Google DeepMind launches AlphaGenome Atlas for predicting DNA changes

In short: Google DeepMind has launched AlphaGenome Atlas, a tool that predicts what could happen if any single “letter” in human DNA is changed.

What happened

Google DeepMind announced a new online resource called AlphaGenome Atlas. The company says it is a “predictive map” covering every possible single-letter change across the human genome.

DNA uses four chemical “letters”, A, C, G, and T. The human genome has about three billion letter pairs, and there are roughly nine billion possible one-letter substitutions. Figuring out which tiny changes matter is hard, a bit like finding which single typo in a huge library changes the meaning of a whole paragraph.

DeepMind says Atlas predicts how each of those potential changes could affect biology at a molecular level, for example by changing how much of a protein the body makes. It is based on AlphaGenome, a model DeepMind introduced last year, and it expands beyond protein-coding regions to include much of the DNA that helps control how genes behave.

Google is also releasing a Variant Impact Score, or AVI, to help researchers rank which DNA changes look most important so they can decide what to study first. DeepMind said the full Atlas dataset is about 1 petabyte in size (roughly like hundreds of thousands of high-definition movies).

The company says Atlas is available starting today for noncommercial research use through its website. Commercial access is planned “soon” through Google Cloud.

Why it matters

This does not diagnose anyone by itself, but it could speed up early research into why certain genetic changes are linked to disease. If researchers can narrow down the most likely harmful mutations faster, it may help them design better studies and, over time, support work on new treatments.

Source: The Verge AI