A searchable map of possible changes
Google DeepMind announced AlphaGenome Atlas on September 8, 2026, the same day The Verge reported the launch. According to DeepMind’s announcement, the resource precomputes predicted molecular effects for nine billion possible single-letter DNA variants.
The accompanying AlphaGenome Variant Impact score combines predictions from AlphaGenome and AlphaMissense to help rank variants. The number describes possible changes, not nine billion patients or laboratory-confirmed discoveries.
What researchers can use
Google’s overview frames the problem as understanding both protein-coding DNA and the much larger non-coding portion. A change outside a protein’s blueprint can still affect how genes are regulated.
The research platform offers an interactive portal and API access, alongside skills for Google Antigravity. That gives computational and browser-based workflows different entry points into the resource.
The practical distinction
TechKili analysis: a ranked candidate is a starting point for investigation. Readers should distinguish a predicted molecular effect from evidence that a variant causes a particular disease or that a treatment works. The useful question for research teams is which hypotheses to test next, and what independent evidence would confirm them.