Method and system for determining personalized treatments - Patent Application 20070122997

Abstract

To provide a method for generating immunity scores.SOLUTION: A method includes the following steps of: (i) determining a qualitative and / or quantitative evaluation of tumor infiltration lymphocytes in a specimen; (ii) determining a qualitative and / or quantitative evaluation of T cell receptor signal transduction in a specimen; (iii) determining a qualitative and / or quantitative evaluation of mutation load in a specimen; and (iv) applying a predictive algorithm to generate an immunity score based on the determined qualitative and / or quantitative evaluation of tumor infiltration lymphocytes, the determined qualitative and / or quantitative evaluation of T-cell receptor signal transduction, and the determined qualitative and / or quantitative evaluation of mutation load.SELECTED DRAWING: Figure 7

Description

[Technical Field]

[0001] The present disclosure relates generally to methods and systems for generating oncology treatment recommendations. [Background technology]

[0002] In patients with melanoma and non-small cell lung cancer (NSCLC), a high rate of somatic mutations—the so-called "mutational burden" (MuB)—and elevated intratumoral expression of checkpoint blockers, including the immunosuppressive molecule CD274 (best known as programmed death-ligand 1 protein (PD-L1)), have been shown to correlate with improved clinical responses to immune checkpoint blocker (ICB)-based immunotherapy. However, predicting the response of patients with other malignancies to immunotherapy—and perhaps combining immunotherapy with other targets for immune checkpoint blockade—requires deeper deconvolution of the immunological tumor microenvironment. Similarly, detailed characterization of the immunological makeup of malignancies may be necessary to aid clinical decision-making, especially for patients who fail standard ICB-based immunotherapy.

[0003] For example, therapeutic antibodies targeting immune checkpoint molecules have been approved by the FDA for the treatment of several types of cancer. However, assessment of tumor checkpoint blockade has been limited to FDA-approved IHC assays measuring the status of the programmed death-ligand 1 (PD-L1) protein, which are subjective and analytically ineffective. As the number of antibodies targeting immune checkpoints increases, assays capable of evaluating additional biomarkers in tumor specimens are needed to accurately predict patient response to these drugs. Summary of the Invention [Problem to be solved by the invention]

[0004] Therefore, there is a need for assays that can characterize the immunological tumor microenvironment to guide therapeutic decisions. [Means for solving the problem]

[0005] The present disclosure relates to an innovative method for determining the susceptibility of cancer cells to one or more personalized tumor treatments. According to one aspect of the present invention is a method for generating an immune score, the method comprising the steps of: (i) determining a qualitative and / or quantitative assessment of tumor infiltrating lymphocytes in a sample; (ii) determining a qualitative and / or quantitative assessment of T cell receptor signaling in the sample; (iii) determining a qualitative and / or quantitative assessment of mutational burden in the sample; and (iv) using a predictive algorithm to generate an immune score based on the determined qualitative and / or quantitative assessment of tumor infiltrating lymphocytes, the determined qualitative and / or quantitative assessment of T cell receptor signaling, and the determined qualitative and / or quantitative assessment of mutational burden.

[0006] According to one embodiment, the method further comprises determining one or more possible therapeutic therapies based on the immune score. According to one embodiment, the predictive algorithm is trained and updated using machine learning.

[0007] According to one aspect of the present invention is a method of analyzing a tumor, the method comprising the steps of: (i) determining a qualitative and / or quantitative assessment of tumor-infiltrating lymphocytes in a sample; (ii) determining a qualitative and / or quantitative assessment of T-cell receptor signaling in the sample; (iii) determining a qualitative and / or quantitative assessment of mutational burden in the sample; and (iv) classifying the sample as a responder to treatment or a non-responder to treatment based on the determined qualitative and / or quantitative assessment of tumor-infiltrating lymphocytes, the determined qualitative and / or quantitative assessment of T-cell receptor signaling, and the determined qualitative and / or quantitative assessment of mutational burden.

[0008] According to one embodiment, the method further comprises determining the response of the tumor to one or more possible therapeutic therapies based on the determined classification. According to one embodiment, the sample is classified as an indeterminate responder to treatment. According to one embodiment, samples classified as non-responders to treatment and non-responders to treatment may be further classified as being at risk for rapid cancer progression.

[0009] According to one embodiment, the method further comprises correlating the determined classification of the sample with a second classification. According to one embodiment, the method further comprises determining the response of the tumor to one or more possible therapeutic therapies based on the correlation.

[0010] According to one embodiment, the method further comprises determining immunohistochemical data of the sample, and said classifying step is further based on the determined immunohistochemical data.

[0011] According to another embodiment is a method for analyzing tumors, the method comprising the steps of: (i) determining a quantitative assessment of the expression levels of at least four genes in a sample; (ii) classifying the sample as a responder to treatment if the determined quantitative assessment of the expression level of a first gene of the four genes exceeds a predetermined threshold, or proceeding to the next step if the determined quantitative assessment of the expression level of the first gene of the four genes does not exceed a predetermined threshold; and (iii) classifying the sample as a non-responder to treatment if the determined quantitative assessment of the expression level of a second gene of the four genes is greater than a predetermined threshold, or proceeding to the next step if the determined quantitative assessment of the expression level of the second gene of the four genes does not exceed a predetermined threshold. (iv) classifying the sample as a responder to treatment if the determined quantitative assessment of the expression level of the third gene of the four genes exceeds a predetermined threshold, or proceeding to a next step if the determined quantitative assessment of the expression level of the third gene of the four genes exceeds a predetermined threshold; and (v) classifying the sample as a non-responder to treatment if the determined quantitative assessment of the expression level of the fourth gene of the four genes exceeds a predetermined threshold, or classifying the sample as a responder to treatment if the determined quantitative assessment of the expression level of the fourth gene of the four genes exceeds a predetermined threshold.

[0012] According to another embodiment is a method for analyzing a tumor, the method comprising the steps of: (i) determining a qualitative and / or quantitative assessment of a plurality of genes associated with immune cell infiltration in the sample; (ii) determining a qualitative and / or quantitative assessment of a plurality of genes associated with T cell activation in the sample; (iii) determining a qualitative and / or quantitative assessment of a plurality of genes associated with cytokine signaling in the sample; (iv) determining a qualitative and / or quantitative assessment of a plurality of genes associated with immune response regulation in the sample; (v) normalizing each of the determined qualitative and / or quantitative assessments; and (vi) proceeding to analyze a normalized qualitative and / or quantitative assessment of T cell activation in the sample if the normalized qualitative and / or quantitative assessment of immune cell infiltration exceeds a predetermined threshold, wherein the sample is deemed to be a tumor if the normalized qualitative and / or quantitative assessment of T cell activation exceeds a predetermined threshold. (vii) if the normalized qualitative and / or quantitative assessment of immune cell infiltration is below the predetermined threshold, proceeding to the next step; (viii) if the normalized qualitative and / or quantitative assessment of the immunomodulatory response is above the predetermined threshold, the sample is classified as a difficult responder, or if the normalized qualitative and / or quantitative assessment of the immunomodulatory response does not exceed the predetermined threshold, proceeding to the next step; and (ix) if the normalized qualitative and / or quantitative assessment of cytokine signaling is above the predetermined threshold, the sample is classified as a difficult responder, or if the normalized qualitative and / or quantitative assessment of cytokine signaling is below the predetermined threshold, the sample is classified as a non-responder.

[0013] According to another aspect is a method for providing a comprehensive immune profiling clinical report to a patient's clinician, the method comprising the steps of: (i) obtaining one or more samples from the patient's tumor; (ii) generating, from the one or more samples, RNA sequencing data comprising information about the expression of a plurality of tumor-infiltrating lymphocyte proteins and a plurality of T-cell receptor signaling proteins; (iii) generating, from the one or more samples, DNA sequencing data comprising mutational burden information for a plurality of genes; (iv) generating, from the one or more samples, immunohistochemistry and fluorescence in situ hybridization (FISH) data to measure patterns of tumor-infiltrating lymphocyte protein expression and copy number gain, and CD3 and CD8, among other proteins; (v) calculating, from the RNA sequencing data, DNA sequencing data, and immunohistochemistry and FISH data, likelihoods of the patient's tumor responding to a plurality of possible treatments; and (vii) providing a comprehensive immune profiling clinical report to the patient's clinician, the report comprising the calculated likelihoods of the patient's tumor responding to a plurality of possible treatments.

