Key Notes
- AlphaGenome 2 predicts the effects of DNA changes in specific cell types and states.
- Researchers can condition the model on their own single-cell data without fine-tuning it.
- Access is available for non-commercial research, while clinical decision-making is prohibited.
Google DeepMind has released AlphaGenome 2, an AI model that predicts how DNA changes affect particular cell types, giving researchers a more focused way to investigate the biology behind disease.
Google said on November 11 that the model was trained on more than 107 million individual cell profiles across 68 tissues. On the TenK10K immune-cell dataset, it identified 30% of causal variants, versus 16% for the earlier AlphaGenome fine-tuned on single-cell data. Those figures describe a research evaluation, not diagnostic accuracy in patients.
A lab’s own cells become part of the input
The practical change is that a researcher can ask about a particular biological condition rather than rely only on a tissue-wide prediction. A DNA variant may have a different effect in one cell state than another, so the choice of cells becomes part of the question posed to the model.
Google’s tutorial shows how researchers can supply their own single-cell gene-expression measurements and obtain predictions without fine-tuning the model on that dataset. The example compares a custom sample of immune cells with a built-in reference condition, then examines predicted changes in gene activity and DNA accessibility.
The input still needs preparation. The tutorial calls for raw counts with compatible gene identifiers, accepts up to 256 cells per call and recommends at least 16 when available to reduce sampling noise. That makes this a tool for a research workflow, rather than a chatbot that interprets an uploaded genetic test.
Research access comes with clear boundaries
The AlphaGenome documentation offers free API access for non-commercial use, with query rates dependent on demand. Google says AlphaGenome 2 is also entering a private preview for commercial research.
The documentation warns that predictions may not capture cell states far outside the training data. It also distinguishes benchmarked outputs from exploratory predictions, which should be treated as hypotheses. The service’s terms prohibit clinical decision-making; Google says the models have not been validated or approved for diagnostic use.
The release builds on work AIstify reported with AlphaGenome Atlas, a database of predicted effects for nine billion possible single-letter DNA changes. AlphaGenome 2 adds a more specific cellular context to the investigation. Its value will depend on how well those predictions guide experiments that researchers can verify.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
