AI model maps tumor tissue to improve cancer care
EPFL researchers have created VirTues, a foundation model that learns from spatial proteomics data across 31 clinical cohorts, 12,000 images, and 5,000 patients. The model uses a new Transformer architecture to integrate diverse protein panels and can be applied to multiple analyses, from cell‑level mapping to patient outcome prediction. VirTues has been published in Nature and is designed to be flexible, allowing new tissue samples to be compared within the same framework.
EPFL researchers have created VirTues, a foundation model that learns from spatial proteomics data across 31 clinical cohorts, 12,000 images, and 5,000 patients. The model uses a new Transformer architecture to integrate diverse protein panels and can be applied to multiple analyses, from cell‑level mapping to patient outcome prediction. VirTues has been published in Nature and is designed to be flexible, allowing new tissue samples to be compared within the same framework.
VirTues exemplifies Swiss AI’s leadership in precision oncology by providing a scalable, foundation‑model approach that can be integrated into national cancer care pathways.
Original source record. Open the original record to verify the underlying announcement.