Harvard AI predicts cancer immunotherapy response

COMPASS: The AI Ending the "Black Box" of Cancer Immunotherapy | Harvard’s Precision BreakthroughWhy do some cancer patients experience "miracle" recoveries while others see no benefit from the same drugs? In today’s episode, we investigate a major breakthrough published in Nature Medicine: COMPASS, a new AI foundation model from Harvard Medical School designed to predict patient response to immune checkpoint inhibitors (ICIs). We break down how this model outperformed 22 existing methods by an average of 8.5%, potentially marking the difference between a successful clinical trial and a failure.We dive into the "Concept Bottleneck" architecture: unlike traditional "black box" AI, COMPASS provides a clear rationale by analyzing 16,000 genes and mapping them to 44 interpretable immune concepts. We also discuss the strategic "regulatory gap"—why using bulk RNA-seq data instead of DNA makes COMPASS a powerful adaptive tool today, but a complex challenge for the FDA tomorrow. Join us as we explore how this technology could have refined treatments for patients like Jimmy Carter and what it means for the next generation of personalized oncology.Become a supporter of this podcast: https://www.spreaker.com/podcast/scitech-brief--6846358/support.This episode includes AI-generated content.