The human genome consists of roughly three billion base pairs, but only about two percent of those sequences code for proteins, leaving the remaining 98 percent largely uncharted. According to the source, DeepMind’s AlphaGenome model had previously shown that single changes in non‑coding regions can disrupt molecular processes such as protein production, yet the broader impact remained unclear. To address this gap, the company has introduced AlphaGenome Atlas, a database that pre‑calculates the regulatory effect of every possible single‑letter change across the genome. Using the AlphaGenome AI model, researchers computed the impact of all nine billion single‑nucleotide variants, producing a dataset that occupies approximately one petabyte of storage. The Atlas enables scientists to query this massive collection quickly. To simplify interpretation, the source explains that the Atlas introduces the AlphaGenome Variant Impact (AVI) score, a single metric that combines predictions for both coding and non‑coding regions. This score lets researchers prioritize variants without having to examine thousands of individual data points. The source provides two concrete examples of how the tool is already being used. At the Broad Institute, Laura Covill and her team applied the AVI score to unsolved rare‑disease cases. The tool highlighted a critical variant in the DNM1 gene, predicting that it created an incorrect splice site, which supplied crucial evidence that helped solve the case. In a separate effort, Dr. Gareth Hawkes used AlphaGenome Atlas with genotype data from over fifty‑four thousand UK Biobank participants. By grouping variants according to their predicted molecular effects, he uncovered 22 percent more non‑coding genetic associations than previous approaches. Focusing on the top one percent of impactful variants, he identified nineteen genetic regions linked to body mass index (BMI), pointing to targets for further study. The source notes that AlphaGenome Atlas is available today through an intuitive website portal that requires no coding skills, aiming to democratize access for clinical researchers and biologists worldwide. This release is described as part of DeepMind’s ongoing commitment to accelerate genomic discovery and scientific progress for everyone.
