AlphaGenome Atlas: Google DeepMind Launches 1-Petabyte Database Predicting Effects of 9 Billion DNA Variants
Google DeepMind has released AlphaGenome Atlas, an AI database predicting the molecular effects of all 9 billion possible single-letter DNA changes in the human genome. The 1-petabyte resource includes a new impact score and is freely available for academic research to aid disease studies.
Google DeepMind has launched AlphaGenome Atlas, a searchable AI-powered database that predicts the molecular effects of every possible single-letter change across the human genome. The resource covers all 9 billion single-nucleotide variants and was announced on 8 September 2026.
Demis Hassabis, co-founder and chair of Google DeepMind, highlighted the advance, noting that AlphaFold mapped the protein universe while AlphaGenome Atlas now charts the human genome. The tool aims to help scientists better understand disease by showing how DNA variants affect biological processes. Google DeepMind CEO Demis Hassabis Predicts AGI Within 5 to 8 Years, Hails 'Golden Era' of Science at India AI Impact Summit 2026.
What AlphaGenome Atlas Provides?
The atlas uses predictions from the AlphaGenome AI model, previously released by DeepMind. Researchers pre-calculated the regulatory impact of every possible single DNA letter change, producing a dataset of around 1 petabyte. This is more than 30 times larger than the AlphaFold Database.
It includes the AlphaGenome Variant Impact (AVI) score. This combines outputs from AlphaGenome and the earlier AlphaMissense model with other features into a single number. The score ranks variants from low to high impact and indicates potential molecular effects, such as disruption to gene switches or RNA splicing.
Scientists can query the data through an intuitive website that requires no coding skills, via the AlphaGenome API, or as a skill in Google Antigravity. Access is free for non-commercial academic research. Commercial use on Google Cloud is planned soon.
How Researchers Can Use the Atlas?
The resource supports work on rare genetic diseases and complex traits. Early applications include prioritising variants in unsolved rare disease cases and identifying additional non-coding associations linked to traits such as body mass index in large biobank data.
Most of the human genome does not code for proteins but regulates gene activity. AlphaGenome Atlas extends predictions across both coding and non-coding regions, giving researchers a genome-wide view of potential functional effects.
Background on DeepMind’s Genomics Tools
AlphaGenome, released in 2025 with further details in early 2026, analyses DNA sequences up to one million base pairs long. It predicts how variants influence gene expression, splicing, chromatin features and related processes across many cell types.
It builds on AlphaFold, the protein structure prediction system that contributed to the 2024 Nobel Prize in Chemistry shared by Hassabis and colleagues, and on AlphaMissense for assessing protein-altering mutations. The new atlas makes the underlying model’s insights more accessible by pre-computing results for the entire genome. Gemini Omni Launched: Google Unveils New AI Video Generator at I/O 2026.
DeepMind states the predictions are for research purposes only and must not be used for clinical decision-making. The company positions the atlas as a foundational tool to accelerate genomic discovery for the scientific community.
(The above story first appeared on LatestLY on Sep 08, 2026 10:38 PM IST. For more news and updates on politics, world, sports, entertainment and lifestyle, log on to our website latestly.com).
