Google DeepMind expands from proteins to the human genome with AlphaGenome Atlas

Google DeepMind has introduced the AlphaGenome Atlas, an AI-powered map for studying how billions of possible DNA changes could affect the human body. The new tool could help researchers better understand diseases and speed up genetic research.

In Short

• Google DeepMind introduces AlphaGenome Atlas
• It can learn patterns between DNA changes and our human body
• It is available for academic research through a free web portal.

Eight years ago, Google’s DeepMind began its investigation into the science of life. Year 2024 saw its co-founder Demis Hassabis and researcher John Jumper win the Nobel Prize in Chemistry for a historic breakthrough in predicting protein 3D structures with AlphaFold. Now, DeepMind is taking that success one step further, moving from understanding proteins to exploring something bigger – the human genome.

The company has introduced AlphaGenome Atlas, an AI-powered platform designed to help scientists understand what happens when the tiny letters that make up our DNA change. And how it affects the human body. In the coming days, this new tool could broaden the horizons of biology, accelerating research and ultimately leading to new ways to treat diseases.

A map of billions of DNA changes

Think of DNA as a huge instruction book for the human body. It is written using just four chemical letters: A, C, G and T. The human genome contains roughly three billion pairs of these letters.

A tiny change in just one letter may mean nothing at all. But in some cases, it can affect how our genes work and may contribute to disease.

The problem is scale. There are roughly nine billion possible single-letter changes that researchers could study. Testing every one of them in a laboratory would be an enormous task. And this is where AlphaGenome Atlas comes in.

DeepMind says the platform contains predictions for how all nine billion possible single-letter DNA changes could affect molecular processes in the body. Instead of scientists having to start from scratch, the Atlas gives them a giant map to explore.

Finding the DNA changes that matter

Of course, giving scientists nine billion possibilities is not exactly making their job easier. They still need to figure out which changes deserve attention.

To help with that, DeepMind has introduced the AlphaGenome Variant Impact, or AVI, score. It gives each genetic variant a score based on its predicted impact, helping researchers rank changes and focus on the ones that may be more important.

The system can look at both the small part of the genome that directly helps make proteins and the much larger portion that does not. That second part is important because it can help control when and how genes work.

From rare diseases to common traits

DeepMind says researchers working on unsolved rare diseases used the AVI score to identify a genetic variant linked to the DNM1 gene, which is associated with epileptic encephalopathy. Further experiments supported the prediction and showed how the variant could interfere with the way genetic instructions are processed.

Researchers have also used the Atlas to study genetic changes linked to common traits. One study involving more than 54,000 UK Biobank participants found 22 per cent more non-coding genetic associations by grouping rare variants according to their predicted molecular effects.

A giant AI-powered genome map

DeepMind says AlphaGenome Atlas is a 1-petabyte dataset, making it more than 30 times larger than the AlphaFold Database. The company is making the Atlas available for non-commercial research through its website, while commercial access through Google Cloud is expected later.

AlphaFold helped scientists understand the shapes of proteins. AlphaGenome Atlas is now trying to help them understand how changes in the DNA code can influence biology.

- Ends