[0014] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. [Effects of the Invention]

[0015] As described above, the present invention provides a method for generating an immune score. [Brief explanation of the drawings]

[0016] [Figure 1] 1 is a graph of results for a 54-gene model using a panel of retrospective samples for training, according to one embodiment. [Figure 2] FIG. 1 is a schematic diagram of the feature space of a model utilizing 54 genes and mutation burden (MuB), according to one embodiment. [Figure 3]FIG. 1 is a schematic diagram of a decision tree for a four-gene model, according to one embodiment. [Figure 4] 1 is a graph of results for a four-gene model using a panel of retrospective samples for training, according to one embodiment. [Figure 5] FIG. 1 is a schematic diagram of a decision tree for an immune function model, according to one embodiment. [Figure 6] 1 is a graph of the results of an immune function model using a panel of retrospective samples for training, according to one embodiment. [Figure 7] FIG. 1 is a schematic diagram of Bayesian model averaging (BMA) for final prediction using outputs from a 54-gene model, a 4-gene model, and an immune function model, according to one embodiment. [Figure 8] 1 is a graph of Bayesian model averaging results using a panel of 87 retrospective samples for training, and results from a 54-gene model, a 4-gene model, and an immune function model, according to one embodiment. [Figure 9] 1 is a table of overall results from four different models for determining the sensitivity of cancer cells to one or more personalized tumor treatments, according to one embodiment. [Figure 10] 1 is a flowchart of a method for providing a report to a patient's clinician, according to one embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0017] In the drawings, like reference characters generally refer to the same parts throughout the different views. Also, the drawings are not necessarily to scale, emphasis instead being placed upon illustrating the principles of the invention.

[0018] The present disclosure relates to embodiments of methods and systems for determining the susceptibility of cancer cells to one or more personalized tumor treatments. According to one embodiment, the present disclosure relates to a method for applying a multi-analyte assay (MAAA) with an algorithmic analytical approach to predict therapeutic efficacy.

[0019] Immune Advance (IA) is a next-generation sequencing (NGS) assay designed to provide information about tumor-infiltrating lymphocytes (TILs), mutational burden (MuB), T cell receptor signaling (TCRS), immune-related drug target score (IRDTS), and overall immune activation status (i.e., immune score). While TIL assessment focuses on CD8+ T cells, it provides additional information about other subsets of T cells and related immune effector cells, such as B cells and macrophages. MuB assessment is designed to provide information about the total number of somatic mutations (high vs. low) within a tumor and is not intended to report on specific mutations at the gene level. TCRS, or T cell signaling with both neoplastic cells and other immune-related cells, utilizes expression information from genes that represent both receptors and ligands in the interaction of these various cell types. IRDTS is a set of genes that are direct targets of one or more immunomodulatory agents, such as CTLA4 and ipilimumab. The Immune Score (IS) utilizes information from TIL, MuB, and TCRS to provide an assessment of the activation of the immune status of the examined tumor in the context of a reference database of previous clinical outcomes for patients treated with one or more checkpoint inhibitors (CPIs).

[0020] Immune Advance (IA) is a next-generation sequencing assay that uses RNA-seq to assess mRNA expression of numerous immune-related (IR) genes and multiple expression-regulated genes, and DNA-seq to assess mutation burden in a 1.5 Mb target capture. The RNA-seq component validates expression of the most clinically important genes, while the DNA-seq component reports the number of non-synonymous mutations without reference to the specific genes involved. For example, a representative list of 767 genes that can be utilized in one or more of RNA-seq, DNA-seq, and mutation burden analyses is provided herein as Table 2.

[0021] The RNA-seq component of the IA interrogates a large number of genes, representing over 40 unique gene functions at the highest level. The IA focuses on the most clinically important genes, many of which are directly related to T-cell receptor signaling (TCRS) and several others for classifying tumor-infiltrating lymphocytes (TILs). The clinical validation portion of the IA for all identified clinically important genes compares RNA-seq data from multiple formalin-fixed, paraffin-embedded (FFPE) specimens with custom TaqMan assay results and publicly available information on the same sample sets for whole-transcriptome analysis by the Cancer Genome Atlas project (TCGA).

[0022] Genes associated with TCRS can be expressed on immune infiltrating cells or neoplastic cells and are typically classified as receptor-ligand pairs. For all genes classified as IA as associated with TCRS, either the ligand or the receptor is expressed by one or more subsets of T cells. The current approach to this list of TCRS genes is to divide them into co-stimulatory or co-inhibitory T cell functions, or checkpoint pathways. TCRS-associated genes can be further classified into genes that are direct targets of one or more checkpoint inhibitors and those that are not. The direct target of a checkpoint inhibitor can be either the receptor or the ligand, but not both simultaneously. Examples of checkpoint inhibitors and targets are ipilimumab and CTLA-4, a receptor expressed on activated T cells.

[0023] Genes associated with TILs encompass a wide variety of infiltrating immune cells, and IA employs existing classifications and associated gene expression markers. Classically, "immunoscores" have been reported as prognostic markers in multiple tumor types using three or fewer TIL markers. IA utilizes three or more TIL-associated genes to stratify patients for response to CPIs.

[0024] The DNA sequence component of IA is, for example, a 1.5 Mb AmpliSeq capture of many different cancer-related genes. For IA, specific mutations may not be reported, but rather the number of nonsynonymous mutations. Thus, the output of the DNA sequence component of IA is an assessment of mutation burden.

[0025] Drug-target gene expression analysis focuses on each gene targeted by one or more checkpoint inhibitors and classifies each result as highly relevant, moderately relevant, or low relevant based on a predetermined threshold for the specific drug target. High, moderate, or low relevance for each drug target goes beyond a simple assessment of the reads per million for the target gene and encompasses upstream and downstream effectors for that target gene. For example, the primary function of CTLA-4 signaling is to downregulate T cell activation by counteracting the costimulatory signal delivered by the second receptor, CD28. Both CTLA-4 and CD28 share the same ligands, CD80 (also known as B7.1) and CD86 (also known as B7.2). CTLA-4 has a higher affinity for both of these ligands than CD28 for ligand binding, resulting in an overall co-inhibitory signal when expressed at comparable levels. Clinically, this is counteracted by administering the CTLA-4 checkpoint inhibitor ipilimumab.

[0026] TIL analysis focuses on semiquantitative and / or quantitative assessment of TILs and qualitative and / or quantitative assessment of additional immune effector cell subsets. In the first example, tumor-infiltrating lymphocytes (TILs) are reported as the number of expressed CD3+ and CD8+ transcripts (reads). To aid in the interpretation of TIL expression levels, direct immunohistochemical comparisons of CD3 and CD8 are performed using the Aperio image analysis platform. In the second example, additional qualitative and / or quantitative TIL markers, such as FOXP3 for T-regulatory cells, CD163 and CD68 for macrophages, ICOS and CD28 for T-helper cells, and several other markers of immune cell subsets, are each classified as high, medium, or low relative expression results based on a predetermined transcript read threshold. In contrast to drug-target evaluation, analysis of various additional TIL markers is simpler and reflects a ranking relative to previously observed values.

[0027] Mutation burden analysis focuses on the number of nonsynonymous mutations equivalent to exome sequencing. Results from a 1.5 Mb target capture sequence are provided in the context of exome sequencing to validate the DNA sequence component of IA. These results are presented as high, moderate, or low mutation burden. A mutation burden greater than the equivalent of 200 exonic nonsynonymous mutations is reported as high, a mutation burden less than 200 and greater than 150 is reported as moderate, and a mutation burden less than 150 is reported as low. The cutoff values ​​are arbitrary and reflect a summary of what has been reported in the literature to date regarding clinical response to one or more checkpoint inhibitors. Therefore, many other thresholds are possible.

[0028] To summarize the results of drug-target gene expression, immune cell infiltration, and mutation burden, the IA provides a single immune score ("Immune Score") on a scale of 0 to 100 for a global assessment of immune activation. The IS represents the results of a single patient tested for clinical utility in this validation.

[0029] According to another embodiment, the present disclosure relates to a method and system for generating tumor treatment recommendations using at least three independently developed unique models or approaches that can be utilized for comparison purposes and, optionally, for Bayesian modeling. The models address both machine learning and biological approaches to tumor treatment. According to one embodiment, the at least three models provide similar responses to checkpoint inhibitors, and a Bayesian average model can be utilized to represent the best fit.

[0030] The first model, referred to as the 54-gene model and discussed in more detail below, is a polynomial machine learning regression model. According to one embodiment, the 54-gene model or approach uses 11 genes representing TILs and 43 genes for TCRS in combination with MuB for prediction, although other genes and combinations are possible.

[0031] The second model, referred to as the four-gene model and discussed in more detail below, represents a biological approach at the gene level. According to one embodiment, the four-gene model or approach utilizes a decision tree model to select the best minimal set of TIL or T cell activation genes for prediction.

[0032] The third model, referred to as the immune function model or gene function cluster model and discussed in more detail below, represents a biological approach at the functional level. According to one embodiment, this model or approach utilizes 13 genes representing immune cell infiltration, 23 genes for T cell activation, 10 genes for cytokine signaling, and 8 genes for immune response regulation, although other genes and combinations are possible.

[0033] An optional fourth model is referred to as a primary immune marker model or approach. A primary immune marker model or approach analyzes one or more immune markers. According to one embodiment, the primary immune marker model or approach uses immunohistochemistry to analyze PD-L1 protein expression and / or tumor-infiltrating lymphocyte (TIL) expression (including, but not limited to, CD3 and / or CD8). According to another embodiment, the primary immune marker model or approach utilizes fluorescence in situ hybridization (FISH) methodology to analyze PD-L1 and / or PD-L2 copy number gain.

