Loading…
Google DeepMind launched the AlphaGenome Atlas on 8 September 2026 as an open-access database of precomputed AI predictions for all 9 billion possible single-letter mutations in human DNA.
The dataset is about 1 petabyte, which DeepMind says is more than 30 times the size of the AlphaFold Database.
It introduces the AlphaGenome Variant Impact (AVI) score, a single metric for ranking both coding and non-coding mutations.
The Atlas focuses on non-coding DNA, which forms about 98% of the human genome and contains many disease-linked variants.
It is free for non-commercial research; DeepMind states that AlphaGenome has not been validated or approved for clinical use.
Coding DNA carries protein-making instructions. The remaining ~98% is non-coding, and includes regulatory regions such as enhancers and promoters that decide when genes switch on and off. Many disease-linked variants sit in this region, outside any protein-coding sequence.
Simple Analogy: The genes are the recipes; the non-coding DNA is the kitchen timer.
Artificial intelligence research laboratory that developed AlphaFold, AlphaGenome and the AlphaGenome Atlas.
DeepMind's protein structure prediction system, presented as AlphaFold2 in 2020. The AlphaFold Database holds over 200 million predicted protein structures; the AlphaGenome Atlas is described as more than 30 times larger.
Awarded one half to David Baker for computational protein design, and the other half jointly to Demis Hassabis and John Jumper of Google DeepMind for protein structure prediction.
Launched 25 June 2025, it is a sequence-to-function model: it takes DNA up to a million letters long and predicts thousands of molecular properties describing that stretch's regulatory activity.
Laura Covill's team at the Broad Institute used the AVI score to identify a rare-disease mutation in the DNM1 gene, and Dr Gareth Hawkes used the Atlas on data from more than 54,000 UK Biobank participants.
GS Paper 3 > Science & Technology > Biotechnology and Artificial Intelligence
General Awareness > Science & Technology
A change in a single letter of the DNA sequence, also called a point mutation. There are about 9 billion possible such changes across the human genome.
Regulatory stretches of non-coding DNA that control when and how strongly a gene is expressed.
AlphaGenome Variant Impact score - a single metric that lets coding and non-coding mutations be ranked against each other in one framework.