Google DeepMind launched the AlphaGenome Atlas on 8 September 2026 as an open-access database for human DNA analysis. The resource contains 1 petabyte of precomputed artificial intelligence predictions for all 9 billion possible single-letter mutations in human DNA.
AlphaGenome Atlas and Human Genome
The human genome contains about 3 billion base pairs, and single-letter mutations are also called point mutations or single-nucleotide variants. The AlphaGenome Atlas is designed to map both coding DNA, which carries protein-making instructions, and non-coding DNA, which forms about 98% of the human genome.
Non-coding DNA and Gene Regulation
Non-coding DNA includes regulatory regions such as enhancers, promoters, and other control elements that influence when genes switch on and off. The Atlas focuses on these regions because many disease-linked variants occur outside protein-coding sequences.
AlphaGenome Variant Impact Score
Google introduced the AlphaGenome Variant Impact (AVI) score as a unified metric for ranking coding and non-coding mutations. The score is intended to help researchers compare variant effects across different genomic regions using one framework.
Important Facts for Exams
- The AlphaGenome Atlas is an interactive web-based resource that works in a standard browser without coding.
- The database is free for academic researchers.
- The DNM1 gene is associated with a rare disease mutation that can disrupt splicing.
- The UK Biobank is a large biomedical database that includes genetic data from more than 54,000 participants in the cited analysis.
Research Collaborations and Genomic Analysis
Melanie Weilert and Dr. Julia Zeitlinger of the Stowers Institute for Medical Research co-published a preprint on bioRxiv on 9 September 2026 on regulatory DNA patterns. Laura Covill’s team at the Broad Institute used the AVI score to identify a rare disease mutation in the DNM1 gene.
Applications in Population Genomics
Dr. Gareth Hawkes used the Atlas to analyse genomic data from more than 54,000 UK Biobank participants. The analysis found 22% more non-coding genetic associations and identified 19 genetic regions linked to body mass index.
