Google DeepMind releases AlphaGenome 2 for cell-specific DNA variant effects

Google DeepMind released AlphaGenome 2 on November 11, 2026, a DNA sequence model that predicts how genetic variants affect specific human cell types and states, trained on more than 107 million individual cell profiles. Full details are in the AlphaGenome 2 blog post.

What changed in AlphaGenome 2

• Trained on over 107 million individual cell profiles across 68 tissues, covering bulk and single-cell data from human and mouse.
• Predicts effects of DNA changes in specific human cell types and states, unlike the cell-agnostic predictions of AlphaGenome 1.
• Identifies nearly twice as many expression-altering variants as AlphaGenome 1.
• On the TenK10K single-cell immune dataset across 28 cell types, recovers 30% of causal expression-linked variants at 90% precision, versus 16% for AlphaGenome 1 fine-tuned on single-cell data, and recovers 43% of directional gene effects versus 23% for the prior model.
• Redesigned as a conditional model that predicts effects for one specified biological condition at a time, adding new DNA methylation predictions and cross-modality imputation.
• Covers all approximately 9 billion possible single-letter DNA changes in the human genome.

What it means for people who build with AI

AlphaGenome 2 is a major capability jump for a model that developers and researchers can already access via API. It shows DeepMind extending frontier model techniques into cell-specific biological prediction, nearly doubling detection accuracy compared with its predecessor. For people building computational biology tools, the conditional architecture and the new methylation and cross-modality predictions open up applications that were not possible with a cell-agnostic model.

Availability and what's next

AlphaGenome 2 is available now for non-commercial research through the AlphaGenome API, with a private preview open to commercial and drug-discovery partners. Google DeepMind is also planning AlphaGenome 2 Skills for Google Antigravity in a future release. The model remains research-only and is not approved for diagnostic use.

Sources

• AlphaGenome 2: Deciphering how cells read DNA to better understand disease (Google DeepMind, November 11, 2026, official)

Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.