AI boosts rare diagnosis

A new AI model called popEVE shows promise in accelerating the diagnosis of rare genetic disorders. It functions by predicting which mutations in a patient's genome are most likely to cause severe disease. The model was trained on genome data from over 31,000 patients with rare developmental disorders.

PopEVE combines evolutionary information with population sequence data to rank variants by severity. It identifies 442 genes in a cohort of developmental disorder cases, including evidence of 119 novel genetic disorders. The model can pinpoint likely causal candidates even for extremely rare single-patient disorders.

When tested, popEVE achieved a 15x enrichment for pathogenic variants, outperforming existing methods. Researchers are now testing popEVE in clinical settings to see if it can speed accurate diagnoses of rare, single-variant genetic diseases.