Google DeepMind has introduced AlphaGenome 2, a next-generation DNA sequence model designed to predict how genetic variants influence gene activity in specific human cell types. The announcement, published on November 11, 2026, highlights a shift from broad tissue-level predictions toward a more detailed view of the biological context in which a DNA variant acts.
What AlphaGenome 2 is designed to do
Every cell in the human body carries essentially the same DNA, but different cell types use that genetic information in different ways. A variant that changes gene activity in an immune cell may have a different effect, or no meaningful effect, in a heart cell. AlphaGenome 2 is designed to model those differences by conditioning its predictions on a target tissue, cell type, or cell state alongside the DNA sequence.
DeepMind says the model was trained on more than 100 million individual cell profiles spanning 68 tissues. The company reports that on the TenK10K single-cell dataset, AlphaGenome 2 identified nearly twice as many causal variants as its original model fine-tuned for single-cell data: 30% compared with 16%. These are company-reported benchmark results, not a clinical validation or a guarantee that the model will identify causal variants in every research setting.
Why cell-specific predictions matter
Many genetic studies identify regions of DNA associated with a disease or trait but do not immediately reveal which variant is responsible, which gene it affects, or the cell type where the effect occurs. A model that can narrow those possibilities may help researchers prioritize experiments, connect genetic signals to cell biology, and investigate mechanisms that broad tissue averages can miss.
DeepMind also describes a capability to work with new single-cell experiments without retraining the model. Researchers can provide a small sample of gene-activity measurements to represent a cell state and ask the model to estimate other molecular readouts. This could make computational analysis more adaptable as scientists study specialized or disease-associated cell populations.
Important limitations and availability
AlphaGenome 2 is available through the AlphaGenome API for non-commercial research, while commercial research access is in private preview. DeepMind explicitly says the system is intended for research use and has not been validated or approved for clinical or diagnostic purposes. It should therefore be treated as a research aid, not a medical test, diagnosis, or substitute for laboratory evidence.
The next meaningful test will come from independent research groups using the model on their own datasets and reporting where its predictions are useful, uncertain, or wrong. For genetics and drug-discovery teams, the announcement is notable because it points toward more context-aware AI analysis of the genome rather than simply larger general-purpose models.
Source
Image credit: Julia Koblitz / Unsplash. Used as an illustrative laboratory image.
