Labels as a Feature: Network Homophily for Systematically Discovering human GPCR Drug-Target Interactions

Abstract Machine learning (ML) has revolutionized drug discovery by enabling the exploration of vast, uncharted chemical spaces essential for discovering novel patentable drugs. Despite the critical role of human G protein-coupled receptors (hGPCRs) in FDA-approved drugs, exhaustive in-distribution drug-target interaction (DTI) testing across all pairs of hGPCRs and known drugs is rare due to significant economic and technical challenges. This often leaves off-target effects unexplored, which poses a considerable risk to drug safety. In contrast to the traditional focus on out-of-distribution (OOD) exploration (drug discovery), we introduce a neighborhood-to-prediction model termed Chemical Space Neural Networks (CSNN) that leverages network homophily and training-free graph neural networks (GNNs) with Labels as Features (LaF). We show that CSNN’s ability to make accurate predictions strongly correlates with network homophily. Thus, LaFs strongly increase a ML model’s capacity to enhance in-distribution prediction accuracy, which we show by integrating labeled data during inference. We validate these advancements in a high-throughput yeast biosensing system (3773 DTIs, 539 compounds, 7 hGPCRs) to discover novel DTIs for FDA-approved drugs and to expand the general understanding of how to build reliable predictors to guide experimental verification.