A computational network approach to identify predictive biomarkers and therapeutic combinations for anti-PD-1 immunotherapy in cancer
AbstractBackgroundDespite remarkable success, only a subset of cancer patients have shown benefit from the anti-PD1 therapy. Therefore, there is a growing need to identify predictive biomarkers and therapeutic combinations for improving the clinical efficacy.ResultsBased upon the hypothesis that aberrations of any gene that are close to MHC class I genes in the gene network are likely to deregulate MHC I pathway and affect tumor response to anti-PD1, we developed a network approach to infer genes, pathway, and potential therapeutic target genes associated with response to PD-1/PD-L1 checkpoint immunotherapies in cancer. Our approach successfully identified genes (e.g. B2M and PTEN) and pathways (e.g. JAK/STAT and WNT) known to be associated with anti-PD1 response. Our prediction was further validated by 5 CRISPR gene sets associated with tumor resistance to cytotoxic T cells. Our results also showed that many cancer genes that act as hubs in the gene network may drive immune evasion through indirectly deregulating the MHC I pathway. The integration analysis of transcriptomic data of the 34 TCGA cancer types and our prediction reveals that MHC I-immunoregulations may be tissue-specific. The signature-based score, the MHC I association immunoscore (MIAS), calculated by integration of our prediction and TCGA melanoma transcriptomic data also showed a good correlation with patient response to anti-PD1 for 354 melanoma samples complied from 5 cohorts. In addition, most targets of the 36 compounds that have been tested in clinical trials or used for combination treatments with anti-PD1 are in the top list of our prediction (AUC=0.833). Integration of drug target data with our top prediction further identified compounds that were recently shown to enhance tumor response to anti-PD1, such as inhibitors of GSK3B, CDK, and PTK2.ConclusionOur approach is effective to identify candidate genes and pathways associated with response to anti-PD-1 therapy, and can also be employed for in silico screening of potential compounds to enhances the efficacy of anti-PD1 agents against cancer.