Google DeepMind on Tuesday released AlphaGenome Atlas, a searchable database predicting the molecular effects of all 9 billion possible single-letter substitutions in human DNA. The free, non-commercial resource is meant to help scientists identify which tiny changes among the genome’s vast background noise deserve lab attention.¹

The launch tackles a problem left open after the Human Genome Project mapped the sequence: knowing the letters is not the same as understanding their consequences. As DeepMind research chief Pushmeet Kohli put it, “we bought the book, but we did not understand how to read it.” The atlas combines predictions across gene regulation, splicing and protein effects into an AlphaGenome Variant Impact score, including for the non-coding 98% of DNA that has been especially difficult to interpret.¹

Google framed access as the point. Sundar Pichai said the browser-based tool requires no coding and is free for academic researchers, promising “discoveries to come.”Sundar Pichai (sundarpichai)² DeepMind also said commercial access would follow through Google Cloud licensing, leaving the terms unspecified.¹

Before release, beta users offered evidence for the upside. In one unresolved case of epileptic encephalopathy, Atlas highlighted a DNM1 variant whose predicted splicing effect was later confirmed experimentally and reclassified as likely pathogenic. In a separate analysis of UK Biobank data, researchers found 22% more associations after filtering variants through the tool; one region fell from 526 candidates to four. “We can use it to shrink the haystack,” said University of Exeter researcher Gareth Hawkes.¹

Yet DeepMind’s own genomics lead, Žiga Avsec, drew the boundary sharply: the outputs are “accurate enough to really point us in the right direction,” but not “the universal truth.” The atlas can miss some effects, particularly in enhancers, and remains only one link in the evidence chain for a clinical diagnosis.¹

That restraint matters. The company calls Atlas a “predictive map of every possible DNA letter change,” a potentially transformative research shortcut—but one whose most consequential findings still have to survive the bench.³