Google DeepMind released AlphaGenome Atlas, a freely available database containing predicted molecular effects for all approximately nine billion possible single-letter changes to human DNA2,4. The dataset, described as a 1-petabyte catalogue, was made available to researchers globally on September 8.
The atlas was built using Google DeepMind's AlphaGenome AI model, which generated precomputed predictions of the biological consequences of every possible single-nucleotide variant in the human genome3. The human genome contains roughly three billion letter pairs, written in an alphabet of four chemical bases: A, C, G, and T1. Each position can be changed to one of three alternative letters, producing the more than nine billion variants catalogued in the atlas.
AlphaGenome Atlas includes molecular effect predictions and AVI scores for each variant. Google DeepMind described it as "the most comprehensive catalogue of how genetic mutations affect molecular biology". The company's VP of Science and Chief Scientist for Google Cloud were listed in connection with the release.
Google DeepMind researchers said the platform "could help unravel the mysteries of the human genome and transform our understanding of biology, accelerating scientific research and ultimately paving the way for new treatments for diseases". The researchers published a blog post accompanying the release.
ANALYSIS The scale of the release is notable: precomputing predictions for every possible single-letter DNA change eliminates the need for researchers to run individual variant queries through the model, converting what would be a per-query inference task into a lookup operation against a static, petabyte-scale dataset. Making the atlas freely available positions it as a public research resource in the same lineage as Google DeepMind's earlier protein-structure databases, where open access drove broad adoption across academic and pharmaceutical research groups.