AlphaGenome 2 Pins DNA Variants to Specific Cells
The update predicts variant effects at cell-type resolution, a model available now to researchers via API, with weights unreleased.
Updated on Nov. 11, 2026 in Biotech

Google DeepMind released AlphaGenome 2 on November 11, 2026, an update to its DNA-sequence model that predicts how genome variants change gene activity in specific cell types rather than whole tissues. The model is available to non-commercial researchers through the AlphaGenome API, while commercial users receive a private preview; the model weights have not been released. This is a released model, accessible through the API and private preview.
Why it matters
The original AlphaGenome worked mostly at the level of tissues and bulk samples, which washes out signals from individual cell types. Genome-wide association studies have found thousands of disease-linked variants without identifying the responsible cell, and AlphaGenome 2 is designed to narrow that gap.
AlphaGenome 2 reads sequences of one million DNA letters at a time and was trained on more than 100 million individual cell profiles drawn from 68 tissues in humans and mice. On TenK10K, it identified causal variants 30% of the time versus 16% for the first model, and identified the direction of effect correctly 43% of the time versus 23%.
The players
Google DeepMind
The Google-owned AI research lab that develops AlphaGenome, a DNA-sequence model built on deep learning and trained on large-scale genomic and single-cell data.
Genomics Limited
A company integrating AlphaGenome 2 into its Mystra platform.
The details
AlphaGenome 2 is trained on single-cell profiles, meaning measurements taken from individual cells rather than averaged across a tissue sample. It predicts how a change in a DNA letter, called a genome variant, shifts how strongly a gene is switched on in a particular cell type. The model adapts directly to new single-cell data without retraining, according to the source.
Timeline
2025: Original AlphaGenome released.
November 11, 2026: AlphaGenome 2 released.
November 30, 2026: Community spotlight program for API credits closes.
The Tech Race
AlphaGenome 2 extends the original AlphaGenome released in 2025, shifting prediction from tissues and bulk samples to individual cell types. The update targets the problem of genome-wide association studies, which have identified thousands of disease-linked variants without identifying the responsible cell.
Non-commercial researchers can access AlphaGenome 2 now through the AlphaGenome API, and commercial users can request a private preview. The model weights have not been released, so the model cannot yet be run independently outside the API.
The takeaway
The shift from tissue-level to cell-type prediction is the trajectory to watch, since it addresses the gap between thousands of disease-linked variants and their responsible cells. Track whether the November 30, 2026 close of the community spotlight program is followed by broader access or a public weight release.
Further reading
For more on the research tools reshaping genetics, see the Biotech section.






