COMPASS Model Lifts ICI Response Prediction Accuracy by 8.5% via Gene-Activity Bottleneck
COMPASS improves ICI patient selection by 8.5% using interpretable gene-activity concepts. The advance supports precision-medicine economics by reducing ineffective use of expensive checkpoint inhibitors. Prospective validation remains the critical next step.
If validated prospectively, COMPASS could serve as a pre-screening gate for high-cost immunotherapy, reducing the number needed to treat and sharpening statistical power in next-generation combination trials. Remaining questions include performance across underrepresented ancestries and cancers with sparse ICI data, plus integration with emerging spatial and single-cell readouts.
COMPASS: A 200-patient prospective validation cohort will show AUC >0.82 and reduce non-responder enrollment by 20% within 24 months of trial start.
Sources (3)
- [1]Primary Source(https://doi.org/10.1038/s41591-026-04502-7)
- [2]Supporting Source(https://www.nejm.org/doi/full/10.1056/NEJMoa1800429)
- [3]Supporting Source(https://www.nature.com/articles/s41591-022-01772-0)