Researchers from Harvard Medical School unveiled COMPASS, an AI model built to predict patient response to immune checkpoint inhibitors (ICIs), with results described as improving accuracy in anticipating outcomes. The system was developed by a team led by Associate Professor Marinka Zitnik, using data patterns intended to forecast how patients may respond to immunotherapy drugs. In clinical immuno-oncology, response prediction remains a major bottleneck for treatment selection, particularly because current biomarkers often fail to capture the full variability between patients. COMPASS is positioned as a tool for more refined patient stratification for ICIs. If the model performs robustly across external datasets, it could influence how clinicians and developers evaluate immunotherapy candidates and optimize treatment planning.