Google DeepMind announced AlphaGenome Atlas in September 2026, an AI platform containing predictions for how every possible single-letter DNA change in the human genome could affect the body at a molecular level. The company is making it available to researchers for noncommercial use immediately, with commercial access on Google Cloud to follow.
The human genome is written in four chemical letters — A, C, G, and T — spanning roughly three billion letter pairs. That structure yields approximately nine billion potential single-letter substitutions, each of which could be harmless, account for ordinary variation between people, or contribute to disease. AlphaGenome Atlas provides predictions for all nine billion variants, covering not only protein-coding regions but also the stretches of DNA that regulate how genes are switched on and off. Google describes it as “the most comprehensive catalogue of how genetic mutations affect molecular biology.” The resulting dataset is approximately one petabyte in size.
To help researchers prioritize which mutations warrant closer investigation, Google is also releasing a Variant Impact Score that draws on the company’s other DNA-prediction models. Scientists can access Atlas through a web portal, through Google’s agentic development platform Antigravity, and via its AlphaGenome interface.
Ziga Avsec, DeepMind’s genomics lead, explained that while the underlying AlphaGenome model was released previously, computing predictions across the full genome took additional time. “Basically it took us some time to really precompute and also analyze this many variants because the space is so big,” he said. AlphaGenome was trained on public databases of human and mouse genomes.
Atlas builds on earlier DeepMind biology tools including AlphaGenome, released last year, and AlphaMissense, which focused on predicting the effects of small protein-altering mutations. The broader context is DeepMind’s expanding AI-for-science portfolio, which also includes AlphaFold — the protein-structure model that earned Demis Hassabis and John Jumper the 2024 Nobel Prize in Chemistry.
The platform could help scientists identify the genetic drivers of disease more efficiently and may ultimately accelerate the development of new treatments, though the practical impact will depend on how researchers apply its predictions.
Source: The Verge