"A predictive map of every possible DNA letter change in the human genome." DeepMind precomputed AlphaGenome's molecular-effect predictions for all 9 billion single-nucleotide variants, a one-petabyte dataset it describes as more than 30 times the AlphaFold Database, and added an AlphaGenome Variant Impact (AVI) score that condenses AlphaGenome and AlphaMissense predictions into one number for ranking coding and non-coding variants. Each variant carries thousands of regulatory-effect predictions, AVI feature attributions that point at disrupted processes such as splicing, and matches against more than 2,500 DNA sequence motifs. Free for academic use through a portal and API, with commercial access to follow on Google Cloud; collaborators include the Broad Institute, Boston Children's, Stowers, Exeter, Harvard, MSK, and MGH. The underlying AlphaGenome model (2025) is not yet tracked here.

Dataset

sciencebiologydataset

Related