Generating counterfactual explanations of tumor spatial proteomes to discover therapeutic strategies for enhancing immune infiltration

Abstract

AbstractImmunotherapies can halt or slow down cancer progression by activating either endogenous or engineered T cells to detect and kill cancer cells. For immunotherapies to be effective, T cells must be able to infiltrate the tumor microenvironment. However, many solid tumors resist T-cell infiltration, challenging the efficacy of current therapies. Here, we introduce Morpheus, an integrated deep learning framework that takes large scale spatial omics profiles of patient tumors, and combines a formulation of T-cell infiltration prediction as a self-supervised machine learning problem with a counterfactual optimization strategy to generate minimal tumor perturbations predicted to boost T-cell infiltration. We applied our framework to 368 metastatic melanoma and colorectal cancer (with liver metastases) samples assayed using 40-plex imaging mass cytometry, discovering cohort-dependent, combinatorial perturbations, involving CXCL9, CXCL10, CCL22 and CCL18 for melanoma and CXCR4, PD-1, PD-L1 and CYR61 for colorectal cancer, predicted to support T-cell infiltration across large patient cohorts. Our work presents a paradigm for counterfactual-based prediction and design of cancer therapeutics using spatial omics data.