Google DeepMind launches AlphaGenome Atlas to map 9 billion DNA mutations

Google DeepMind has introduced AlphaGenome Atlas, a petabyte-scale platform designed to map the human genome by predicting the molecular effects of 9 billion single nucleotide variants. This free resource for academic research covers both the 2% protein-coding region and the 98% non-coding genome, providing a detailed catalog of genetic changes.

The platform utilizes the AlphaGenome Variant Impact (AVI) index, which combines data from AlphaGenome and AlphaMissense models to provide a single score for assessing the potential biological harm of mutations. This allows researchers to quickly sort variants and identify specific molecular disruptions, such as RNA maturation or chromatin accessibility.

Beyond mapping existing DNA, the AlphaGenome model can interpret long strands of DNA—up to one million base pairs—and predict how subtle mutations affect biological processes. Researchers suggest this technology could enable the design of synthetic DNA to act as “synthetic switches,” allowing gene therapies to activate or deactivate genes only in specific target tissues, such as the liver or retina, potentially ushering in a new era of personalized medicine.