Google DeepMind maps 9 billion DNA changes with AlphaGenome AtlasPlease

• Courtesy of Google

Google DeepMind has launched AlphaGenome Atlas, an artificial intelligence powered resource that maps the predicted molecular effects of every possible single letter change in the human genome, giving researchers a new tool to study genetic variation and disease.

The platform contains predictions for around nine billion possible single nucleotide variants across the human genome, making it what Google DeepMind describes as its most comprehensive catalogue yet of how genetic mutations could affect molecular biology.

AlphaGenome Atlas was built by running Google DeepMind’s AlphaGenome AI model at scale and precomputing its predictions across the genome. The resulting dataset is about one petabyte in size, more than 30 times larger than the AlphaFold Database.

While roughly two percent of the human genome directly codes for proteins, the remaining 98 percent includes non-coding regions that play important roles in controlling gene activity. Many genetic variants associated with diseases and physical traits are found within these regions.

AlphaGenome is designed to help researchers understand these areas by predicting how DNA sequences and genetic variants affect biological processes involved in gene regulation.

The model can analyze DNA sequences of up to one million base pairs and predict thousands of molecular properties, including RNA production and splicing, chromatin accessibility, protein binding and interactions between different parts of DNA.

Alongside the Atlas, Google DeepMind introduced the AlphaGenome Variant Impact or AVI score, which combines predictions from AlphaGenome and AlphaMissense into a single measure intended to help scientists prioritize genetic variants for further investigation.

Researchers have already begun testing the technology in areas including rare diseases and complex traits.

At the Broad Institute, researchers used the AVI score to narrow down potential variants in an unresolved rare disease case and identified a variant involving the DNM1 gene. Laboratory experiments later supported the model’s prediction, according to Google DeepMind.

The technology could also help researchers investigate genetic mechanisms associated with conditions such as cancer and Alzheimer’s disease by examining how variants affect gene regulation across different tissues and cell types.

However, Google DeepMind stressed that AlphaGenome Atlas is a research tool and has not been validated or approved for clinical use. Its predictions are not intended to replace professional medical advice, diagnosis or treatment.

AlphaGenome Atlas is available to researchers through a free web portal, the AlphaGenome API and AlphaGenome Skills in Google Antigravity. Google DeepMind also offers AlphaGenome through Google Cloud for commercial and enterprise research workflows.