[0034] According to another embodiment, the method may also optionally utilize PCR analysis or any other methodology to analyze microsatellite instability (MSI), among other possible factors.

[0035] Each of the two approaches is described in more detail below. Approach 1 - IA Multi-Analyte Assay (MAAA) with Algorithmic Analysis According to the first approach to deriving personalized tumor therapy, a multifactorial analysis called Immune Advance (IA) provides qualitative and / or quantitative assessments of tumor-infiltrating lymphocytes (TILs), mutational burden (MuB), and T-cell receptor signaling (TCRS). The three values ​​are used to derive an overall score, called the Immune Score, which is a predictor of tumor response to one or more checkpoint inhibitors (CPIs). By way of example only, the Immune Score can be reported on a scale of 1 to 100 and categorized into three clinically relevant groups, such as high, moderate, or low overall response rate to CPIs, among many other possible reporting mechanisms.

[0036] The IA utilizes proprietary algorithmic analysis to provide qualitative and / or quantitative assessments of tumor-infiltrating lymphocytes (TILs), mutational burden (MuB), and T-cell receptor signaling (TCRS). These three values ​​are then used to derive an overall immune score. All of these analyses utilize a reference database of previous IA results developed through this validation for future evaluation. This reference database is updated, e.g., quarterly, as previous clinical results are added to the existing results. These updates to the reference database will not affect clinical results prior to the time of update. The assessment of TILs, MuB, and TCRS is independent of known clinical outcomes but relies on the observed values ​​of these parameters in the reference database of previous IA results. In contrast, the immune score requires the use of clinical outcomes from the reference database regarding response to current FDA-approved checkpoint inhibitors (CPIs).

[0037] An approach for IA analysis was developed by first analyzing a group of co-expressed TCRS genes discovered using RNA-seq data from a cohort of specimens. The concept of using TILs and MuB was added to the information provided in the peer-reviewed literature regarding response to CPIs and prognostic outcomes in various tumor types. Based on analysis of the peer-reviewed literature and RNA-seq data from the RPCI TCGA cohort, several genes are available for TIL evaluation. This approach shows that TILs and TCRS are highly correlated and co-expressed, whereas MuB is somewhat, but not as, consistent with these two parameters.

[0038] The final approach in IA analysis was to develop a mathematical formula (i.e., algorithm) that could be used to evaluate TIL, MuB, TCRS, and IAS in the context of patient stratification for response to CPIs. The endpoint of this analysis was to normalize each of these parameters to a score of 100 to provide a numerical reference for the results. The algorithm utilizes three distinct analyses or steps in a unique sequence to derive a final patient stratification for response to CPIs.

[0039] Step 1 of the algorithmic analysis is based on ranking or scoring the observed test results against a reference database for TIL, MuB, and TCRS. For the purposes of reporting patient outcomes, the TCRS score is an unreported intermediate value, while the TIL score and MuB score are reported values. The formula used for this ranking or scoring is as follows:

[0040] TIL score means the number of reference samples below the average normalized log2 reads per million of the TIL-identified gene for the test sample / total number of reference samples * 100, rounded to the nearest integer.

[0041] TCRS score means the number of reference samples below the mean normalized log2 reads per million of the TCRS identified genes for the test sample / total number of reference samples * 100, rounded to the nearest integer.

[0042] MuB score means the number of somatic mutations for the test sample less or equal to the number of reference samples / total number of reference samples*100, rounded to the nearest integer. In step 2 of the algorithmic analysis, the scores for TIL, MuB, and TCRS are combined into a single value, the Immune Activation Weighted Score (IAWS), using weighted values ​​for each of these parameters. The formula used for this Immune Activation Weighted Score (IWS) is:

[0043] IWS score = (TIL score x TIL weighted value) + (MuB score x MuB weighted value) + (TCRS score x TCRS weighted value), rounded to the nearest integer.

[0044] The weights for TIL, MuB, and TCRS can be based, for example, on a machine learning approach that trains a classifier to generate optimal IWS (see step 3) for known treatment responders versus non-responders in a reference dataset. The classifier functions such that the IWS of the two classes is best differentiated, such that the weight of TIL + the weight of MuB + the weight of TCR = 1.0. Within the ranked IWS of the reference dataset with the responder / non-responder designation, two thresholds are further determined to call high / low scores, so that all reference samples above the high threshold have a PPV of 95% or higher in predicting responders, and all reference samples below the low threshold have a NPV of 95% or higher in predicting non-responders.

[0045] Optional step 3 of the algorithmic analysis involves converting the IWS to an immunity score (IS) based on a ranking or score of the observed test IWS results. For the purposes of reporting patient outcomes, the IWS is an intermediate value that is not reported, while the IS is a reported value. The formula used for this ranking or score is as follows:

[0046] IS = (number of samples in the reference dataset that are less than or equal to the IWS score of the test sample) / total number of reference samples * 100, rounded to the nearest integer. Although the IA algorithm will remain the same, the reference database of patient responses to the CPI will continue to expand with future clinical trials and follow-up of patient responses. Once a score is assigned to either the qualitative and / or quantitative assessment of TIL, MuB, TCRS, IWS, or IS, that value will not change with future additions to the reference database.

[0047] Approach 2 - Integrating three or more models According to a second approach to deriving personalized tumor therapy, three independent models (a 54-gene model, a 4-gene model, and an immune function model) are utilized individually or in one or more possible combinations. For example, two or more of the three models may be utilized for comparison purposes and / or used for Bayesian or other types of modeling. Each of the three models provides a prediction of tumor response to one or more checkpoint inhibitors, and advanced comparison and modeling techniques may use the output from two or more of the three models to provide an average or overall prediction.

[0048] According to the second approach to deriving personalized tumor therapies, three independent models (a 54-gene model, a 4-gene model, and an immune function model) are utilized individually or in one or more possible combinations. The first model, referred to as the 54-gene model, is a polynomial machine learning regression model that uses 11 genes representing TILs and 43 genes for TCRS in combination with MuB for prediction. The second model, referred to as the 4-gene model, represents a biological approach at the gene level and utilizes a decision tree model to select the best minimal set of TIL or T cell activation genes for prediction. The third model, referred to as the immune function model, represents a biological approach at the functional level and utilizes 13 genes representing immune cell infiltration, 23 genes for T cell activation, 10 genes for cytokine signaling, and 8 genes for immune response regulation. As described below, the three models can be utilized independently and / or jointly to generate or derive personalized tumor therapies.

[0049] According to a further embodiment of the second approach, a fourth model, referred to as the primary immune marker model or approach, is also utilized, which analyzes one or more immune markers using immunohistochemistry as described herein or otherwise contemplated.

[0050] Model 1-54 Gene Model According to one embodiment, the 54-gene model is a polynomial machine learning regression model that uses 11 genes representing TILs and 43 genes for TCRS in combination with MuB for prediction, although other gene combinations are possible.

[0051] The 54-gene model was derived by benchmarking different combinations of training and test data sizes based on multiple retrospective samples with known treatment regimens and tumor responses. For a training size of N samples, 20,000 iterations were trained using N randomly drawn samples each, and the remaining ((total number of retrospective samples) - N) test samples were used to evaluate the classifier's performance.

[0052] The overall performance of the classifier for N training samples was calculated from the average of 20,000 benchmark runs. From internal benchmarks, we observed that the ROC / AUC performance of the classifier converged when N reached approximately 50% of the total initial retrospective population. The published performance metrics (ROC plots, AUC scores, PPV, NPV, etc.) were calculated using a more explicit leave-one-out test, i.e., 87 iterations of unique N = 86 tests, where each iteration uses 86 samples for training while leaving one unique sample for testing purposes. However, the final predictive model for future testing purposes used all retrospective samples for training, which totaled 87 samples in one experimental study.

[0053] Referring to Figure 1, in one embodiment, there is a graph of results for a 54-gene model using a panel of retrospective samples for training. As shown in the graph, the results have a positive predictive value (PPV) of 96% for 26% of the population, a negative predictive value (NPV) of 90% for 49% of the population, and a difficult-to-distinguish group representing 25% of the population.

[0054] According to one embodiment not shown in Figure 1, a 54-gene model can be determined or designed to classify tumors as responders, difficult responders, or non-responders to treatment, and can be determined or designed to predict the risk of hyper-progression of cancer. For example, a 54-gene model can be determined or designed to identify a subset of non-responders at risk for hyper-progressive disease. The risk of hyper-progressive disease can be qualitative and / or quantitative.

[0055] According to one embodiment, the 54-gene model utilizes expression information for 54 different genes and / or mutation burden as the sum of the number of mutations in 409 genes. Table 1 contains a representative list of genes for which expression information can be utilized in the 54-gene model. However, many other genes are possible for expression analysis in this model, including but not limited to the genes identified in Table 2.

[0056] [Table 1]

[0057] Referring to Figure 2, there is shown a schematic representation of the feature space of a model utilizing 54 genes or 54 features, in one embodiment, combined with mutational burden (MuB) as an additional feature. According to one embodiment, MuB as an individual feature is treated equivalently to 54 genes or 54 features.

