Scientists have developed an artificial intelligence model capable of identifying previously unknown human genetic mutations likely to cause disease, offering new hope for patients with rare and undiagnosed conditions. The research, reported by the Financial Times, describes the AI, called popEVE, as outperforming rivals, including Google DeepMind’s AlphaMissense, in predicting which genetic variants may be harmful.
PopEVE draws on evolutionary data from hundreds of thousands of animal species, combined with human genetic information from the UK Biobank and gnomAD databases. The model focuses particularly on “missense” mutations, which involve changes to a single amino acid in a protein. By comparing genetic sequences across species, the AI can infer whether a mutation is likely destructive, since harmful mutations are often absent in evolutionary records.
In testing, popEVE was applied to genetic data from 31,000 families with children suffering severe developmental disorders. For 513 children with completely novel mutations, the model correctly identified the most damaging variant 98% of the time. It also flagged 123 genes active in brain development that had never previously been linked to developmental disorders. Researchers said popEVE outperformed other models in predicting disease severity and in assessing mutations in non-European populations.
The AI, created through collaboration between the Centre for Genomic Regulation in Barcelona and Harvard Medical School, builds on a 2021 algorithm called EVE (Evolutionary model of Variant Effect). It is energy-efficient and has already been applied in low- and middle-income countries, including Senegal, where it helped guide treatment for a patient with muscular atrophy using vitamin B2 supplementation.
“This latest work will allow scalable application to every gene,” said Damian Smedley, a computational genomics professor at Queen Mary University of London. “Being able to systematically assess the impact of all variants in a patient’s genome is key to fully delivering the promise of genomic sequencing in healthcare.”
The development of popEVE could significantly expand doctors’ ability to diagnose and treat rare genetic diseases, particularly in cases where parental genetic samples are unavailable, potentially transforming care for millions of patients worldwide.
