is already known that rapid and non-invasive determination of infectious etiologies is possible using [this method]. The field of sepsis is very mature in terms of improved precise diagnostic methods, and this can be supported by the clinical interpretation work described herein.

[0243] The technologies described herein enable the following: (1) Using our new findings on the dynamics of immune exhaustion in sepsis patients, we selectively treat patients with immunotherapy and enhance their adaptive immune system at a precise time that reduces the risk of developing a fatal secondary infection; (2) Using our new findings on the dynamics of organ tissue damage in sepsis, we will selectively introduce specific visceral protective measures to patients with sepsis to reduce the risk of fatal multiple organ failure; (3) To gain a deeper understanding of cell-free DNA epigenomics and to clarify the location of exhausted immune cells, that is, not simply that these cells originate from the periphery, but to clarify exactly which tissues they come from; adding spatial elements to these time-series / dynamic studies of cell-free DNA immunogenomics should deepen our scientific understanding; (4) By performing machine learning, the data from the inventors' human-derived sequencing repository obtained by the above technology will be integrated with clinical parameters to develop a combination biomarker with even higher predictive ability. Ultimately, this technology can be modified to utilize cell-free DNA epigenomics for the purpose of understanding the spatiotemporal multifactorial basis of sepsis; doing so would enable the development of precise biomarkers that improve patient outcomes.

[0244] Furthermore, the research described herein may influence research in multiple different clinical fields. For example, in patients with inflammatory diseases, similar methodologies could be applied to non-invasively track tissue types and immune cell status for the purpose of preventing potential inflammation and determining which organs and tissues are being damaged by that inflammation. In patients undergoing deep wound healing, our research may potentially enable accurate and non-invasive monitoring of this process. Thus, while our research in sepsis is highly influential, it may also have a positive impact on research in other clinical areas. References [Table 8] TIFF0007874822000037.tif230161 TIFF0007874822000038.tif230161 TIFF0007874822000039.tif226162 TIFF0007874822000040.tif69162

Claims

1. An in vitro method for determining cell type or cell state, wherein the method is: (a) To provide a sample containing DNA derived from the tumor microenvironment, including tumor-infiltrating leukocytes (TILs), and to create a methylation profile for said DNA; or To provide a methylation profile for the DNA, Herein, the methylation profile is provided to include the co-associated CpG methylation pattern and / or methylation haplotype block (MHB) of the DNA; (b) Counting the co-associated CpG methylation patterns in the DNA, Here, the co-associated CpG methylation pattern includes two or more CpGs in the DNA; or Counting MHB; (c) Attributing the DNA to a cell type or cell state based on a reference CpG value or reference MHB value, Here, a reference CpG value or reference MHB value is determined and assigned from a reference cell type or reference cell state in order to create a cell type-specific methylation profile; (d) Counting the DNA molecules assigned to each reference CpG value or reference MHB value, Here, each reference CpG value or reference MHB value corresponds to one cell type or cell state; count accordingly; and (e) Determining the proportion of different TIL subsets from the cell type-specific methylation profiles identified in the DNA sample, Methods that include...

2. A method according to claim 1, further comprising counting the methylation profiles of a known single CpG to enhance sensitivity.

3. A method according to claim 1, wherein the reference value is a CpG with different methylation derived from DNA of a known cell species and known cellular state, which may be of bacterial, viral, fungal, or eukaryotic parasite origin.

4. The method according to claim 1, wherein the sample is a blood sample, plasma, urine, saliva, feces, tissue, or biopsy sample.

5. A method according to claim 1, wherein the DNA is cell-free DNA and / or circulating DNA of a rare cell type and is derived from plasma.

6. A method according to claim 1, further comprising providing the method according to claim 1 or a cell state-specific signature of the sample, or, as already provided, the method determining a cell state-specific signature.

7. A method according to claim 1, wherein the sample comprises cell-free DNA (cfDNA).

8. A method according to claim 1, wherein the DNA is cell-free tumor ctDNA.

9. A method according to claim 1, wherein the cellular state measured herein is determined by measuring cell-free DNA circulating in the blood and originates from tumor-infiltrating leukocytes (TILs).

10. A method according to claim 1, further comprising classifying DNA as originating from normal leukocytes, tumor-associated cells, or tumor-infiltrating leukocytes.

11. A method according to claim 1, wherein a methylation profile of DNA in the sample is prepared using a microarray or bisulfite sequencing method.

Images