SYSTEMS AND METHODS FOR IDENTIFYING RESPONDERS AND NON-RESPONDERS TO IMMUNE CHECKPOINT BLOCKADE THERAPY

1 . A system, comprising:

at least one computer hardware processor; and

at least one non-transitory computer-readable storage medium storing processor-executable instructions that, when executed by the at least one computer hardware processor, cause the at least one computer hardware processor to perform:

obtaining, for each subject in a plurality of subjects having responders to a checkpoint blockade therapy and non-responders to the checkpoint blockade therapy, expression data indicating expression levels for a plurality of genes;

determining, for the plurality of genes, expression level differences between the responders and the non-responders using the expression data;

identifying, using the determined expression level differences, a subset of genes associated with the checkpoint blockade therapy in the plurality of genes, wherein identifying the subset of genes associated with the checkpoint blockade therapy comprises identifying genes that are differentially expressed between the responders and non-responders with at least a threshold level of statistical significance;

training, using the expression data, a statistical model for predicting efficacy of the checkpoint blockade therapy, the training comprising:

identifying at least some of the subset of genes as a predictor set of genes to include in the statistical model; and

estimating, using the expression data, parameters of the statistical model that are associated with the predictor set of genes;

obtaining additional expression data for an additional subject; and

determining, using the additional expression data and the statistical model, whether the additional subject is likely to respond positively to the checkpoint blockade therapy and/or whether the additional subject is not likely to respond positively to the checkpoint blockade therapy.

2 . The system of claim 1 , wherein the expression data is RNA expression data, DNA expression data, or protein expression data.

3 . The system of claim 1 , wherein training the statistical model comprises training a generalized linear model having a plurality of regression variables, the plurality of regression variables including a regression variable for each of the predictor set of genes.

4 . The system of claim 1 , wherein training the statistical model comprises training a logistic regression model having a plurality of regression variables, the plurality of regression variables including a regression variable for each of the predictor set of genes of genes.

5 . The system of claim 4 , wherein the logistic regression model comprises a respective plurality of weights for the plurality of regression variables, wherein estimating the parameters of the statistical model comprises:

estimating the plurality of weights using the expression data for the plurality of subjects and information indicating which of the plurality of subjects responded to the checkpoint blockade therapy and/or which of the plurality of subjects did not respond to the checkpoint blockade therapy.

6 . The system of claim 1 , wherein training the statistical model comprises iteratively adding regression variables for respective genes to the statistical model, wherein iteratively adding regression variables comprises:

identifying a candidate gene in the subset of genes;

augmenting a current statistical model with a regression variable for the candidate gene to obtain an augmented statistical model;

evaluating performance of the augmented statistical model; and

determining to add the regression variable for the candidate gene to the current statistical model based on results of evaluating the performance.

7 . The system of claim 1 , wherein training the statistical model comprises training a generalized linear model having a plurality of regression variables, each of the plurality of regression variables representing a ratio of a pair of genes for respective pairs of members of the predictor set of genes.

8 . The system of claim 1 , wherein training the statistical model comprises training a logistic regression model having a plurality of regression variables, each of the plurality of regression variables representing a ratio of a pair of genes for respective pairs of members of the predictor set of genes.

9 . The system of claim 8 , wherein the logistic regression model comprises a respective plurality of weights for the plurality of regression variables, wherein estimating the parameters of the statistical model comprises:

estimating the plurality of weights using the expression data for the plurality of subjects and information indicating which of the plurality of subjects responded to the checkpoint blockade therapy and/or which of the plurality of subjects did not respond to the checkpoint blockade therapy.

10 . The system of claim 9 , wherein training the statistical model comprises iteratively adding regression variables for respective genes to the statistical model, wherein iteratively adding regression variables comprises:

identifying a candidate gene in the subset of genes;

augmenting a current statistical model with a regression variable for the candidate gene to obtain an augmented statistical model;

evaluating performance of the augmented statistical model; and

determining to add the regression variable for the candidate gene to the current statistical model based on results of evaluating the performance.

11 . The system of claim 10 , wherein evaluating performance of the augmented statistical model comprises obtaining an area under a receiver operating characteristic curve (ROC AUC) statistic.

12 . The system of claim 1 , wherein the statistical model comprises a first set of at least three dependent variables, each representing a ratio of a pair of genes, wherein the genes are selected from: BRAF, PRKAG1, STX2, AGPAT3, FYN, CMIP, ROBO4, RAB40C, HAUS8, SNAP23, SNX6, ACVR1B, MPRIP, COPS3, NLRX1, ELAC2, MON1B, ARF3, ARPIN, SPRYD3, FLI1, TIRAP, GSE1, POLR3K, PIGO, MFHAS1, NPIPA1, DPH6, ERLIN2, CES2, LHFP, NAIF1, ALCAM, SYNE1, SPINT1, SMTN, SLCA46A1, SAP25, WISP2, TSTD1, NLRX1, NPIPA1, HIST1H2AC, FUT8, FABP4, ERBB2, TUBA1A, XAGE1E, SERPINF1, RAI14, SIRPA, MT1X, NEK3, TGFB3, USP13, HLA-DRB4, IGF2, and MICAL1.