[0058] In addition to listing genes that may be suitable for the 54-gene model, Table 2 also provides genes for which expression information can be utilized in the immune function models described herein. Thus, the genes listed in Table 2 can be utilized in one or more of RNA-seq, DNA-seq, and mutational burden analyses.

[0059] [Table 2-1]

[0060] [Table 2-2]

[0061] [Table 2-3]

[0062] [Table 2-4]

[0063] [Table 2-5]

[0064] Model 2-4 Gene Model According to one embodiment, the four-gene model utilizes a decision tree model to select the best minimal set of TIL or T cell activation genes for prediction. The four-gene model was derived using a decision tree machine learning approach independent of an initial selection of, for example, 54 genes (although other genes were possible). The machine learning algorithm selected a subset of genes to construct a human-interpretable decision tree that best distinguished responders from non-responders from the entire population of 87 training samples. The four automatically selected genes include two genes associated with T cell activation, a gene associated with immune response regulation, and a gene associated with cytokine signaling.

[0065] Referring to Figure 3, in one embodiment, a schematic diagram of a decision tree for a four-gene model is shown. In the four-gene model, the identities and specific cutoff values ​​of the four genes utilized can vary. According to one embodiment, the four genes are PD-L1, TGFB1, TBX21, and BTLA, with the cutoff values ​​for each of these genes being 171, 2043, 70.56, and 26.38, respectively. According to a further embodiment, the first gene is PD-L1, the second gene is TGFB1, the third gene is TBX21, and the fourth gene is BTLA. The decision tree utilizes the expression level (in normalized reads per million (nRPM)) of each gene to provide a series of YES or NO decisions. For example, if the nRPM of gene 1 is below a particular threshold or cutoff, the tumor is determined to be a responder. If the nRPM of gene 1 is greater than the threshold or cutoff, the decision tree proceeds to the next gene. The outcome of the decision tree classifies tumors as either responders or non-responders, as shown in Figure 3.

[0066] According to one embodiment not shown in Figure 3, a four-gene model can be determined or designed to classify tumors as responders or non-responders to treatment, and can be determined or designed to predict the risk of rapid cancer progression. For example, a four-gene model can be determined or designed to identify a subset of non-responders at risk for aggressive disease. The risk of aggressive disease can be qualitative and / or quantitative.

[0067] Referring to Figure 4, in one embodiment, there is a graph of the results for a four-gene model using a panel of 87 retrospective samples for training. As shown in the graph, the results had a PPV of 72% for 43% of the population, an NPV of 92% for 49% of the population, and no difficult groups.

[0068] Model 3 - Immune Function Model According to one embodiment, the immune function model utilizes 13 genes representing immune cell infiltration, 23 genes for T cell activation, 10 genes for cytokine signaling, and 8 genes for immune response regulation, although other genes are possible. The immune function model, like the four-gene model, utilizes a decision tree learning method. However, instead of assessing the predictive importance of individual genes, the immune function model takes a weighted average relative rank of multiple genes in a given immune function group, including immune cell infiltration, immune response regulation, T cell activation, and cytokine signaling. The relative rank is established by ranking the normalized expression value (nRPM) of a gene against that of a reference population and further normalizing the rank to the same range from 0 to 100. The immune function relative rank collectively reflects the degree of expression of multiple genes with the same function compared to the reference population.

[0069] Referring to Figure 5, in one embodiment, a schematic diagram of a decision tree for the immune function model is shown. The identity of the gene set can vary. The decision tree utilizes relative rank cutoff values ​​for each of four different immune function groups (immune cell infiltration, immune response regulation, T cell activation, and cytokine signaling) to provide a series of YES or NO decisions and can classify tumors as responders to treatment, non-responders to treatment, or difficult-to-distinguish responders to treatment.

[0070] According to one embodiment not shown in FIG. 5 , the immune function model can be determined or designed to classify tumors as responders, difficult responders, or non-responders to treatment, and can also be determined or designed to predict the risk of rapid cancer progression. According to one embodiment, the four functions are immune cell infiltration, immune response regulation, T cell activation, and cytokine signaling, with cutoff values ​​of 58.9, 60.25, 42.98, and 69.78, respectively. For example, the immune function model can be determined or designed to identify a subset of non-responders at risk for rapidly progressive disease. The risk of rapidly progressive disease can be qualitative and / or quantitative.

[0071] Referring to Figure 6, in one embodiment, there is a graph of results for an immune function model using a panel of 87 retrospective samples for training. As shown in the graph, the results have a PPV of 72% for 21% of the population, an NPV of 85% for 55% of the population, and a difficult-to-distinguish group representing 24% of the population.

[0072] According to one embodiment, the immune function model utilizes expression information for 54 genes (13 genes representing immune cell infiltration, 23 genes for T cell activation, 10 genes for cytokine signaling, and 8 genes for immune response regulation), although many other gene combinations are possible, including but not limited to the genes identified in Table 2. Table 3 contains a list of genes for which expression information can be utilized in the immune function model.

[0073] [Table 3]

[0074] Model 4 - Primary immune marker model The primary immune marker model or approach is optional and utilizes immunohistochemistry to analyze one or more immune markers. According to one embodiment, the primary immune marker model or approach utilizes immunohistochemistry to analyze PD-L1 protein expression and / or tumor-infiltrating lymphocyte (TIL) expression (including but not limited to CD3 and / or CD8). According to another embodiment, the primary immune marker model or approach utilizes fluorescence in situ hybridization (FISH) methodology to analyze PD-L1 and / or PD-L2 copy number gain.

[0075] Multiple Model Correlation According to one embodiment, the outputs from two or more of the 54-gene model, the 4-gene model, and the immune function model are combined to provide a final recommendation for personalized tumor therapy.

[0076] For example, referring to Figure 7, in one embodiment, the outputs from the 54-gene model, the 4-gene model, and the immune function model were combined using Bayesian model averaging (BMA) for the final prediction. According to one embodiment, the BMA algorithm is similar to the concept of majority voting, but the algorithm also utilizes the prior probability distribution of the performance of each individual model to optimize the final prediction.

[0077] According to a further embodiment as shown in Figure 7, the outputs from the 54-gene model, the 4-gene model, the immune function model, and the primary immune marker model were combined using BMA for the final prediction.

[0078] Referring to Figure 8, in one embodiment, there is a graph of results for a BMA using a panel of 87 retrospective samples for training, as well as results from a 54-gene model, a 4-gene model, and an immune function model. As shown in the graph, the results have a PPV of 96% for 30% of the population, an NPV of 90% for 70% of the population, and no difficult groups.

[0079] Referring to FIG. 9, in one embodiment, there is a table of overall results from each of the 54-gene model, the 4-gene model, the immune function model, and Bayesian model averaging (BMA).

[0080] The various approaches and models described herein or otherwise contemplated utilize retrospective training panels and machine learning to derive a 54-gene model, a 4-gene model, and an immune function model. As the various approaches and models utilize larger and / or different retrospective training panels, the various genes and models may vary.

[0081] Immunity Report Card 10 , a method 100 for generating a report including the likelihood that a patient's tumor microenvironment will respond to immunotherapy is shown. According to one embodiment, the report also includes one or more personalized treatment options based on the patient's comprehensive immune profile. According to one embodiment, the method utilizes three or more of five data inputs to generate the data necessary to determine the likelihood(s) and personalized treatment options. According to one embodiment, these five data inputs can include at least the following:

[0082] 1. RNA-seq to measure relative transcript levels of genes associated with tumor-infiltrating lymphocytes (TILs) and T-cell receptor signaling (TCRS) genes associated with anti-cancer immune responses and immunotherapy targets.

[0083] 2. DNA sequencing to estimate mutation burden (MUB). 3. Immunohistochemistry (IHC) to measure the expression of PD-L1 protein and the pattern of expression (CD3 and CD8) on tumor-infiltrating lymphocytes (TILS).

[0084] 4. PCR to assess microsatellite instability (MSI), and / or 5. Fluorescence in situ hybridization (FISH) to detect PD-L1 / L2 copy number gain.

[0085] Thus, in step 110 of the method, one or more tumor samples or specimens are collected. The samples may be collected using any method now known or developed in the future. The samples may be obtained directly from the patient and / or tumor, or may be derived from a sample previously obtained from the patient and / or tumor. The samples may be tumor samples or non-tumor samples obtained from an individual. The samples may be analyzed immediately and / or stored for future analysis. Thus, the samples may be shipped, stored, and / or processed for other current or future uses.

[0086] In step 120 of the method, RNA-seq data input is obtained. According to one embodiment, the RNA-seq data input comprises a next-generation sequencing (NGS) assay that interrogates 395 genes representing immune-related gene function using amplicon-based targeted NGS for digital gene expression detection. One embodiment focuses on 54 of these 395 genes, with an additional 10 genes used as controls. In this or other embodiments, one or more additional genes (such as those listed in one or more tables herein) can be analyzed and provided in the results of the method.

