On 2026-09-09, Google DeepMind announced "AlphaGenome Atlas," a resource designed to predict the effects of approximately 9 billion single-nucleotide variant (SNV) patterns within the human genome.

AlphaGenome Atlas is a comprehensive catalog of pre-computed predictions from the AI model "AlphaGenome," detailing how variations across the genome influence molecular biology. The platform will be made available to academic researchers via a web portal and API.

Along with the atlas, the company released the "AlphaGenome Variant Impact (AVI) score" as a metric for evaluating the severity of mutations. This score combines AlphaGenome with AlphaMissense—a model for predicting DNA mutations that alter proteins—allowing researchers to rapidly prioritize the potential impact of specific variants. The AVI score functions not only for protein-coding regions but also for mutations in non-coding regions that regulate gene activity.

AlphaGenome Atlas is a massive dataset exceeding one petabyte, significantly surpassing the scale of the previous AlphaFold Database. Validation by collaborators has indicated that the resource is helpful in identifying mutations that cause rare diseases and discovering non-coding region mutations associated with common traits.

Source: