Identification of immune susceptibility features to predict response in cancer immunotherapy

Abstract

Cancer immune checkpoint blockade (ICB) shows durable clinical benefits in treating melanoma, but only limited number of patient responds to such therapy. Combination ICB therapies have shown to increase number of responding patients. A reliable predictor of ICB response is needed to ascertain patients who will respond to ICB prior to treatment and to efficiently seek novel ICB combination drugs. Here I present anti-PD-1 Immunotherapy Signature (aPIMS), a melanoma-intrinsic predictor of anti-PD-1 ICB response. It is an unbiased, machine learning based signature that is able to predict anti-PD-1 ICB response in patient-derived data as well as cell line data. I also use aPIMS on cell line perturbation data to screen for novel anti-PD-1 ICB combination drugs.