Infrastructure · Biology · Medicine
AlphaGenome Atlas releases genome-wide variant predictions with research case studies
A genome-wide catalogue of predicted molecular effects for nine billion single-nucleotide variants, accompanied by variant-prioritization and experimental research case studies.
Summary
DeepMind released AlphaGenome Atlas with precomputed variant effects and interpretable impact scores. Collaborators report experimental confirmation of a DNM1 splice-altering prediction and increased detection of non-coding genetic associations in UK Biobank analyses. This is a downstream resource and application release distinct from the original AlphaGenome model.
AI role
AlphaGenome predicts regulatory consequences; an impact score combines these with AlphaMissense to prioritize variants for association analyses and laboratory follow-up.
Narrative role
The combination of genome-wide predictions and concrete collaborator findings shows how a model becomes a usable research resource beyond its initial benchmark publication.
Caveat
Most entries are predictions, not validated variants. Case studies are reported by the developer and collaborators; the linked large paper could not be inspected through the web parser. No clinical efficacy or field-wide causal acceleration is established.