AlphaGenome Atlas Predicts Effects of All 9 Billion Human DNA Variants

Google DeepMind introduced AlphaGenome Atlas on September 8, 2026, a platform containing predicted molecular effects for all 9 billion possible single-letter DNA variants in the human genome, accompanied by a new variant-ranking score and a companion technical paper.

The resource is built by precomputing the predictions of AlphaGenome, the lab’s sequence-to-function model, across the entire genome rather than running the model one variant at a time. DeepMind’s announcement describes Atlas as the most comprehensive catalogue of how genetic mutations affect molecular biology. The resulting dataset is roughly 1 petabyte, which the company says is more than 30 times larger than the AlphaFold Database it expanded in 2022.

What the Atlas Contains

According to the technical paper, the team performed in silico saturation mutagenesis across the human genome, generating predictions for every possible single-nucleotide variant in the hg38 reference assembly plus more than 100 million insertions and deletions observed in the gnomAD, UK Biobank, and All of Us population datasets. Each variant carries an average of 27,000 experiment-specific scalar predictions spanning hundreds of human and mouse cell types and tissues.

Alongside the raw predictions, DeepMind released the AlphaGenome Variant Impact (AVI) score, which condenses AlphaGenome and AlphaMissense predictions with conservation and protein loss-of-function features into a single number per variant. Each AVI score is paired with feature attributions that decompose it into the molecular processes driving it, such as RNA splicing or gene expression. The Atlas also includes a compendium of 2,601 recurrent DNA sequence motifs, the regulatory “words” of the genome, with their locations mapped across cell types. DeepMind says the AVI score works across both the coding regions that make up about 2 percent of the genome and the non-coding 98 percent, and the paper reports state-of-the-art performance across variant pathogenicity and rare disease benchmarks, with the strongest gains on non-coding variants.

Early Research Applications

DeepMind says external collaborators have already used the resource. Working with the GREGoR Consortium, Laura Covill and Anne O’Donnell-Luria of the Broad Institute applied the AVI score to unsolved rare disease cases and prioritized a deep intronic variant in DNM1, a gene strongly linked to epileptic encephalopathy. The underlying AlphaGenome predictions showed the variant created a brain-specific cryptic splice site that extended the resulting protein by 13 amino acids, and the paper reports that experimental screens validated the prediction and identified nearby variants with similar effects, supporting a Likely Pathogenic classification.

At the University of Exeter, Medical Research Council fellow Gareth Hawkes applied Atlas to whole-genome data from more than 54,000 UK Biobank participants. By grouping rare variants according to their predicted molecular effects, Hawkes uncovered 22 percent more non-coding genetic associations with circulating protein levels than would otherwise have been detectable, pinpointing regulatory variants affecting proteins including PLA2G7, which is linked to aging, and EGLN1, a cellular oxygen sensor. Focusing on the 1 percent of non-coding variants Atlas predicts to be most impactful, he also identified 19 genetic regions associated with body mass index. At the Stowers Institute for Medical Research, Julia Zeitlinger and Melanie Weilert used the motif resource to categorize which transcription factors only affect DNA accessibility versus which can also switch genes on and off.

Availability and Stated Limits

The AlphaGenome Atlas website is open for non-commercial research use from September 8, 2026, alongside the AlphaGenome API and a skill in Google Antigravity, the company’s agentic development platform. DeepMind said commercial access on Google Cloud is coming soon, and the technical paper notes that a static download of AVI scores is permissively licensed for commercial and non-commercial use.

The authors state in the paper that Atlas and AVI are research tools that predict molecular effects and can serve only as part of the evidence chain leading to clinical diagnoses, not as sufficient evidence on their own. DeepMind’s disclaimer adds that AlphaGenome has not been validated or approved for any clinical use. The paper also notes gaps in AlphaGenome’s training data, including missing cell types and non-polyadenylated RNAs, and says the team views this first-generation Atlas as a baseline that will grow more precise as the underlying models improve.