Transcriptional analysis of immune modulatory genes in melanoma treated with PD-1 blockade

Abstract/Summary We aimed to characterize immunological features of melanoma patients treated with PD-1 blockade using tumor transcriptomic datasets. Response-dependent and response-independent predictors based on biological knowledge were investigated. Domain knowledge-driven regression-based analysis identified CEACAM1, CD40, B7-H3 , and CD112 as key genes that determine the melanoma immune status. We devised the transcriptional deviance score ( TDS ) representing the individual sample-wise contribution to the immune network. The TDS not only showed good predictive power for immune checkpoint inhibitor (ICI) responses but also suggested specific gene interactions that determine ICI responses. Dynamic TDS changes following ICI treatment were related to long survival, indicating immune network modulation by ICIs occurred in responders. A predictive model incorporating B7-H3 and CEACAM1 expression, mutational status, clinical features, and the TDS showed excellent performance for ICI response. Thus, our approaches suggest a novel measure for the tumor immune temperature and provide insight into melanoma immunobiology. Highlights We applied outcome-independent and outcome-dependent methods to investigate melanoma immunobiology. CEACAM1, CD40, B7-H3 , and CD112 expression levels are key determinants of immune status. We devised a TDS that could measure tumor immune network status at the individual level. Incorporating regression and correlation approaches greatly improves predictive power.