Researchers at Harvard Medical School have developed a new AI model called popEVE that ranks genetic variants across the entire human genome by how likely they are to cause disease, enabling clinicians to spot the most harmful ones with far greater precision. By combining evolutionary data from many species with human population genetics, the model can identify whether a variant is likely benign, pathogenic, inherited or de novo, and even predict whether it might cause early childhood death. In tests on thousands of undiagnosed patients, popEVE helped diagnose roughly one-third of cases and even proposed more than 100 new disease-causing gene candidates. The model also showed little bias across different ancestries and is being deployed to clinicians and researchers as a diagnostic support tool that could accelerate the discovery of rare diseases and their treatments.
