The work behind AlphaGenome Atlas comes from a large team at Google DeepMind, led by researchers working alongside scientists from the University of Exeter Medical School, the Broad Institute of MIT and Harvard, Boston Children’s Hospital, the Stowers Institute for Medical Research, and several other research centers in the US and UK. Together, they have built something that sounds almost impossible: a catalog of the likely effects of every single possible one-letter change across the entire human genome.
Why This Matters
Our DNA is made of about 3 billion letters, and there are roughly 9 billion ways a single letter could change. Each of these changes, called a variant, might do nothing at all, or it might disrupt a gene, alter how a protein is made, or throw off a process that keeps a cell running properly. Figuring out which variants actually matter has been one of the toughest problems in genetics. Testing each one by hand in a lab would take lifetimes and resources nobody has.
Only about 2% of our genome codes for proteins. The other 98% doesn’t build proteins directly, but it still controls when, where, and how strongly genes switch on. Scientists have struggled for years to make sense of this non-coding portion because the rules governing it are subtle and spread out across long stretches of DNA.
From a Single Model to a Full Map
The team had already built AlphaGenome, a model that predicts how a genetic variant affects biological processes. It works well when you want to check one variant at a time. But researchers often need a bigger picture, something that lets them scan across the whole genome and compare variants against each other.
The team ran AlphaGenome’s predictions across all 9 billion possible single-letter changes and organized the results into one enormous, searchable resource, about 1 petabyte of data in total. To put that in perspective, it’s over 30 times larger than the AlphaFold Database, which mapped the 3D shapes of proteins and became a go-to tool for biologists worldwide after it launched a few years back.
Alongside the raw predictions, the team introduced something called the AlphaGenome Variant Impact score, or AVI for short. It combines AlphaGenome’s predictions with another model, AlphaMissense, which focuses on protein-altering changes. The result is a single number that tells researchers how disruptive a given variant is likely to be, whether it sits in the coding 2% or the non-coding 98% of the genome. On top of that, each AVI score comes with an explanation of which biological process, such as splicing or gene expression, is being affected most.
The Atlas also includes a library of more than 2,500 recurring DNA sequences, sometimes described as the “words” of the genome, along with where they show up. These sequences often mark the spots where proteins bind to DNA and switch genes on or off.
Already Making a Difference
Researchers at the Broad Institute, working with the GREGoR Consortium on unsolved rare disease cases, used the AVI score to sift through thousands of candidate variants in one patient. They found a mutation in a gene called DNM1, tied to a severe form of childhood epilepsy. The prediction showed exactly how it went wrong: the variant created a faulty splice site that led to an abnormal, extended protein. Lab experiments backed up the finding.
At the University of Exeter, a researcher applied Atlas to genetic data from more than 54,000 people in the UK Biobank. By grouping rare variants according to their predicted molecular effects, he uncovered 22% more genetic associations than standard methods would have caught, including links to genes tied to aging and oxygen sensing in cells. He also used the tool to narrow down 19 genetic regions connected to body mass index.
And at the Stowers Institute, scientists used the motif library to sort out which DNA-binding proteins simply open up access to DNA versus which ones actually flip genes on or off, a distinction that matters a great deal for understanding how cells make decisions.
What Comes Next
AlphaGenome Atlas is free for academic use through a website, an API, and as a tool inside Google Antigravity. A version for commercial use on Google Cloud is expected soon. The team describes it as a starting point rather than a finished product, with the expectation that future versions of the underlying models will make the maps sharper and more reliable over time.
For a field that has long struggled to interpret the vast majority of the genome, having a resource like this, freely available and covering every possible variant, marks a real shift in how researchers can approach genetic disease and trait discovery. Rather than hunting for a handful of known suspects, scientists can now start with the whole genome in view.
Note: AlphaGenome Atlas is intended for research purposes and has not been validated or approved for clinical use.
Sources: Reference Paper | Reference Article
Disclaimer:
The research discussed in this article was conducted and published by the authors of the referenced paper. CBIRT has no involvement in the research itself. This article is intended solely to raise awareness about recent developments and does not claim authorship or endorsement of the research.
Important Note: This blog is based on a company announcement and a preprint that have not yet undergone peer review. As a result, it is important to note that these papers should not be considered conclusive evidence, nor should they be used to direct clinical practice or influence health-related behavior. It is also important to understand that the information presented in these papers is not yet considered established or confirmed.
Follow Us!
Learn More:
