ADVERTISEMENT

DeepMind Launches AlphaGenome Atlas, an AI Map of 9 Billion Genetic Mutations

DeepMind Launches AlphaGenome Atlas, an AI Map of 9 Billion Genetic Mutations

Google DeepMind has launched AlphaGenome Atlas, a massive open-access database that maps how every possible single-letter DNA mutation impacts human biology. By combining two advanced AI systems, this breakthrough tool allows scientists to quickly identify harmful genetic changes – and pinpoint the root causes of unsolved diseases. The launch marks a major leap forward in genetic research, making complex molecular predictions freely available to medical researchers worldwide.

At its core, AlphaGenome Atlas is a platform containing predictions for the effects of 9 billion single-nucleotide variants — every single-letter change possible — in the human genome, according to DeepMind news.

AlphaGenome Atlas works like a giant biological dictionary. It contains predictions for all 9 billion possible single-letter DNA changes in the human genome. Testing these mutations individually in a lab would take decades – but DeepMind precomputed every result so researchers can instantly look up how any mutation alters cell behavior.

To make this massive dataset easy to navigate, DeepMind introduced the AlphaGenome Variant Impact (AVI) score. This score combines predictions from two AI models into a single number that ranks how dangerous a genetic change might be. It works for both the 2% of our DNA that makes proteins – and the remaining 98% that controls gene activity.

Scientists are already using the atlas to solve medical mysteries. In rare disease research, scientists used the AVI score to find a hidden genetic mistake in a gene linked to severe childhood epilepsy. Laboratory tests confirmed that the AI accurately predicted how the mutation disrupted normal cell instructions.

The tool also helps researchers study common health conditions across large populations. By analyzing genetic data from over 54,000 people, scientists uncovered 22% more genetic links than traditional methods could detect. This approach helps identify specific non-coding variants connected to aging, oxygen processing, and body mass index.

AlphaGenome Atlas is available for free to academic researchers through an online website portal, an API, and Google Antigravity. By making 1 petabyte of genetic predictions accessible to everyone, DeepMind aims to speed up medical discoveries and help create new treatments for genetic disorders.