[0087] According to one embodiment, the TIL gene component of the RNA-seq data input includes genes identified to classify subsets of infiltrating immune cells. The genes may be the 11 genes identified herein (CD163, CD2, CD3D, CD3E, CD3G, CD4, CD68, CD8A, CD8B, FOXP3, and CD20) or may include different and / or additional genes. The application of TILs as predictive immune biomarkers or prognostic markers of survival has been investigated in a wide variety of tumor types. Evidence indicates that CD3+ CD8+ cytotoxic lymphocytes are positively associated with response to CPIs, whereas FOXP3+ Tregs are negatively associated. The pattern of TILs in tumors has also been shown to be important.

[0088] According to one embodiment, the T cell receptor signaling (TCRS) component of the RNA-seq data input includes genes expressed on immune infiltrating cells, neoplastic cells, or other cells of the tumor microenvironment, and are immunophenotypically classified as directly involved in checkpoint blockade or as involved in other functions related to the adaptive immune response. The genes may be the 43 genes identified herein (see, e.g., Table 1), or may include different and / or additional genes.

[0089] According to one embodiment, immune phenotypes associated with checkpoint blockade are typically classified as receptors and associated ligands. For subsets of TCR-associated genes, either ligands or receptors are expressed by one or more T cell subsets. Genes associated with checkpoint blockade can be further classified as those that are direct targets of one or more checkpoint inhibitors and those that are not. A direct target of a checkpoint inhibitor can be either a receptor or a ligand, but not both simultaneously. Examples of checkpoint inhibitors and targets are ipilimumab and the receptor CTLA-4 expressed on activated T cells. According to one embodiment, potential reports and personalized treatment options also report genes associated with checkpoint blockade as checkpoint blockade (PD-1, CTLA-4), checkpoint blockade (other), or primed T cells (T-cell primed), as shown in Table 3. Checkpoint blockade and checkpoint blockade (other) are associated with co-inhibitory signaling to effector T cells, while primed T cells are co-stimulatory. Other immune phenotypes associated with the adaptive immune response that indirectly influence TCRS include myelosuppression, pro-inflammatory responses, anti-inflammatory responses, and metabolic immune evasion.

[0090] [Table 4]

[0091] The expression of each gene is compared to a reference population and normalized to a value between 1 and 100, referred to as the relative rank. According to one embodiment, the reference population on which this method is based consisted of RNA-seq results from 167 unique tumors. The top 95th percentile of scores (or relative ranks), with values ​​of 95 or greater, is interpreted as very high expression, and the 85th to 94th percentile is interpreted as high expression. The bottom 50th percentile of scores (or relative ranks), with values ​​less than 49, is considered low or very low expression. A score of 50 to 85 is considered moderate expression. The interpretation of the immunophenotype is derived from the average of all genes in that class, ranked as normalized values ​​in the same manner as the expression of individual genes.

[0092] In step 130 of the method, DNA sequence data input is obtained. According to one embodiment, the DNA sequence data input includes 1.75 Mb AmpliSeq® capture of 409 cancer genes with complete exon coverage, assessing a total of 6,602 exons covering 1,165,294 base pairs of unique exonic DNA in a whole-exon mutation profiling assay. Mutation burden (MUB) is reported as the number of mutations per megabase (Mb) of exonic DNA. MuB can be calibrated against a subset of samples using whole-exome sequencing for the development of a variant calling pipeline that provides 20x coverage of over 90% of the unique exonic DNA in the IRC panel. According to one embodiment, MuB was calibrated against four clinically relevant peer-reviewed publications reporting a correlation of high mutation burden with response to checkpoint inhibitors in melanoma. Using comparable whole exomes and these four references as calibrators, a cutoff value for MuB, measured by the number of mutations per Mb of DNA, was established using an internal reference population of 167 patients. In this context, MuB was classified as "very high," "high," "moderate," "low," and "very low." While high and very high MuB classifications are likely to be responders as single biomarkers, this measure lacks sensitivity and specificity and should not be used independently of other assay results in IRC.

[0093] In step 140 of the method, immunohistochemistry data input is obtained. According to one embodiment, the immunohistochemistry data input includes measurements of PD-L1 protein expression and tumor-infiltrating lymphocyte (TILS) expression (CD3 and CD8). According to one embodiment, the immunohistochemistry data input is obtained utilizing an automated DAKO platform and commercially available antibodies to provide expression data for PD-L1, CD3, and CD8. The method reports protein expression patterns for all three analytes to better elucidate the multidimensional interactions occurring in the tumor microenvironment, as well as semi-quantitative measurements of expression for PD-L1.

[0094] According to one embodiment, for melanoma, PD-L1 is measured using the PD-L1 IHC 28-8 FDA-approved assay, following scoring guidelines for reporting the percentage of neoplastic cells that exhibit membrane staining of any intensity. The PD-L1 22 C3 FDA-approved assay is used to test non-small cell lung cancer and other tumor types, and PD-L1 protein expression is determined using the Tumor Proportion Score (TPS), which is the percentage of viable tumor cells that exhibit partial or complete membrane staining of any intensity.

[0095] According to one embodiment, the TILS expression pattern (CD3 and CD8 as measured by IHC) is reported as "infiltrating," "non-infiltrating," or "minimal to absent." An "infiltrating" pattern refers to TIL staining within the group of neoplastic cells in the majority of the tumor examined. "Non-infiltrating" refers to TILs that are present but inconsistent with the pattern of infiltrating groups of neoplastic cells in the majority of the tumor examined. A non-infiltrating pattern of CD3 and CD8 staining includes cases where there are abundant TILs at the leading edge of the tumor but no infiltration of neoplastic cells. A minimal to absent pattern refers to essentially minimal to zero TILs present in any part of the tumor.

[0096] According to one embodiment, CD3 highlights T cells (called TILs in this assay) and is useful for identifying T cell populations associated with neoplasia. CD8 highlights cytotoxic T cells, which (when found in the midst of neoplastic cells) tend to reflect a response to checkpoint inhibitors (CPIs). The information provided in this report may be used by physicians to determine whether immunotherapy with one or more FDA-approved checkpoint inhibitors would be beneficial for this patient.

[0097] According to one embodiment, the PCR component of the method uses five markers to detect microsatellite instability (MSI), including two mononucleotide repeat markers (BAT-25, BAT-26) and three dinucleotide repeat markers (D2S123, D5S346, and D17S250). According to another embodiment, the NGS component of the method uses up to 100 homopolymer, dinucleotide, trinucleotide, and / or tetranucleotide markers to detect microsatellite instability (MSI). In either case, the results are reported as "MSI-high," "MSI-low," or "MSS" (microsatellite stable). MSI, commonly used as a prognostic marker for survival in the setting of Lynch syndrome, also known as hereditary nonpolyposis colorectal cancer (HNPCC), is an FDA-recognized marker of response to checkpoint inhibitors in colorectal cancer with no other treatment options and second-line therapy in solid tumors.

[0098] Colorectal cancer, endometrial cancer, and other types of neoplasms with MSI-H may be sporadic (i.e., microsatellite instability is found only in the neoplasm and therefore not part of an inherited disease) or secondary to Lynch syndrome (i.e., a familial genetic mutation in a DNA repair gene, typically associated with MLH1, PMS2, MSH2, or MSH6). While MSI testing can determine the presence of microsatellite instability, it cannot determine which specific DNA repair gene is affected. Immunohistochemistry for MLH1, PMS2, MSH2, or MSH6 can be used to identify the specific protein affected. Depending on the pattern of expression loss and the carcinoma tissue of origin, slightly different strategies are indicated. Specifically, we recommend reviewing the National Comprehensive Cancer Network (NCCN) guidelines. The presence of Lynch syndrome is reliably determined when a pathogenic DNA repair gene mutation is found in the patient's non-neoplastic tissue. In such cases, genetic counseling is recommended, involving family members at risk for Lynch syndrome. MSI-H colorectal cancer (sporadic and Lynch syndrome-associated) is known to have a different clinical profile (prognosis and chemotherapy response) than MSS (microsatellite stable) colorectal cancer.

[0099] According to one embodiment, the FISH component of the method measures the copy number of PD-L1 (CD274) and PD-L2 (PDCD1LG2), two PD-1 ligands that, when amplified, correlate with PD-L1 expression. The two genes are located 40 kb apart on 9p24.1 and are detected using a pool of fluorescently labeled BAC clones mapping to the gene region (RP11-635N21, RP11-812M23, and RP11-485M14). Results are reported using the ASCO-CAP HER2 testing guideline recommendations for determining copy number as amplified, equivocal, or unamplified. An amplified result is defined as a PD-L1 / 2 to CEP9 ratio of 2.0 or greater for any PD-L1 / 2 copy number value, or a PD-L1 / 2 to CEP9 ratio of less than 2.0 and a PD-L1 / 2 copy number value of 6.0 or greater. An equivocal result occurs when the PD-L1 / 2 to CEP9 ratio is <2.0 and the PD-L1 / 2 copy number value is ≥4.0 but <6.0. A non-amplified result occurs when the PD-L1 / 2 to CEP9 ratio is <2.0 and the PD-L1 / 2 copy number value is ≥4.0. The IRC will use the previous NYS-CLEP-approved OmniSeq "PD-L1 and PD-L2 Amplification and Validation Data SOP" for PD-L1 / 2 copy number testing.

