Google DeepMind releases AlphaGenome 2, with predictions refined to the single-cell type
Google DeepMind has launched AlphaGenome 2, a new-generation genomic AI model that can predict the effects of DNA sequences and their variants on specific human cell types and states. The model was trained on over 107 million single-cell atlases from 68 tissues, covering human and mouse samples, and the number of gene expression-related variants it identifies is nearly twice that of the previous-generation model. DeepMind has opened the model to non-commercial research through the AlphaGenome API, while commercial research is in a limited preview phase.
Why read this
AI score 70
Compared with the previous generation, which only analyzed whole-tissue data, the new model refines predictions to the single-cell type and provides details on its training data and open-access approach.
Topics
Body · AI translation
Google DeepMind Releases AlphaGenome 2: Tracing the Cellular Roots of Disease
Phoenix Tech Download the App for Exclusive Early Access
November 12, 2026, 07:13:50 From Beijing
Updated November 12, 2026, 07:13:01
Phoenix Tech News, November 12 — Google DeepMind recently announced the launch of its next-generation genomic AI model, AlphaGenome 2, which can predict the effects of DNA sequences and their variants on specific human cell types and states, helping researchers find disease-related genetic clues at the cellular level.
Most cells in the human body possess the same DNA, but different cells turn different genes on or off depending on their type and state. As a result, the same genetic variant may have different effects in heart muscle cells, skin cells, or immune cells. Compared with the previous-generation model, which mainly analyzed whole-tissue data, AlphaGenome 2 can refine predictions further down to individual cell types.
DeepMind said the model was trained on more than 107 million single-cell atlases from 68 tissues, covering human and mouse samples. Test results show that the number of gene-expression-related variants it identifies is nearly double that of the previous-generation model, and it can predict indicators such as gene expression, chromatin accessibility, and DNA methylation.
AlphaGenome 2 can also analyze previously unseen cell types and biological states such as disease and vaccination without retraining, and infer missing information in experimental data to help researchers determine the direction of follow-up experiments. DeepMind has opened the model to non-commercial research through the AlphaGenome API, while commercial research is currently in a small-scale preview phase. (Author/Song Shumei)
For more first-hand news, please download the Phoenix News app and subscribe to Phoenix Tech. For in-depth reporting, search "Phoenix Tech" on WeChat.