Researchers led by Associate Professor Marinka Zitnik at Harvard Medical School introduced COMPASS, an AI model designed to predict patient response to immune checkpoint inhibitor therapy. The report states the system improves accuracy for identifying who is likely to benefit from checkpoint inhibitors, aiming to support more personalized immunotherapy decisions. The study frames COMPASS as an approach to handle heterogeneity in treatment response, where clinical and molecular factors can be insufficient for reliable prediction on their own. For oncology developers, the signal is that model-driven stratification could influence trial enrichment, biomarker strategy, and ultimately prescribing. The dataset described here focuses on model capability rather than regulatory action or trial endpoints, but it adds to the growing push for AI-supported patient selection in immuno-oncology.