MERIT: Mechanism driven model predicts drug outcomes and nominates indications for failed drugs

Abstract

Drug development depends on efficacy and safety, but many trial-outcome prediction models incorporate trial design, prior development history or compound identity, enabling compound memorization and inflating apparent performance. We developed MEchanism-Resolved Inference of Trial outcomes (MERIT), a model that predicts trial outcomes from molecular and disease features without using information on similar-compound success. MERIT integrates the disease and drug of interest with large-scale drug-protein, protein-metabolite and immune interaction maps to link a drug's intended and potential off-target effects to tissue-specific efficacy and safety. Across 753 small-molecule drugs and 3,133 trials, MERIT achieved an overall AUROC of 0.770 (0.765 for efficacy and 0.784 for safety) under structure-holdout evaluation, outperforming two published methods retrained on the same trials and folds. Scored by a mechanism-fit profiling, MERIT ranked the approved indication above the failed one for 83% (50 of 60) of a preselected set of 60 cohort compounds that failed one indication but are approved for another. Finally, we registered locked, outcome-blind predictions for 55 drug-indication pairs in ongoing phase III trials for prospective evaluation.