DEV Community

News
News

Posted on

DeepMind unveils AlphaGenome 2, a DNA model that predicts variant effects in specific cell types

Infographic cover

๐Ÿ“… November 11, 2026 โ€ข Google DeepMind โ€” Google DeepMind announced AlphaGenome 2, a next-generation DNA sequence model that predicts how DNA sequences and their variations affect specific human cell types and states. The headline change is architectural: instead of predicting all bulk-tissue outputs in parallel like the original, version 2 is a conditional model that predicts effects for one condition โ€” a tissue, a cell type or a state โ€” at a time, with the condition fed in at every stage, per DeepMind.

Training data grew to more than 100 million individual cell profiles across 68 tissues, figures The Verge repeats. On TenK10K, a single-cell map of 28 primary immune cell types, AlphaGenome 2 reliably identified nearly twice as many causal variants (30 percent versus 16) and directional gene effects (43 percent versus 23) compared with the original AlphaGenome fine-tuned on single-cell data and other specialized tools, according to DeepMind. Against the previous state-of-the-art model it identifies nearly twice as many variants that change gene expression in cells.

Several capabilities matter for lab workflows. Given one measurement such as single-cell gene activity, the model imputes unmeasured properties like chromatin accessibility, helping researchers prioritize experiments. It generalizes to unseen single-cell experiments without retraining โ€” a snapshot from a few dozen cells is enough โ€” and matched or outperformed models fine-tuned directly on those datasets. On COVID-19 mRNA vaccine study data it distinguished vaccine-activated immune T cells from resting ones, per DeepMind.

It also computationally reproduced a published IBD finding, singling out the variant that overactivates the ETS2 gene in gut macrophages while leaving other cells largely unaffected. DNA methylation joins gene expression, histone modifications and chromatin accessibility as a new readout. Availability, per DeepMind and The Verge: free for non-commercial research via the AlphaGenome API starting today, with commercial research in a limited private preview.

The Verge adds balance: genomics lead Ziga Avsec says the finer predictions are more computationally expensive than the predecessor, which is not being replaced โ€” it remains the tool for genome-wide scans like the AlphaGenome Atlas covering all 9 billion possible single-letter DNA changes. For teams choosing between them, the split is one of scope: version 2 for cell-type-specific questions, the original for broad genomic sweeps.

Sources:
deepmind.google/blog/alphagenome-2-deciphering-how-cells-read-dna-to-better-understand-disease
www.theverge.com/ai-artificial-intelligence/1031524/google-deepmind-alphagenome-2-genetic-ai-model

Key figures

IMPACT checklist


Continue the conversation: https://t.me/AIAutomationArchitecture ยท Bluesky

AI Automation

Top comments (0)