November 11, 2026 — Google DeepMind announced AlphaGenome 2 today, a next-generation DNA model that predicts how genetic variants affect specific human cell types and cell states — not just broad tissues. Why it matters: many disease-linked variants only act inside one specialized cell type, and tissue-level averages can wash that signal out entirely.
What happened
Every cell in your body carries the same DNA, but a heart muscle cell and an immune cell switch very different genes on and off. Last year DeepMind launched AlphaGenome to predict how DNA variants affect molecular activity across broad tissues, then expanded it with AlphaGenome Atlas, which holds predictions for all 9 billion possible single-letter changes in the human genome. AlphaGenome 2 is a redesigned successor: instead of predicting for many tissues in parallel, it is a conditional model that takes one biological condition at a time — a tissue, a cell type, or a cell state — alongside a DNA sequence up to one million letters long, and feeds that condition through every stage of the network.
The details
DeepMind says AlphaGenome 2 was trained on bulk and single-cell data from human and mouse samples, including more than 107 million individual cell profiles across 68 tissues. It predicts the same core readouts as the original model — gene expression, histone modifications, chromatin accessibility — and adds DNA methylation. On the TenK10K dataset, a single-cell map of 28 primary immune cell types, DeepMind reports the model identified nearly twice as many putatively causal variants as the original AlphaGenome fine-tuned on single-cell data (30% vs. 16%) and more directional gene effects (43% vs. 23%), at the precision reported in its paper. As a check, the model reproduced a published finding in which a variant in a distant regulatory switch overactivates the ETS2 gene in gut macrophages and drives inflammatory bowel disease — singling out the causal variant and the specific cell type.
The model can also work with cell states it has never seen. A researcher can upload a single-cell RNA snapshot from a few dozen cells, and AlphaGenome 2 compresses it into a cell-state representation it can predict from — no retraining. DeepMind says it used this on data from a COVID-19 mRNA vaccine study, distinguishing vaccine-activated T cells from resting ones and showing that a variant linked to severe COVID-19 disrupts gene activity more strongly in the activated state. DeepMind has also precomputed AlphaGenome 2 predictions for trait-associated variants across hundreds of human cell and tissue conditions, including enhancer–gene maps in primary human cells such as brain astrocytes.
Availability — and the limits
AlphaGenome 2 and interactive Colab tutorials for expression conditioning are available today for non-commercial research via the AlphaGenome API, and in private preview for commercial research. DeepMind says AlphaGenome 2 Skills for Google Antigravity, its agentic development platform, are coming soon. Two cautions come straight from DeepMind: the model is for research use only and has not been validated or approved for clinical or diagnostic purposes, and even at high precision it recovers only a minority of the variants statistical analyses nominate as likely causal. It is a prioritization tool for researchers, not a diagnostic.
Why it matters
Genome-wide association studies keep flagging stretches of DNA linked to disease without naming the exact variant, gene, or cell responsible. A model that can point at all three — variant, gene, and the cell type where it acts — turns a long candidate list into a short one that labs can actually test. That is the same playbook that made AlphaFold valuable, applied one level down: not what a protein looks like, but where in the body a genetic typo actually does its damage. For drug developers, whose therapies target proteins in particular cell types, that cell-level map is the useful part.
Frequently asked questions
What is AlphaGenome 2?
It is Google DeepMind’s next-generation DNA sequence model, announced November 11, 2026. It predicts how DNA sequences and their variants affect gene regulation in specific human cell types and cell states.
How is AlphaGenome 2 different from the first AlphaGenome?
The original predicted molecular activity across broad tissues in parallel. AlphaGenome 2 is a conditional model: it takes one tissue, cell type, or cell state as an input and predicts for that condition, which DeepMind says roughly doubles the causal variants it recovers on single-cell benchmarks.
Who can use AlphaGenome 2?
It is available today for non-commercial research through the AlphaGenome API; commercial research access is in private preview. It is research-only and not approved for clinical or diagnostic use.
What data was AlphaGenome 2 trained on?
Human and mouse bulk and single-cell data, including more than 107 million individual cell profiles across 68 tissues, according to DeepMind.
Sources: Google DeepMind (announcement and technical materials), Unite.AI, The Verge via Zubiqo.