Atlases have long helped humans navigate unfamiliar terrain. And now a new atlas from Google DeepMind aims to better map the terrain of our biology: the human genome.
On Tuesday the company released AlphaGenome Atlas, a database that charts predictions of the possible effects of nine billion single variants in the human genome.
The database is based on the AlphaGenome, a DeepMind artificial intelligence model released last year that analyzes stretches of noncoding DNA, which primarily regulates gene activity. The model then predicts how much those variants affect molecular function. With the new atlas, all that analysis has been done ahead of time, and researchers can search for any DNA variant to see its predicted effects down to the level of specific tissues.
On supporting science journalism
If you’re enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
“This represents the first time that any researcher in the world can access a comprehensive map of the human genome and its variations by simply opening a browser,” said Pushmeet Kohli, vice president of science at Google DeepMind, in a press briefing.
Researchers have been puzzled by the intricacies of the genome, especially noncoding DNA—the 98 percent of DNA that doesn’t code for specific proteins but still plays a major role in the body’s function. AlphaMissense, another DeepMind tool, makes predictions for coding DNA, or the genes that make proteins. The new atlas covers everything else and contains a whopping petabyte of data—a quadrillion bytes.
Single variants in coding DNA can cause diseases such as Tay-Sachs and sickle cell anemia. Noncoding DNA variants have also been linked to some types of cancer. To trace these mutations, researchers typically must painstakingly search through billions of data points to search for signals, but AlphaGenome Atlas could offload a lot of that work. Jonathan Sebat, a psychiatric geneticist at the University of California, San Diego, says it could help speed up such research.
“Our own workflows in the lab can be streamlined quite a bit because we don’t actually have to compute anything,” Sebat says. “We literally can just look up everything.”
In some ways, the atlas acts as a successor to the database for AlphaFold, an earlier DeepMind project that effectively solved the protein prediction problem, winning then DeepMind chief Demis Hassabis a share of the 2024 Nobel Prize in Chemistry. It has since been used by scientists worldwide for disease research and drug discovery. AlphaGenome Atlas, by contrast, focuses on DNA—the instructions to make proteins and regulate their function—not proteins themselves. The new atlas is also 30 times larger than AlphaFold’s database, although it is far less accurate.
Rather the new database is meant to serve as a starting point, says Žiga Avsec, genomics lead at Google DeepMind. It may be especially useful for research questions that involve more than one variant across the genome, for example, he says.
While AlphaFold’s database was free to use—which was reportedly the cause of some recent shake-ups at DeepMind, including Hassabis’s move to Alphabet, Google’s parent company—commercial users such as drugmakers will have to license AlphaGenome Atlas.
The new atlas also gives genetic variants a so-called AlphaGenome Variant Impact (AVI) score. The measure combines the AlphaGenome and AlphaMissense predictions and ranks each variant by how much it affects function.
The database looks at single-letter changes in the genome, meaning a single change in the base pair—if you imagine that a double helix of DNA is a twisting ladder, a base pair is a rung on that ladder. These tiny changes won’t be able to give the full picture of all genetic diseases on their own, but they can provide clues.
“If somebody is studying a disease, and they don’t have any idea about what cell types to look for or what molecular processes are impacted, then starting with an AVI score, which is a single score, is a great starting point to help you prioritize variants and try to find that needle in the haystack,” Avsec says.
Subscribe to Support Independent Journalism
Great science journalism requires human expertise, time, effort and creativity. And it costs money. That’s why I and the journalists here at Scientific American hope you’ll join our community.
When you subscribe, you are supporting staff and freelance journalists who are passionate about telling science stories that are true, important and compelling. Our editors and reporters are often experts in their fields, which means they understand the nuances of big discoveries and can untangle the breakthroughs from the hype. With a subscription, you are also supporting rigorous fact-checking to ensure the words we publish are precise and accurate. And you’re supporting original illustrations, graphics and photos that bring you closer to an advanced laboratory, an ice sheet in Antarctica or a space mission in orbit. You’re helping us craft other types of high-quality journalism as well: Our newsletters are carefully written, edited and curated by staffers you have or will come to know and love. Our Science Quickly podcast is based on original reporting, collaboration with editors and scientists and exacting production.
Subscriptions keep this engine running so we can continue to deliver thoughtful, rigorous and independent science journalism to you. In an era of viral misinformation, this work is crucial. If you value what we do, I hope you’ll consider joining us as a subscriber