[0100] In step 150 of the method, the input data from one or more of steps 120, 130, and 140 are collated and analyzed. For example, according to one embodiment, the outputs are combined to provide a final recommendation for personalized tumor treatment. For example, according to one embodiment, the outputs are combined using Bayesian model averaging (BMA) for the final prediction. According to one embodiment, the BMA algorithm is similar to the concept of majority voting, but the algorithm also utilizes the performance prior probability distribution of each individual model to optimize the final prediction. However, many other methods for collating data from two or more data inputs are possible.

[0101] In step 160 of the method, a clinical report is provided to the patient's clinician. The report includes at least the likelihood that the patient's tumor microenvironment will respond to a particular immunotherapy, and / or combination immunotherapy, and / or immunotherapy clinical trial. According to one embodiment, in step 170 of the method, the data entry, collated data, and / or final analysis are utilized to generate one or more personalized treatment options based on the patient's comprehensive immune profile. Thus, the report provided to the patient may also include one or more personalized treatment options.

[0102] According to one embodiment, the report provided to the patient's clinician may include one or more of the following: Priority Immune Markers - At-a-glance information on clinically relevant immune markers for treatment management, including but not limited to PD-L1 expression, tumor-infiltrating lymphocyte (TILS) pattern and expression, microsatellite instability (MSI) mutational burden (MUB), and PD-L1 / L2 copy number gain. Evidence of therapeutic relevance for each marker can be provided in the context of the tumor type tested.

[0103] Abstract interpretation - A pathologist's interpretation and characterization of the overall immune status of this tumor based on both preferential immune markers and TCRS gene expression. The abstract interpretation can provide an overview of the assertion of likely response to FDA-approved checkpoint inhibitors and clinical trial opportunity trials based on immunophenotypic assessment.

[0104] Immunophenotype Summary - A summary of gene expression for several immunophenotypes, showing highly expressed genes within each applicable phenotype. The genes associated with each immunophenotype are not intended to include all genes with similar or identical effects, but rather genes whose expression has been rigorously validated. Several genes within a given immunophenotype are related immunotherapeutic agents available as either FDA-approved treatments or in clinical trials.

[0105] Immunophenotyping details - Gene-level expression rank and interpretation of TCRS genes by immunophenotype. Expression of each gene is provided as a rank and interpretation as previously described in the RNA-seq component. Gene-associated immunotherapies are listed regardless of tumor type tested. With the exception of PD-L1, there may be limited evidence supporting overexpression of any immunophenotyping gene and response to any associated immunotherapy.

[0106] Tumor-infiltrating lymphocytes - Gene-level expression rank and interpretation of TILS genes associated with the differentiation of various immune-related cells. Clinical Trials - Clinical trials are displayed for overexpressed markers ranked "high" or "very high" for the tumor type tested and therapies in clinical development. If no markers are ranked "high" or "very high," clinical trials are displayed for markers ranked "moderately high." Clinical trials on lymphocyte recruitment to tumors are displayed whenever the tumor is considered non-inflamed.

[0107] Example - Analysis of immune responses in solid tumors According to one embodiment, immune responses in formalin-fixed, paraffin-embedded (FFPE) tumor specimens were analyzed utilizing methodologies described herein or otherwise contemplated to provide a characterization of the immunological tumor microenvironment to guide therapeutic decisions for patients with solid tumors.

[0108] As described herein, the analysis utilized RNA-seq data to semi-quantitatively measure levels of transcripts associated with anti-cancer immune responses and levels of transcripts reflecting the relative abundance of tumor-infiltrating lymphocytes (TILs), as well as DNA-seq data to estimate mutation burden. Although not described in this example, the analysis may include primary marker immune data as described or otherwise contemplated herein.

[0109] An embodiment of a methodology for obtaining and / or analyzing RNA sequence data and DNA sequence data is described below. However, it is understood that this is only one possible embodiment and is therefore by no means limiting. Other methodologies for obtaining and / or analyzing RNA sequence data and DNA sequence data are possible.

[0110] Unlike existing mutational profiling assays, this assay uses a broad range of biomarkers to accurately match patients to immunotherapeutic treatments based on the immunological makeup of their tumors.

[0111] method To evaluate the analytical performance of the assay, we obtained a subset of 167 FFPE specimens and matched fresh-frozen (FF) tissues from patients with NSCLC, melanoma, renal cell carcinoma, head and neck squamous cell carcinoma (HNSCC), and bladder cancer. Specimens were collected under the institution's banking policy with patient consent, and the study followed the institution's policy for non-human subjects research and was approved by Internal Review Board review (IRB protocol number BDR073116).

[0112] Specimens included fine-needle aspiration biopsies, punch biopsies, needle core biopsies, incisional biopsies, excision biopsies, and resection specimens from 2002 to 2016. For a subset of specimens, whole-exome sequencing and whole-transcriptome RNA-seq data were available as part of The Cancer Genome Atlas (TCGA) project for comparative purposes. Four human cell lines, lymphoblastoid GM12878 cells (ATCC, Manassas, VA), colorectal carcinoma KM-12 cells (NCI-Frederick Cancer DCTD, Bethesda, MD), NSCLC HCC-78 cells (DSMZ, Braunschweig, Germany), and large cell lymphoma SU-DHL-1 cells (ATCC) processed as FFPE blocks, were also used for development and as internal study controls.

[0113] A board-certified anatomic pathologist examined hematoxylin and eosin (H&E)-stained tumor sections to identify the area to be examined. The tumor surface area on H&E-stained sections was ≥ 2 mm per slide. 2 The tumor cellularity was ≥50%, and necrosis was ≤50%. To test for potential preanalytical interference, non-malignant and necrotic tissues were also macroscopically dissected from the corresponding unstained slides for nucleic acid extraction. The areas identified by the pathologist were used as a guide to scrape tissue from three to five unstained slides. Genomic DNA and total RNA were simultaneously extracted from this material using the turXTRAC FFPE Extraction Kit (Covaris, Inc., Woburn, MA) according to the manufacturer's instructions with some modifications. After purification, RNA and DNA were eluted in 30 and 50 μL of water, respectively, and yields were determined by the Quant-iT RNA HS Assay and the Quantifiler Human DNA Quantification Kit (both Thermo Fisher Scientific, Waltham, MA) according to the manufacturer's recommendations. Predefined yields of 10 ng RNA and 30 ng DNA were used as acceptance criteria to ensure proper library preparation.

[0114] Run controls were established and used for library preparation, enrichment, and NGS. They included both positive (MuB-DNA, GEX-RNA) and negative (MuB-DNA negative, GEX-RNA negative) controls, as well as a no-template control (NTC, water). The positive controls provide templates for all targets for qualification and downstream normalization purposes, while the negative controls monitor assay specificity. For RNA-seq, NTCs are used to identify the detection limit at the individual sample level. For DNA-seq, NTCs are used to identify false-positive thresholds at the run level. The performance characteristics of these five run controls were evaluated over multiple weeks to develop thresholds and filters that serve as daily QC parameters for runs and samples.

[0115] The assays utilized the Oncomine Immune Response Research Assay (OIRRA) for GEX and the Comprehensive Cancer Panel (CCP) (Thermo Fisher Scientific) for MuB. Both panels use multiplexed gene-specific primer pairs and next-generation sequencing to amplify nucleic acids extracted from FFPE slides. OIRRA was adapted to quantify the expression of 54 target genes using 10 constitutively expressed housekeeping (HK) genes as normalizers. All 409 cancer-related genes included in the CCP panel were used to estimate MuB from genomic DNA.

[0116] OIRRA libraries were prepared using Ion AmpliSeq targeted sequencing technology (Thermo Fisher Scientific). Briefly, 10 ng of RNA was reverse transcribed to cDNA, and the target was amplified with a multiplex primer pool. For DNA sequencing, 30 ng of DNA was used to prepare CCP libraries. Barcode adapters were ligated to the partially digested amplicons, purified, and normalized to 50 pM. Up to 16 equimolar RNA and DNA libraries were pooled prior to enrichment and template preparation using the Ion Chef system (Thermo Fisher Scientific). 200-bp sequencing was performed on an Ion S5XL540 chip, yielding 1.5–2.5M RNA-seq mapping reads per sample and an average DNA sequencing depth of 100–150X.

