Science

DeepMind releases AlphaGenome Atlas predicting 9 billion human genetic variants

Google DeepMind has launched a 1-petabyte repository mapping all single-letter mutations in a reference genome, Hacker News reports.

Updated Sep 11, 2026 · 19:47 UTC1 reports

The short version

Google DeepMind publicly released the AlphaGenome Atlas on September 8, providing precomputed AI predictions for all 9 billion single-letter changes in a reference genome.[Hacker News]
The 1-petabyte repository enables researchers to evaluate genetic variants through a simplified impact score.[Hacker News]
Researchers note that the model is computationally heavy and limited to a 1 million base pair range, which may miss distant enhancer effects.[Hacker News]

Google DeepMind publicly released the AlphaGenome Atlas on September 8.[Hacker News]
The repository contains precomputed predictions covering all 9 billion possible single-letter alterations in a reference human genome.[Hacker News]
The entire dataset encompasses approximately 1 petabyte of information.[Hacker News]
AlphaGenome is constrained by a 1 million base pair field of view, limiting its capacity to map distant regulatory enhancer sequences.[Hacker News]

Carl de Boer warned that the Atlas's single-number variant impact score is prone to misinterpretation by researchers.[Hacker News]

Outlet counts describe coverage, not independent confirmation. Reports may share a wire service or original source.