AlphaGenome Atlas: A predictive map of every possible DNA letter change in the human genome

Interpreting how genetic mutations affect biology is a fundamental challenge, since testing every one of the roughly 9 billion possible single-letter changes in the human genome in the lab is impossible. DeepMind's AlphaGenome AI model could already predict the impact of genetic variants, but it was primarily useful for analyzing specific variants; researchers needed a genome-wide perspective.

AlphaGenome Atlas provides that perspective by precomputing AlphaGenome predictions for every possible single-nucleotide variant. It is a massive 1-petabyte dataset, more than 30 times larger than the AlphaFold Database. To help scientists quickly find the most impactful changes, DeepMind also released the AlphaGenome Variant Impact (AVI) score, which combines AlphaGenome and AlphaMissense predictions into a single number so researchers can rank and interpret variants.

The atlas is available through a free-to-use website portal, the AlphaGenome API, and as a skill in Google Antigravity. External collaborators have already used it to experimentally verify key variants in unsolved rare disease research and to identify rare variants associated with common traits.

DeepMind compares the release to its 2022 expansion of the AlphaFold Database, which grew available 3D protein structure information from roughly 190K experimental structures to over 200M predictions and made large-scale analysis accessible to researchers without coding experience. AlphaGenome Atlas aims to similarly democratize genomic variant analysis.