In reporting published on November 11, 2026, Google DeepMind unveiled AlphaGenome 2, an AI model designed to predict how genetic variants affect individual cell types and states. The goal is to help researchers investigate why a DNA change may matter in one kind of cell but not another—a question relevant to understanding disease, though the model does not itself establish causes or produce treatments.
Nearly every cell in the body carries the same DNA, but different cells switch different genes on and off. To model those differences, AlphaGenome 2 was trained on more than 100 million individual cell profiles across 68 tissues. Google says it can identify nearly twice as many variants that change gene expression in cells as a previous state-of-the-art model. DeepMind researchers also reported that it can make predictions for cell types and biological states it was not retrained on.
Adam Gayoso, a DeepMind researcher on the project, told The Verge that the advance lies in connecting human genetics with cell biology. DeepMind genomics lead Žiga Avsec said cell-level detail matters when investigating disease and developing drugs that target particular proteins and cell types. These are research aims, not evidence that AlphaGenome 2 has yielded an effective treatment.
Avsec said the more detailed predictions are computationally more expensive than those of the original AlphaGenome, so the newer model is not replacing it outright. DeepMind is also working to improve coverage of cell states that are poorly represented in current data. As reported on November 11, Google said AlphaGenome 2 was available through its API for non-commercial research, with a limited private preview for commercial research.
