Google Releases AlphaGenome Atlas to Predict Effects of 9 Billion Human Genome Variants

Google DeepMind has released AlphaGenome Atlas, a database with precomputed regulatory impacts for 9 billion single nucleotide variants (SNVs) in the human genome. Previously, only 2% of the 3 billion base pairs in the human genome, specifically the protein-coding regions, were well understood, but this tool predicts the impact of variants in the remaining 98% non-coding regions on molecular processes.

Key Features and Technology

AlphaGenome Atlas is a 1 petabyte dataset that introduces the AlphaGenome Variant Impact (AVI) score. This score is a single metric combining predictions from both coding and non-coding regions, allowing researchers to quickly prioritize the most promising research directions without manually reviewing thousands of data points.

Real-World Research Cases

• Rare Disease Diagnosis: The team led by Laura Covill at the Broad Institute used AVI scores to analyze unresolved rare disease variants. They contributed to solving cases by predicting and confirming variants in the DNM1 gene that create incorrect splice sites.
• Complex Trait Research: Dr. Gareth Hawkes analyzed data from over 54,000 individuals in the UK Biobank. By grouping variants based on predicted molecular effects, they discovered 22% additional non-coding genetic associations and identified 19 genetic regions associated with BMI among the top 1% of impactful variants.

Accessibility

It is available via the alphagenome.google/atlas web portal, making it easily accessible to clinical researchers and biologists worldwide without requiring coding skills.