Harvard’s COMPASS AI Predicts Cancer Treatment Success

Researchers at Harvard Medical School have introduced an artificial intelligence model called COMPASS to forecast patient responses to cancer immunotherapy ^[1]. This tool specifically targets immune checkpoint inhibitors, which are vital drugs used to treat various malignancies ^[1][2]. The study, published July 3, 2026, in Nature Medicine, reveals that the AI improves prediction accuracy by 8.5% compared to existing clinical standards ^[1].

COMPASS analyzes tumor gene expression through 44 specific immune concepts to determine if a patient will react positively to treatment ^[1][3]. By using biologically grounded data, the model aims to help clinicians make more informed decisions regarding personalized cancer care ^[2]. Scientists tested the tool across multiple cancer types to ensure its effectiveness in diverse medical scenarios ^[1].