Harvard Medical School researchers developed COMPASS, an AI model designed to predict patient response to immune checkpoint inhibitors (ICIs), and reported that it outperforms ICD-10-based approaches for capturing relevant clinical data features. COMPASS is led by Associate Professor Marinka Zitnik’s team and uses model-driven patient data integration to forecast whether patients may respond to ICI therapy. The work is positioned around practical decision support as immunotherapy becomes more individualized. The focus on prediction targets a major bottleneck for checkpoint therapy—identifying responders early enough to guide treatment selection and avoid ineffective exposure. As with other precision medicine models, the field will watch for external validation, calibration across sites, and how the model performs across cancer types and lines of therapy.