Google DeepMind has launched AlphaGenome Atlas, which predicts all 9 billion single nucleotide variations in the human genome.

On September 8, 2026, Google DeepMind released the AI platform called AlphaGenome Atlas, aimed at unraveling the mysteries of the human genome and facilitating the development of new therapies. The platform contains approximately 1 PB of data, covering predictions of all 9 billion single nucleotide variations in the human genome, which is more than 30 times larger than the size of the AlphaFold database. The platform provides integrated resources such as molecular effect predictions, AVI scores, and DNA sequence motifs, supports analysis of variations in both coding and non-coding regions, and consolidates the results of AlphaGenome and AlphaMissense models into a single value to assist in ranking. To date, external collaborative teams have used this platform to identify DNM1 gene variations in rare disease research and additionally discovered 22% of non-coding gene associations. AlphaGenome Atlas is now available for academic research through a portal and API; non-commercial research can access it for free, and commercial access will be offered later on Google Cloud.