Google DeepMind introduced a new tool this week called AlphaGenome Atlas. It is designed to help scientists understand how changes in human DNA affect the body.
The human genome contains about 3 billion base pairs of DNA. Only 2% of that DNA codes for proteins, which scientists already understand fairly well.
The other 98% is much harder to study. Scientists have struggled to figure out what this non-coding DNA actually does.
Google DeepMind had already built a model called AlphaGenome. It showed how single changes in non-coding DNA regions can affect processes like protein production.
What the Atlas Does
The new Atlas takes that work further. It uses the AlphaGenome model to pre-calculate the effects of all 9 billion possible single-letter genetic changes in the human genome.
That process created a massive dataset, about 1 petabyte in size. To put that in perspective, one petabyte can hold roughly 500 billion pages of text.
To make the data easier to use, the Atlas includes a new scoring system called the AlphaGenome Variant Impact score, or AVI. This single score combines predictions from both coding and non-coding regions of DNA.
Researchers can use the AVI score to quickly identify which genetic variants are worth studying further. This saves them from manually sorting through thousands of data points.
Early Research Results
The tool is already being used in real research projects. At the Broad Institute, researcher Laura Covill and her team used the AVI score to study unsolved rare disease cases.
The tool flagged a variant in the DNM1 gene. It predicted the variant created an incorrect splice site in the DNA.
That finding gave the team evidence needed to solve a previously unsolved case. The gene is linked to certain rare neurological conditions.
Separately, Dr. Gareth Hawkes used the Atlas to study genetic data from more than 54,000 participants in the UK Biobank. He grouped variants based on their predicted molecular effects.
That approach uncovered 22% more non-coding genetic associations than earlier methods found. Hawkes then focused on the top 1% of the most impactful variants.
That analysis identified 19 genetic regions linked to body mass index. Researchers said this will help guide future studies into the topic.
Google DeepMind said the Atlas is built for scientists who do not have coding experience. It is available through a website portal that requires no programming skills.
The company said the goal is to make genomic research tools available to clinical researchers and biologists around the world. The Atlas launched on September 8, 2026.
Pushmeet Kohli, VP of Science at Google DeepMind, and Žiga Avsec, who leads the company's Genomics Initiative, announced the release. Google DeepMind said it plans to continue expanding tools in this area.