[0117] The accuracy of RNA-seq was assessed by comparison with qRT-PCR for all reported genes for all samples, and by comparison with IHC for CD8A (based on automated image analysis for a subset of 57 samples containing suitable material). For orthogonal qRT-PCR analysis, 100 ng of RNA was reverse transcribed, amplified, and measured in triplicate by TaqMan® gene expression assays monitoring 54 targets and 10 HK genes using a QuantStudio7 Real-Time PCR System (Applied BioSystems, Foster City, CA). For IHC, 5-μm-thick sections from tissue microarrays (TMAs) with three 0.6-mm tissue cores arranged per tumor were stained with antibodies specific for CD8A (C8 / 144B, Dako, Agilent Technologies, Santa Clara, CA) according to standard procedures, and PD-L1 was stained specifically for tumor histology (22C3, Dako, or 28-8, Dako, or SP142, Ventana) according to FDA guidelines. Quantitative IHC data for CD8A+ T cell counts were obtained using 20x bright-field light microscopy with an Aperio Scanscope (Aperio Technologies, Inc., Vista, CA). Images were analyzed using eSlide Manager v12.2.1 (Aperio Technologies), and the number of positive cells per square millimeter (sq mm) was counted for each TMA core. Quantitative IHC expression data for PD-L1 were obtained by a pathologist trained in interpreting the tumor proportion score (TPS), H score (HS), and modified H score (MHS). At least two of three evaluable cores were required for inclusion in the final analysis. The mean number of CD8A+ cells per square millimeter, TPS, HS, and MHS were derived from each sample during individual analysis of at least two cores.

[0118] Sequencing data were initially processed using Torrent Suite software (v5.2.0) for reference mapping and base calling, during which validation-defined QC specifications for mapped reads, on-target reads, mean read length, mean depth, uniformity, and percent valid reads were used as acceptance criteria. To ensure high-quality results, a QC system was developed based on the NGS data generated during validation. QC criteria were established for several metrics at the run, sample, amplicon, and base pair levels for each nucleic acid type and run control threshold, with defined values ​​used to accept or reject one or more aspects of sequencing. Similarly, specific QC metrics were monitored over time to detect potential long-term assay drift. Quality filters were used to remove counts below the threshold for detection at the amplicon level and at the base pair level for low-quality variant calling.

[0119] RNA-seq absolute reads were generated using the immunResponseRNA (v5.2.0.0) plugin in the Torrent Suite. For each transcript, absolute read counts from the NTC were considered the library preparation background and were therefore subtracted from the absolute read counts for the same transcript in all other samples in the same preparation batch. To allow for comparable evaluation and interpretation of NGS measurements across runs, background-subtracted read counts were subsequently normalized to normalized reads per million (nRPM) values ​​as follows: The background-subtracted read values ​​for each HK gene were compared to a predetermined HK reads per million (HPM) profile. HK RPM profiles were established based on the average RPM of multiple replicates of the GM12878 sample across different validation sequencing runs. This resulted in a fold-change ratio for each HK gene.

[0120]

number

[0121] After this, the median of all HK ratios was used as the normalized ratio for a particular sample. Normalized ratio = median (all HK ratios) The nRPM of all genes (G) in a particular sample (S) was calculated as follows:

[0122]

number

[0123] DNA sequence variant calling was performed using the Variant Caller (version 5.2.0.34) plugin in the Ion Torrent Suite software (version 5.2.0), which requires a minimum minor allele frequency (MAF) of 0.1 and a minimum coverage of 20x. A series of filters was applied to the variants, generating a subset of MuB-qualified variants that met the following criteria: at least one minor allele read on both strands; MAF <0.2% in the 1,000 Genomes, Exome Aggregation Consortium (ExAC), and Exome Sequencing Project (ESP) databases; missense or nonsense; and coexisting somatic variants, as in the Ensembl team's curated database. The number of MuB-qualified variants was further normalized to the number of exonic bases, as reflected in the input BAM file, with a coverage of 20x or greater, to generate a normalized MuB score, interpreted as the number of mutations per million exonic bases. High MuB was defined as more than two standard deviations of the mean number of mutations per megabase of DNA in a tumor reference population of varying histological types.

[0124] For RNA-seq, the suitability of FFPE specimens was established by comparing RNA-seq results obtained from FFPE specimens with the corresponding FF samples. Principal component analysis (PCA) was used to demonstrate that different sections obtained from a given FFPE sample did not yield different RNA-seq results, that potential confounding factors such as the quantity or quality of stroma were well tolerated by the assay, and that the presence of multiple foci or primary lesions in the metastatic tumors tested had minimal impact on the results (data not shown). For MuB, the sensitivity of variant detection was used to assess the minimum threshold for the proportion of tumor nuclei.

[0125] To compare RNA-seq with the gold standard Taqman qRT-PCR, the nRPM values ​​of the target genes were log2-transformed, allowing for proper comparison with the ΔCt values ​​from the qRT-PCR measurements. The Pearson product-moment correlation coefficient (R) was calculated for each gene based on the log-transformed GEX measurements and ΔCt values. Any genes with low correlation values ​​(<0.7) were excluded from the final report. The R values ​​for nRPM of the 54 target genes were used as variables to evaluate GEX correlation for preanalytical, analytical, and reproducibility studies. R was calculated for FFPE vs. unstained sections, FF vs. FFPE sections, various percentages of non-tumorous tissue content, necrotic tissue content, input RNA amount (ng), genomic DNA (gDNA) contamination, batch size, expression linearity, inter-run reproducibility, intra-run reproducibility, inter-operator reproducibility, and inter-day reproducibility. Additionally, the coefficient of determination (R2) was used to demonstrate linearity in absolute reads for different library dilutions. Also, the average coefficient of variation (CV) for nRPM of the 54 target genes was calculated as a measure of the variance of GEX measurements for various batch sizes.

[0126] To evaluate the correlation between MuB counts and the gold-standard TCGA whole-exome data, genomic regions were selected from the MuB panel by filtering the TCGA whole-exome count. The TCGA variant counts mapped to the panel's Browser Extended Data (BED) file were correlated with MuB counts using Pearson's product-moment correlation. DNA stability was tested using a two-tailed Student's t-test between the mean MuB values ​​of FFPE versus unstained sections. The mean CV was calculated as a measure of variability in MuB measurements for FFPE versus unstained sections. CV was also used to demonstrate the effect of DNA input amount (ng) on ​​MuB measurements.

[0127] While various embodiments have been described and illustrated herein, those skilled in the art will readily envision various other means and / or structures for performing the functions and / or obtaining the results and / or advantages described herein, and each such variation and / or modification is deemed to be within the scope of the embodiments described herein. More generally, those skilled in the art will readily appreciate that all parameters, dimensions, materials, and configurations described herein are meant to be exemplary, and that the actual parameters, dimensions, materials, and / or configurations will depend on the particular application or uses for which the teachings are used. Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific embodiments described herein. Accordingly, it is to be understood that the foregoing embodiments are presented by way of example only and that, within the scope of the appended claims and their equivalents, embodiments may be practiced otherwise than as specifically described and claimed. Embodiments of the present disclosure are directed to each individual feature, system, article, material, and / or method described herein. Furthermore, any combination of two or more of such features, systems, articles, materials, and / or methods is within the scope of the present disclosure, provided that such features, systems, articles, materials, and / or methods are not mutually inconsistent.

[0128] The claims should not be construed as limited to the described order or elements unless expressly stated to that effect. It should be understood that various changes in form and detail can be made by those skilled in the art without departing from the spirit and scope of the appended claims. All embodiments that come within the spirit and scope of the following claims and equivalents thereof are claimed.

[0129] The technical concepts that can be understood from the above-described embodiment will be described below as supplementary notes. [Appendix 1] 1. A method for analyzing a tumor, the method comprising: determining at least one of a qualitative and quantitative assessment of tumor infiltrating lymphocytes in the sample, comprising determining RNA expression of a plurality of genes associated with tumor infiltrating lymphocytes; determining at least one of a qualitative and quantitative assessment of T cell receptor signaling in the sample, comprising determining RNA expression of a plurality of T cell receptor signaling genes (TCRS) associated with anti-cancer immune response and / or immunotherapy targets; determining at least one of a qualitative and quantitative assessment of mutational burden in said sample, comprising DNA sequencing and analysis of said sample; providing a classification index for the sample as a responder to treatment or a non-responder to treatment based on the determined qualitative and / or quantitative assessment of tumor infiltrating lymphocytes, the determined qualitative and / or quantitative assessment of T cell receptor signaling, and the determined qualitative and / or quantitative assessment of mutational burden; A method comprising:

[0130] [Appendix 2] 2. The method of claim 1, further comprising determining a tumor response to one or more potential treatment therapies based on the classifier.

[0131] [Appendix 3] 2. The method of claim 1, wherein the sample is provided with the classification index as a difficult responder to treatment.

[0132] [Appendix 4] 2. The method of claim 1, wherein a sample provided with a classification index as a non-responder to treatment is further provided with a classification index as being at risk for rapid cancer progression.

[0133] [Appendix 5] 2. The method of claim 1, further comprising correlating the classification index of the sample with a second classification index.

[0134] [Appendix 6] 6. The method of claim 5, further comprising determining a tumor response to one or more potential therapeutic therapies based on the correlation.

[0135] [Appendix 7] 2. The method of claim 1, further comprising determining immunohistochemistry data for the sample, wherein providing the classification index is further based on the determined immunohistochemistry data.

[0136] [Appendix 8] 2. The method of claim 1, wherein the plurality of genes associated with tumor-infiltrating lymphocytes includes at least CD163, CD2, CD3D, CD3E, CD3G, CD4, CD68, CD8A, CD8B, FOXP3, and CD20.

