Google DeepMind maps the effects of genetic mutations

Google announced AlphaGenome Atlas on September 8, introducing a predictive map of every possible single-letter DNA change in the human genome. Developed by Google DeepMind, the AI model predicts the effects of all 9 billion such changes, helping researchers investigate how genetic variations influence molecular biology and contribute to disease.

The resource builds on DeepMind’s work with AlphaFold, which predicts protein structures. AlphaGenome Atlas is a 1-petabyte dataset, 30 times larger than the AlphaFold dataset, and is available for noncommercial use only. Alongside its predictions, it provides an AlphaGenome Variant Impact score that helps prioritize mutations and assess their genetic significance.

Researchers investigating epileptic encephalopathy have already used the model to identify a disease-linked mutation in the DNM1 gene. Its impact score also indicated how the mutation functioned, illustrating how the resource could help connect genetic changes with disease mechanisms.

Understanding those relationships could support the development of gene-editing treatments and help researchers investigate the unintended effects of altering DNA. Combined with protein-structure predictions, the atlas offers life-sciences researchers another tool for studying disease and identifying potential therapeutic approaches.