AlphaGenome Atlas ships 1-petabyte precomputed SNV impact map covering 9 billion variants

The Atlas pre-calculates AlphaGenome outputs across all 9 billion possible SNVs and exposes them through a no-code portal. At the Broad Institute, the AVI score flagged a DNM1 splice-site variant that resolved one unsolved rare-disease case. In UK Biobank data from 54 000 participants, variant grouping by predicted molecular effect yielded 22 percent additional non-coding associations and 19 BMI-linked loci in the top 1 percent of AVI scores.

Existing maps such as ENCODE and GTEx annotate observed states; AlphaGenome Atlas instead supplies exhaustive in-silico perturbation data. This shifts the bottleneck from variant annotation to experimental validation and statistical power for ultra-rare alleles. The single-score reduction, however, collapses multiple regulatory layers into one metric whose calibration across cell types remains untested at scale.

Operational deployment therefore requires coupling AVI-ranked candidates with targeted assays and larger cohorts. Integration with pangenome references and single-cell atlases will determine whether the 1-petabyte resource accelerates causal-variant discovery or merely increases the volume of hypotheses requiring wet-lab triage.

Next milestones include public API release and cross-cohort replication studies scheduled for 2025.