A New AI Tool Could Transform How We Diagnose Genetic Diseases

Researchers at the Mayo Clinic and Goodfire have developed an AI model, Evo 2, that can predict and explain which genetic mutations cause disease. This approach uses AI interpretability techniques to understand the AI's decision-making process, offering a new way to diagnose and study genetic disorders on a large scale. Evo 2, trained on vast amounts of genomic data, learns patterns of healthy DNA and can identify biological features of mutations, surpassing existing computational tools in predicting pathogenicity. This advancement is significant because it not only predicts disease-causing mutations but also explains the underlying biological reasons, which is crucial for clinical applications. As genome sequencing costs decrease, such interpretability methods could enable personalized therapies. However, further trials and FDA approval are needed before clinical use, and there are ongoing discussions about whether the AI truly discovers new biological concepts or only identifies known ones.

(Source：TIME)