Particle.news
Download on the App Store

Mount Sinai's V2P AI Links Genetic Variants to Likely Disease Types

Published in Nature Communications, the model learned from large variant datasets to prioritize likely causal mutations in patient tests.

Overview

  • V2P predicts disease categories tied to specific DNA variants rather than only labeling mutations as harmful or benign.
  • In de-identified patient evaluations, the correct disease-causing variant often appeared within the model’s top 10 ranked candidates.
  • Training incorporated extensive catalogs of benign and pathogenic variants alongside disease annotations to boost phenotype-specific accuracy.
  • The team highlights potential use in faster genetic diagnostics and in prioritizing genes and pathways for precision therapy development.
  • Outputs currently map to broad categories such as nervous system disorders or cancers, with planned refinements for finer disease specificity and added data sources.