Harvard Medical School researchers led by Marinka Zitnik unveiled COMPASS, an AI model designed to predict patient response to immune checkpoint inhibitors (ICIs). The approach is framed as improving decision-making for personalized cancer immunotherapy by identifying likely responders among patients considering or receiving checkpoint therapy. COMPASS was presented as a model that sets a new standard for predicting response, reflecting a broader push to combine clinical and molecular signals with machine-learning methods to reduce trial-and-error prescribing in oncology. In practice, response prediction can support stratification, trial enrichment, and potentially faster course correction when evidence suggests limited benefit. The excerpt highlights that COMPASS is specifically oriented to ICIs, an “immune checkpoint inhibitors” category that includes PD-1/PD-L1 and CTLA-4 pathway therapeutics. Those agents are widely used, but response rates vary substantially across tumor types and patient profiles. If validated beyond initial cohorts, models like COMPASS can reshape how clinicians and developers prioritize biomarkers and patient selection strategies for checkpoint-based regimens.