Google DeepMind announced AlphaGenome 2, a next-generation DNA sequence model that predicts how genetic variants affect specific human cell types. The launch took place on November 11, 2026.
While the original AlphaGenome (2025) predicted effects across broad tissues, the new version works as a conditional model: it takes a roughly 1-million-letter DNA sequence plus a biological condition (cell type or state) and generates predictions for that specific condition.
Key advances
- Identifies nearly twice as many causal variants that change gene expression (30% vs. 16% on the TenK10K benchmark of 28 immune cell types).
- Predicts the direction of the effect (activation or repression) in 43% of cases, compared with 23% for the previous model.
- Trained on more than 100 million individual cell profiles from 68 human and mouse tissues.
- Adds DNA methylation prediction and supports imputation of unmeasured modalities.
- Generalizes to new cell types from a researcher’s own single-cell RNA-seq data without retraining.
According to DeepMind, the model turns static cell atlases into a predictive engine for human genetics. It computationally reproduced a recent finding about a variant that activates the ETS2 gene in gut macrophages and contributes to inflammatory bowel disease.
AlphaGenome 2 is available today via the AlphaGenome API for non-commercial research. Commercial use remains in private preview. Colab tutorials and the Python package have also been updated.
Sources: Google DeepMind, The Verge.
This article was produced with the assistance of artificial intelligence and editorially reviewed.
Por Grok
