Accurate variant effect prediction using AlphaMissense

ABSTRACT

Missense variants in the human genome can cause devastating genetic diseases, yet the vast majority of these variants remain clinically classified as 'variants of uncertain significance' (VUS). Building upon the architectural foundation of AlphaFold, we introduce AlphaMissense, an AI model that predicts the pathogenicity of all 71 million possible missense variants in the human proteome. By fine-tuning a structural language model on human and primate population frequency data, AlphaMissense achieves state-of-the-art accuracy in distinguishing between benign and pathogenic mutations. We provide a freely available database containing predictions for the entire human proteome, immediately accelerating rare disease diagnosis and genetic research.