[0137] [Appendix 9] 2. The method of claim 1, wherein the plurality of genes associated with tumor-infiltrating lymphocytes consists only of CD163, CD2, CD3D, CD3E, CD3G, CD4, CD68, CD8A, CD8B, FOXP3, and CD20.

[0138] [Appendix 10] 2. The method of claim 1, wherein the plurality of genes associated with tumor-infiltrating lymphocytes and / or the plurality of T cell receptor signaling genes associated with anti-cancer immune response and / or immunotherapy targets consists only of CD163, CD2, CD3D, CD3E, CD3G, CD4, CD68, CD8A, CD8B, FOXP3, and CD20.

[0139] [Appendix 11] 2. The method of claim 1, wherein the plurality of T cell receptor signaling genes associated with anti-cancer immune response and / or immunotherapy targets comprise ADORA2A, CD40LG, TIM3, PD-L2, BTLA, CD80 (B7-1), ICOS, STAT1, VISTA (B7-H5), CD86 (B7-2), ICOSLG, TBX21, CCL2, CSF1R, IDO1, TGFB1, CCR2, CTLA4, IFNG, TNF, SLAMF4, CXCL10, IL10, TNFRSF14, CD27 (TNFRSF27), CXCR6, IL1B, GITR, PD-L1, DDX58, KLRD1, OX40, CD28, ENTPD1, ​​LAG3, CD137, CD38, GATA3, MX1, OX-40L, CD40, GZMB, and PD-1.

[0140] [Appendix 12] 2. The method of claim 1, wherein the plurality of T cell receptor signaling genes associated with anti-cancer immune response and / or immunotherapy targets consists solely of: ADORA2A, CD40LG, TIM3, PD-L2, BTLA, CD80 (B7-1), ICOS, STAT1, VISTA (B7-H5), CD86 (B7-2), ICOSLG, TBX21, CCL2, CSF1R, IDO1, TGFB1, CCR2, CTLA4, IFNG, TNF, SLAMF4, CXCL10, IL10, TNFRSF14, CD27 (TNFRSF27), CXCR6, IL1B, GITR, PD-L1, DDX58, KLRD1, OX40, CD28, ENTPD1, ​​LAG3, CD137, CD38, GATA3, MX1, OX-40L, CD40, GZMB, and PD-1.

[0141] [Appendix 13] The plurality of T cell receptor signaling genes associated with the anti-cancer immune response and / or immunotherapy target and / or the plurality of genes associated with the tumor-infiltrating lymphocytes include CD163, CD2, CD3D, CD3E, CD3G, CD4, CD68, CD8A, CD8B, FOXP3, CD20, ADORA2A, CD40LG, TIM3, PD-L2, BTLA, CD80 (B7-1), ICOS, STAT1, VISTA (B7-H5), CD86 (B7-2), ICOSLG, TBX21, C 2. The method of claim 1, wherein the genes include at least some of the genes selected from CL2, CSF1R, IDO1, TGFB1, CCR2, CTLA4, IFNG, TNF, SLAMF4, CXCL10, IL10, TNFRSF14, CD27 (TNFRSF27), CXCR6, IL1B, GITR, PD-L1, DDX58, KLRD1, OX40, CD28, ENTPD1, ​​LAG3, CD137, CD38, GATA3, MX1, OX-40L, CD40, GZMB, and PD-1.

[0142] [Appendix 14] The plurality of T cell receptor signaling genes associated with the anti-cancer immune response and / or immunotherapy target and / or the plurality of genes associated with the tumor-infiltrating lymphocytes include CD163, CD2, CD3D, CD3E, CD3G, CD4, CD68, CD8A, CD8B, FOXP3, CD20, ADORA2A, CD40LG, TIM3, PD-L2, BTLA, CD80 (B7-1), ICOS, STAT1, VISTA (B7-H5), CD86 (B7-2), ICO 2. The method of claim 1, wherein the antibodies include SLG, TBX21, CCL2, CSF1R, IDO1, TGFB1, CCR2, CTLA4, IFNG, TNF, SLAMF4, CXCL10, IL10, TNFRSF14, CD27 (TNFRSF27), CXCR6, IL1B, GITR, PD-L1, DDX58, KLRD1, OX40, CD28, ENTPD1, ​​LAG3, CD137, CD38, GATA3, MX1, OX-40L, CD40, GZMB, and PD-1.

[0143] [Appendix 15] 2. The method of claim 1, wherein the DNA sequencing and analysis of the sample comprises whole exome sequencing or partial exome sequencing.

[0144] [Appendix 16] 2. The method of claim 1, wherein DNA sequencing and analysis of the sample comprises whole exome sequencing.

[0145] [Appendix 17] 2. The method of claim 1, wherein the mutational load in the sample is the number of mutations per megabase of DNA.

[0146] [Appendix 18] 18. The method of claim 17, wherein the determined mutational burden is classified based on comparison to a reference population.

[0147] [Appendix 19] 19. The method of claim 18, wherein the determined mutational burden is classified as one of: (1) very high; (2) high; (3) medium; (4) low; and (5) very low based on the comparison.

[0148] [Appendix 20] 2. The method of claim 1, wherein providing the classifier for the sample as a treatment responder or a treatment non-responder comprises Bayesian model averaging of at least one of a determined qualitative assessment and a determined quantitative assessment of tumor infiltrating lymphocytes, a determined qualitative assessment and a determined quantitative assessment of T cell receptor signaling, and a determined qualitative assessment and a determined quantitative assessment of mutational burden.

Claims

1. 1. A method for providing an immune profile report, the method comprising: generating RNA sequencing data from one or more samples obtained from the patient's tumor, the RNA sequencing data comprising information about the expression of a plurality of tumor-infiltrating lymphocyte (TIL) proteins and a plurality of T-cell receptor signaling proteins; generating DNA sequence data from the one or more samples, the DNA sequence data including mutation load information for a plurality of genes; generating immunohistochemistry data from the one or more samples to measure PD-L1 protein expression and tumor infiltrating lymphocyte (TILS) expression; generating fluorescence in situ hybridization (FISH) data from the one or more samples to detect PD-L1 copy number gain; calculating, from the RNA-seq data, the DNA-seq data, the immunohistochemistry data, and the FISH data, the likelihood that the patient's tumor will respond to multiple possible treatments using a defined model; providing said immune profile report, said immune profile report including a calculated likelihood that said patient's tumor will respond to said plurality of possible treatments; A method comprising:

2. 10. The method of claim 1, wherein the immune profile report further comprises one or more personalized treatment options based on the calculated likelihood that the patient's tumor will respond to the plurality of possible treatments.

3. 10. The method of claim 1, wherein the likelihood of the patient's tumor responding to the plurality of possible treatments is further calculated from polymerase chain reaction (PCR) to assess microsatellite instability (MSI).

4. 2. The method of claim 1, wherein the information about the expression of multiple TIL proteins includes RNA expression data of multiple genes associated with multiple TIL proteins, and the multiple genes associated with TIL proteins include at least CD163, CD2, CD3D, CD3E, CD3G, CD4, CD68, CD8A, CD8B, FOXP3, and CD20.

5. The information about the expression of the plurality of T cell receptor signaling proteins includes RNA expression data of a plurality of genes associated with a plurality of T cell receptor signaling proteins, and the plurality of genes associated with the plurality of T cell receptor signaling proteins include ADORA2A, CD40LG, TIM3, PD-L2, BTLA, CD80 (B7-1), ICOS, STAT1, VISTA (B7-H5), CD86 (B7-2), ICOSLG, TBX21, 2. The method of claim 1, comprising all of CCL2, CSF1R, IDO1, TGFB1, CCR2, CTLA4, IFNG, TNF, SLAMF4, CXCL10, IL10, TNFRSF14, CD27 (TNFRSF27), CXCR6, IL1B, GITR, PD-L1, DDX58, KLRD1, OX40, CD28, ENTPD1, ​​LAG3, CD137, CD38, GATA3, MX1, OX-40L, CD40, GZMB, and PD-1.

6. 10. The method of claim 1, wherein the immune profile report further comprises an immune phenotypic classifier associated with at least one of checkpoint blockade (PD-1, CTLA-4), checkpoint blockade (other), primed T cells, myelosuppression, anti-inflammatory response, metabolic immune evasion, and pro-inflammatory response.

7. 7. The method of claim 6, wherein the immune profile report further comprises an immune phenotype interpretation.

8. The method of claim 1 , wherein the immunohistochemistry data includes expression data for CD3 and CD8.

9. 2. The method of claim 1, wherein calculating the likelihood that the patient's tumor will respond to the plurality of possible treatments comprises comparing the RNA-seq data, the DNA-seq data, and the immunohistochemistry data to a reference database of RNA-seq data, DNA-seq data, and immunohistochemistry data, and the likelihood is calculated based on ranking the patient relative to reference samples in the reference database.

10. 1. A system for providing an immune profile report, the system comprising: one or more processors; a memory coupled to said one or more processors, said memory being encoded with a set of instructions configured to perform the method of any one of claims 1 to 9; Including, the system.

Images