Google DeepMind has unveiled an AI tool that its scientists say could help unravel the mysteries of the human genome, accelerate scientific research and ultimately open the door to new treatments for diseases. The platform, called AlphaGenome Atlas, contains what researchers describe as a predictive map of every possible DNA letter change in the human genome.
DNA is written using an alphabet of four chemical letters, usually shortened to A, C, G and T, and the human genome contains roughly three billion letter pairs. These letters carry the instructions that make life work, including how, when and where genes are switched on and off, and to what degree. Changes to individual letters can be harmless, contribute to ordinary differences between people, or play a role in disease. Identifying which changes matter is a formidable task, given there are roughly nine billion potential single-letter substitutions.
A genome-wide predictive map
Atlas contains predictions for how each of these nine billion variants could affect the body at a molecular level, such as altering how much of a particular protein is produced. The researchers describe it as the most comprehensive catalogue of how genetic mutations affect molecular biology. Scientists can explore these predictions through a web portal, as a skill within the agentic development platform Antigravity, and via the AlphaGenome interface.
To help researchers sift through the billions of possibilities and concentrate on the mutations that merit closer attention, Google is also releasing a Variant Impact Score, known as AVI, which draws on the company’s other models for predicting the effects of DNA changes. According to Google, researchers can now rapidly rank variants and interpret their molecular effects at the same time.
Building on earlier AI models
The project builds on AlphaGenome, an AI model DeepMind introduced last year to help scientists identify the genetic drivers of disease, as well as AlphaMissense, an earlier tool focused on predicting which small mutations might alter proteins. Atlas extends predictions across the genome, including the vast majority of stretches that do not directly code for proteins but can instead control how genes behave.
In a press briefing, Ziga Avsec, DeepMind’s genomics lead, acknowledged that the underlying AlphaGenome model had already been released, but said turning its capabilities into a genome-wide catalogue took time. He explained that precomputing and analysing this many variants was demanding because the space involved is so large.
AlphaGenome was trained using public databases of human and mouse genomes, enabling it to learn patterns between DNA changes and biological processes. Applying those predictions to billions of possible variants produced a vast dataset that Google says is roughly one petabyte in size. The company is making Atlas available to researchers for non-commercial use through its website from today, and for commercial use on Google Cloud in due course.
Source
Image: theverge.com
