AlphaGenome Atlas maps every potential human DNA variant

Understanding how genetic variations affect biology at a molecular level remains a fundamental challenge in life sciences. Testing billions of single-letter DNA changes in a physical laboratory is practically impossible. To address this, researchers have introduced AlphaGenome Atlas, a platform containing precomputed predictions for the effects of 9 billion single-letter mutations in the human genome.

A comprehensive genomic map

AlphaGenome Atlas is a massive 1-petabyte dataset, more than 30 times larger than the AlphaFold Database. By precomputing predictions from the underlying artificial intelligence model, the platform charts molecular effects across hundreds of human and mouse cell types and tissues. It covers both coding regions—the two percent of the genome that builds proteins—and non-coding regions, which regulate gene activity.

To help scientists interpret the data, the platform introduces the AlphaGenome Variant Impact score. This metric condenses predictions from both the base model and AlphaMissense into a single number, allowing researchers to quickly rank genetic variants by their potential biological disruption. Additionally, the platform catalogs over 2,500 recurrent DNA sequence motifs to help locate transcription factor binding sites and regulatory elements.