Google DeepMind unveils AlphaGenome 2 for cellular genetic analysis

On November 11, 2026, Google DeepMind introduced AlphaGenome 2, a sophisticated DNA prediction model that identifies how genetic changes behave in specific biological settings. While many people often wonder how Google reveals new technological breakthroughs, this latest development focuses on mapping genetic activity across distinct cell types and states rather than general search results. By moving beyond broad tissue measurements, the model helps scientists pinpoint disease-linked effects that might otherwise remain obscured.

How Google reveals genetic insights

The original AlphaGenome model provided fixed outputs based on long DNA sequences. In contrast, AlphaGenome 2 integrates specific target conditions—such as a particular cell type or state—directly into its architecture alongside the DNA sequence. This allows researchers to analyze how the same stretch of genetic material functions across different cellular environments. The system was trained on a massive dataset of over 107 million individual cell profiles from both human and mouse samples, enabling it to learn how DNA activity shifts according to biological context.

AlphaGenome 2 predicts genetic behavior in specific cells to help researchers prioritize targeted laboratory experiments.

One practical application of this technology involves identifying causal variants. In tests using the TenK10K dataset, which covers 28 primary immune-cell types, the new model identified 30% of causal variants, significantly outperforming the original version. DeepMind also demonstrated the model’s utility by computationally reproducing an established biological mechanism involving the ETS2 gene in gut macrophages, a finding previously linked to inflammatory bowel disease in 2024 research. By isolating the variant and its specific impact on the ETS2 gene, the model provides a clearer roadmap for laboratory investigation.

Researchers can also supply their own single-cell gene-activity data to the model to condition predictions on newly observed states without requiring additional training. While the model offers a powerful tool for narrowing down candidate variants, DeepMind emphasizes that these results measure variant identification rather than diagnostic accuracy or individual disease risk. As of the announcement, the system is available via API for non-commercial research, with a private preview for commercial entities. The company maintains that the technology has not been approved for clinical or diagnostic use.

Why it matters

AlphaGenome 2 changes the research landscape by allowing scientists to focus their experiments on the precise cells where genetic effects are predicted to occur. By connecting disease-associated signals to specific genes and cell types, it reduces the guesswork in biological studies.

This development is significant for the global scientific community, including researchers in India and South Asia who are working on complex genetic diseases. By providing a more efficient way to prioritize laboratory work, the model enables faster investigation into the underlying biology of chronic conditions.