Google DeepMind unveils AlphaGenome Atlas mapping every possible human DNA letter change

Google DeepMind on Tuesday unveiled AlphaGenome Atlas, a platform containing predictions for how each of roughly 9 billion possible single-letter DNA changes could affect biology at a molecular level. The company called it a predictive map of every possible DNA letter change in the human genome and the most comprehensive catalogue of how genetic mutations affect molecular biology. DNA is written in an alphabet of four chemical letters — A, C, G, and T — and the human genome contains roughly three billion letter pairs, creating about nine billion potential single-letter substitutions that cannot practically be tested one by one in the lab. The Atlas is a 1-petabyte dataset, more than 30 times larger than the AlphaFold Database. It was built by precomputing predictions from AlphaGenome, an AI model DeepMind released last year to forecast how genetic variants impact biological processes, including in stretches of DNA that do not directly code for proteins but can control how genes behave. DeepMind is also releasing the AlphaGenome Variant Impact, or AVI score, which condenses predictions from AlphaGenome and AlphaMissense — a model for protein-altering DNA variants — into a single number so researchers can rank variants and interpret their molecular effects at the same time. AlphaGenome was trained using public databases of human and mouse genomes. Ziga Avsec, DeepMind’s genomics lead, said precomputing and analyzing this many variants took time because the space is so big. AlphaGenome Atlas is available today for academic research through a free website portal, the AlphaGenome API, and as a skill in Google Antigravity. Trusted external collaborators have already used it to identify and experimentally verify key variants in unsolved rare disease research and to find rare variants associated with common traits. Predictions cover molecular effects such as changing how much of a particular protein is produced. By turning AlphaGenome into an easily accessible genome-wide resource, DeepMind said it wanted to give researchers a big-picture view of variants across the entire genome rather than limiting analysis to specific variants one at a time.