Google DeepMind launches AlphaGenome Atlas, a map of all 9 billion human DNA variants

Google DeepMind has released AlphaGenome Atlas, a database that uses its AlphaGenome AI model to predict the regulatory effect of every possible single nucleotide variant in the human genome: about 9 billion single-letter changes in total, producing a 1-petabyte dataset. The human genome runs to roughly 3 billion base pairs of DNA, and scientists understand the 2% that codes for protein reasonably well, but have only limited knowledge of the remaining 98% of non-coding DNA. AlphaGenome had already shown that single changes in these non-coding regions can disrupt processes like protein production; the Atlas turns that into a pre-calculated, queryable resource covering the whole genome. To make the results usable, the Atlas introduces the AlphaGenome Variant Impact (AVI) score, a single number combining predictions across coding and non-coding regions so researchers can prioritize variants without sorting through thousands of individual data points. Google DeepMind points to two examples of the Atlas already being used in research. At the Broad Institute, Laura Covill and her team used the AVI score to prioritize variants for an unsolved rare disease case; the tool flagged a variant in the DNM1 gene, predicting it created an incorrect splice site, which provided evidence that helped solve the case. Separately, Dr. Gareth Hawkes applied AlphaGenome Atlas to data from more than 54,000 UK Biobank participants, grouping variants by their predicted molecular effects. That approach uncovered 22% more non-coding genetic associations than his prior method, and by focusing on the top 1% of the most impactful variants he identified 19 genetic regions linked to body mass index (BMI). AlphaGenome Atlas is available now through a website portal that Google DeepMind says requires no coding skills, aimed at opening access to clinical researchers and biologists directly rather than only to computational specialists.

Key facts

AlphaGenome Atlas pre-calculates the predicted regulatory effect of all 9 billion possible single nucleotide variants in the human genome, a 1-petabyte dataset built with the AlphaGenome AI model.
It introduces the AlphaGenome Variant Impact (AVI) score, a single combined metric for coding and non-coding regions meant to help researchers prioritize variants quickly.
At the Broad Institute, Laura Covill's team used the AVI score to flag a DNM1 gene variant with an incorrect splice site, providing evidence that helped solve a previously unsolved rare disease case.
Dr. Gareth Hawkes applied the Atlas to over 54,000 UK Biobank participants, finding 22% more non-coding genetic associations and identifying 19 genetic regions linked to BMI among the top 1% of impactful variants.
The Atlas is available now through a no-code website portal aimed at clinical researchers and biologists rather than only computational specialists.

Why it matters

Only about 2% of the human genome codes for protein; the other 98% is non-coding DNA whose function is still poorly understood, even though single changes there can disrupt processes like protein production. Instead of leaving researchers to compute variant effects one at a time, AlphaGenome Atlas pre-calculates predictions for all 9 billion possible single-letter changes across the genome and packages them into a searchable database with a single summary score, the AVI score. That shifts genomic variant analysis from a bespoke computational task to a lookup, which is what let it contribute to solving an actual rare disease case and to a complex-trait study, rather than remaining a research demo.

Who it affects

The Atlas is aimed at clinical researchers and biologists working on rare diseases and complex traits, including groups without in-house computational genomics expertise. The two cited users are an academic rare-disease team at the Broad Institute (Laura Covill and colleagues) and a researcher, Dr. Gareth Hawkes, applying it to large-scale population data from the UK Biobank.

How to use it

AlphaGenome Atlas is available today through a website portal that Google DeepMind describes as requiring zero coding skills, intended to let clinical researchers and biologists query predictions directly rather than running the AlphaGenome model themselves. The source gives no pricing, API access, or rate-limit details beyond describing it as a website portal.

How solid is it

The evidence offered is two applied case studies rather than a benchmark against other variant-effect tools. In the Broad Institute case, the AVI score's flagged DNM1 variant provided supporting evidence that helped solve a previously unsolved rare disease case, though the source does not specify the disease involved or how the finding was clinically confirmed. In the UK Biobank case, grouping variants by predicted molecular effect using the Atlas found 22% more non-coding genetic associations than Dr. Hawkes's prior approach, and narrowing to the top 1% most impactful variants surfaced 19 regions linked to BMI. No comparison figures against a prior AlphaGenome release or other variant-effect prediction tools are given.

Risks and caveats

The source is a Google DeepMind announcement describing its own product, so the case studies are selected examples rather than independent validation. Key details are absent from the text: no institutional affiliation is given for Dr. Hawkes beyond his name, no numeric scale or range is defined for the AVI score itself, and the DNM1 case does not specify the underlying disease or the clinical confirmation process. As with any variant-effect prediction tool, an AVI score is a computational prediction that flags candidates for further investigation, not a standalone diagnosis.