Spatial biomarker discovery via interpretable semantic learning in histopathology.

PathPrism identifies hundreds of spatial biomarkers predictive of survival and molecular alterations in colorectal cancer, advancing precision oncology with 7,000 patient data.

• Why it matters: Discovering spatial biomarkers from complex histopathology images is difficult, limiting personalized treatment strategies and understanding of tumor biology.
• What they did: The study developed PathPrism, an interpretable AI framework that encodes tissue architecture into spatial features, applied to large patient cohorts, and integrated large language models for hypothesis generation.
• The result: PathPrism enables transparent biomarker discovery and virtual experimentation, facilitating scalable, interpretable insights into tumor prognosis and therapy response.