Researchers have developed a new artificial intelligence model capable of signaling whether previously unknown human genetic mutations may cause disease, according to the Financial Times. The technique draws on genetic information from hundreds of thousands of species — mainly animals — and, according to its creators, outperforms existing systems, including Google DeepMind’s AlphaMissense.
The innovation promises to give clinicians additional insight when dealing with cases involving extremely rare or genetically unique conditions. Rare diseases affect hundreds of millions of people worldwide, yet many remain undiagnosed due to limited data on the mutations responsible.
“There are many ways in which individual genetic variants can cause disease, and for this very large number of patients there is often a terrible lack of information available. It is hard to diagnose the disease, it is hard to understand how to treat the disease. We hope that we have provided just a new, very general tool to guide this process,” said Jonathan Frazer from the Centre for Genomic Regulation in Barcelona.
The model, called popEVE, was developed by researchers in Barcelona in collaboration with Harvard Medical School and is described in a study published Monday in Nature Genetics. PopEVE builds on EVE, a 2021 algorithm — the Evolutionary model of Variant Effect — designed to evaluate how genetic changes influence the instructions that guide the body in making proteins.
The scientists focused particularly on missense mutations, changes that alter a single amino acid in a protein. To assess whether these alterations might be harmful, the team analyzed evolutionary patterns across numerous species. If certain mutations do not appear in evolutionary records, it may indicate that they are detrimental, as organisms carrying them might not have survived.
These evolutionary insights were then calibrated using human genetic data from UK Biobank and gnomAD, helping the model determine which variants are tolerated in healthy individuals.
A key advantage of popEVE is its low computational cost, making it suitable for low- and middle-income countries. The model has already been used successfully on patients in Senegal, including in a case of muscle atrophy treated with vitamin B2 supplements.
Photo: freepik.com