13 . The system of claim 1 , wherein the checkpoint blockade therapy is selected from the group consisting of: a PD1 inhibitor and a CTLA4 inhibitor.

14 . The system of claim 1 , wherein the system further comprises providing output to a user of whether the additional subject is likely to respond positively to the checkpoint blockade therapy and/or whether the additional subject is not likely to respond positively to the checkpoint blockade therapy.

15 . A method, comprising:

using at least one computer hardware processor to perform:

obtaining, for each subject in a plurality of subjects having responders to a checkpoint blockade therapy and non-responders to the checkpoint blockade therapy, expression data indicating expression levels for a plurality of genes;

determining, for the plurality of genes, expression level differences between the responders and the non-responders using the expression data;

identifying, using the determined expression level differences, a subset of genes associated with the checkpoint blockade therapy in the plurality of genes, wherein identifying the subset of genes associated with the checkpoint blockade therapy comprises identifying genes that are differentially expressed between the responders and non-responders with at least a threshold level of statistical significance;

training, using the expression data, a statistical model for predicting efficacy of the checkpoint blockade therapy, the training comprising:

identifying at least some of the subset of genes as a predictor set of genes to include in the statistical model; and

estimating, using the expression data, parameters of the statistical model that are associated with the predictor set of genes;

obtaining additional expression data for an additional subject; and

determining, using the additional expression data and the statistical model, whether the additional subject is likely to respond positively to the checkpoint blockade therapy and/or whether the additional subject is not likely to respond positively to the checkpoint blockade therapy.

16 . The method of claim 15 , wherein training the statistical model comprises training a logistic regression model having a plurality of regression variables, each of the plurality of regression variables representing a ratio of a pair of genes for respective pairs of members of the predictor set of genes, wherein the logistic regression model comprises a respective plurality of weights for the plurality of regression variables, and wherein estimating the parameters of the statistical model comprises:

estimating the plurality of weights using the expression data for the plurality of subjects and information indicating which of the plurality of subjects responded to the checkpoint blockade therapy and/or which of the plurality of subjects did not respond to the checkpoint blockade therapy.

17 . The method of claim 16 , wherein training the statistical model comprises iteratively adding regression variables for respective genes to the statistical model, wherein iteratively adding regression variables comprises:

identifying a candidate gene in the subset of genes;

augmenting a current statistical model with a regression variable for the candidate gene to obtain an augmented statistical model;

evaluating performance of the augmented statistical model; and

determining to add the regression variable for the candidate gene to the current statistical model based on results of evaluating the performance.

18 . A system, comprising:

at least one computer hardware processor; and

at least one non-transitory computer-readable storage medium storing processor-executable instructions that, when executed by the at least one computer hardware processor, cause the at least one computer hardware processor to perform:

accessing a statistical model, wherein the statistical model was obtained by:

obtaining, for each subject in a plurality of subjects having responders to a checkpoint blockade therapy and non-responders to the checkpoint blockade therapy, expression data indicating expression levels for a plurality of genes;

determining, for the plurality of genes, expression level differences between the responders and the non-responders using the expression data;

identifying, using the determined expression level differences, a subset of genes associated with the checkpoint blockade therapy in the plurality of genes, wherein identifying the subset of genes associated with the checkpoint blockade therapy comprises identifying genes that are differentially expressed between the responders and non-responders with at least a threshold level of statistical significance;

training, using the expression data, a statistical model for predicting efficacy of the checkpoint blockade therapy, the training comprising:

identifying at least some of the subset of genes as a predictor set of genes to include in the statistical model; and

estimating, using the expression data, parameters of the statistical model that are associated with the predictor set of genes;

obtaining additional expression data for an additional subject; and

determining, using the additional expression data and the statistical model, whether the additional subject is likely to respond positively to the checkpoint blockade therapy and/or whether the additional subject is not likely to respond positively to the checkpoint blockade therapy.

19 . The system of claim 18 , wherein training the statistical model comprises training a logistic regression model having a plurality of regression variables, each of the plurality of regression variables representing a ratio of a pair of genes for respective pairs of members of the predictor set of genes, wherein the logistic regression model comprises a respective plurality of weights for the plurality of regression variables, and wherein estimating the parameters of the statistical model comprises:

estimating the plurality of weights using the expression data for the plurality of subjects and information indicating which of the plurality of subjects responded to the checkpoint blockade therapy and/or which of the plurality of subjects did not respond to the checkpoint blockade therapy.

20 . The system of claim 19 , wherein training the statistical model comprises iteratively adding regression variables for respective genes to the statistical model, wherein iteratively adding regression variables comprises:

identifying a candidate gene in the subset of genes;

augmenting a current statistical model with a regression variable for the candidate gene to obtain an augmented statistical model;

evaluating performance of the augmented statistical model; and

determining to add the regression variable for the candidate gene to the current statistical model based on results of evaluating the performance.