Particle.news
Download on the App Store

AI Model Predicts Cancer Survival From Single-Cell Tumor Data

Peer-reviewed tests in Cancer Discovery showed higher accuracy on melanoma and liver cancer cohorts.

Overview

  • The OHSU team published the scSurvival study Tuesday in Cancer Discovery, introducing a way to forecast outcomes from data on individual tumor and immune cells.
  • The model assigns each cell a weight based on how strongly it relates to survival and then averages the weighted cells to produce a patient risk score.
  • With NIH support, researchers tested scSurvival on clinical data from more than 150 patients with melanoma or liver cancer and found it outperformed standard bulk methods.
  • The approach traces predictions to specific cell states, including cell groups in melanoma linked to responses to immunotherapy.
  • The software and tutorials are open-source on GitHub, Zenodo, and Code Ocean, and the authors say it needs broader validation before use in patient care.