Google DeepMind unveils AlphaGenome 2 to predict DNA variant effects in specific cell types
The model improves variant effect prediction by nearly twofold and is available today for non-commercial research via API and in private preview for commercial use.
1 source · cross-referenced
- AlphaGenome 2 predicts how DNA sequences and their variations impact specific human cell types and disease states.
- The model was trained on more than 100 million individual cell profiles across 68 tissues.
- It identifies nearly twice as many variants that change gene expression in cells compared to its predecessor.
- Available today for non-commercial research via the AlphaGenome API and in private preview for commercial research.
Google DeepMind announced AlphaGenome 2, a next-generation model designed to predict how DNA sequences and their variations impact specific human cell types. The model bridges single-cell biology and human genetics, offering scientists a tool to uncover cellular drivers of disease.
Cells perform specialized functions by activating or suppressing different genes, even though they share the same DNA. Projects like the Human Cell Atlas map gene expression across millions of cells, but predicting how subtle DNA variations affect this expression in specific cell types has remained challenging. AlphaGenome 2 addresses this gap by focusing on cell-type-specific predictions rather than broad tissue averages.
The model represents a step-change improvement over its predecessor, AlphaGenome, by redesigning its architecture to handle one biological condition at a time. This conditional approach allows AlphaGenome 2 to learn how the same DNA sequence is interpreted differently across cell types, improving its ability to capture subtle differences between closely related cell types.
AlphaGenome 2 was trained on more than 100 million individual cell profiles spanning 68 tissues. This training data enables the model to turn static cell atlases into a predictive engine for human genetics, identifying nearly twice as many variants that change gene expression in cells compared to the previous state-of-the-art model.
The model introduces cross-modality imputation, allowing researchers to predict unmeasured properties from limited inputs. For example, if a scientist provides a single measurement like single-cell gene activity, AlphaGenome 2 can predict unmeasured properties such as chromatin accessibility, helping prioritize experimental next steps.
AlphaGenome 2 is available today for non-commercial research via the AlphaGenome API and is in private preview for commercial research. The model builds on earlier tools like AlphaGenome and AlphaGenome Atlas, which provided predictions for all 9 billion possible single-letter changes in the human genome.
- Nov 11, 2026 · The Verge — AI
Google DeepMind unveils AlphaGenome 2 to predict genetic variant effects on cell types
Trust79 - Nov 11, 2026 · arXiv cs.AI
Researchers demonstrate auditable human-agent teams for autonomous synchrotron X-ray experiments
Trust79 - Nov 11, 2026 · arXiv cs.AI
Sparse autoencoders enable compact, prefill-trained topic monitors that generalize to decoding
Trust79