Researchers have developed an AI model named popEVE that uses evolutionary data across hundreds of thousands of species and human genetic variation to assess which protein-altering mutations are likely to cause disease, even when they’ve never been seen in a person before. This tool can rank mutation severity across the entire human proteome, allowing doctors to focus faster on the most damaging variants, and it performed strongly in validation tests by correctly identifying known causal mutations in 98% of cases studied. The model also minimizes bias against people from underrepresented ancestries and is poised to speed diagnosis of rare diseases in settings with limited resources, although it doesn’t replace clinical judgment and only addresses protein-altering mutations.
