DeepMind ships AlphaGenome Atlas: 9B DNA variant predictions

Google DeepMind published predictions for what it says is every possible single-letter DNA change in the human genome, posting AlphaGenome Atlas on September 8. The company describes the resource as covering "9 billion single-nucleotide variants" and the underlying data as "a massive 1-petabyte dataset, more than 30 times larger than the AlphaFold Database."

The centerpiece is the AlphaGenome Variant Impact (AVI) score, which "combines the strengths of AlphaGenome and AlphaMissense," condensing model outputs into a single number for ranking variants. Coverage spans the 2% of the genome that codes for proteins and the 98% that does not, with feature attributions flagging molecular processes like RNA splicing and gene expression.

Early collaborators reported hits at launch. At the Broad Institute, Laura Covill and Anne O'Donnell-Luria's team used the AVI score to identify a variant in the DNM1 gene linked to epileptic encephalopathy that had been overlooked previously. Gareth Hawkes, an MRC fellow at the University of Exeter, ran the resource against UK Biobank data from more than 54,000 participants and reported "22% more non-coding genetic associations," plus regulatory variants affecting PLA2G7 and EGLN1 and 19 genetic regions related to body mass index. At the Stowers Institute, Julia Zeitlinger and Melanie Weilert used it to distinguish transcription factors by their regulatory effects.

Access runs through a free web portal, the AlphaGenome API, and a Google Antigravity skill; non-commercial use is open immediately, with commercial availability on Google Cloud slated to follow. DeepMind states plainly that the output "is not intended to be a substitute for professional medical advice, diagnosis, or treatment," and is not approved for any clinical use.

Three of the researchers on our Who's Who radar shared the post on release day.