Google DeepMind has introduced AlphaGenome Atlas, a database containing predictions for the effects of all 9 billion possible single-letter genetic changes in the human genome. The dataset, which spans 1 petabyte of information, was generated by pre-calculating regulatory impacts using the AlphaGenome AI model.
The tool addresses a significant gap in genetic understanding. While scientists have well-characterized the 2% of the human genome that codes for proteins, the remaining 98% has remained largely mysterious. According to Google DeepMind, the AlphaGenome model has already demonstrated how single changes in non-coding DNA regions can disrupt molecular processes like protein production, but a comprehensive map was previously unavailable.
The Atlas introduces the AlphaGenome Variant Impact (AVI) score, which combines predictions for both coding and non-coding regions into a single metric. This allows researchers to quickly prioritize variants for investigation without manually reviewing thousands of data points.
Google DeepMind highlighted early applications of the tool. At the Broad Institute, Laura Covill’s team used the AVI score to identify a critical variant in the DNM1 gene for unsolved rare disease research. The tool predicted the variant created an incorrect splice site, providing evidence that helped solve the case.
Dr. Gareth Hawkes applied AlphaGenome Atlas to data from over 54,000 UK Biobank participants studying complex traits. By grouping variants based on predicted molecular effects, he uncovered 22% more non-coding genetic associations than previous methods. When focusing on the top 1% of impactful variants, he identified 19 genetic regions linked to body mass index, directing subsequent targeted research.
The database is accessible through a web portal that requires no coding skills, according to Google DeepMind. This design aims to democratize access for clinical researchers and biologists worldwide. Google DeepMind frames the release as part of its commitment to accelerate genomic discovery and science globally.
Key facts
- AlphaGenome Atlas contains predictions for all 9 billion possible single nucleotide variants in the human genome
- The database spans 1 petabyte of data pre-calculated using the AlphaGenome AI model
- The AlphaGenome Variant Impact (AVI) score combines predictions for coding and non-coding regions into a single metric
- At the Broad Institute, the tool helped identify a disease-causing variant in the DNM1 gene
- A study using UK Biobank data found 22% more non-coding genetic associations using the Atlas compared to previous methods
- The tool is accessible through a no-code web portal designed for researchers without programming skills