Google DeepMind's AlphaGenome 2 predicts how DNA variants act in specific cell types

On November 11, 2026, Google DeepMind introduced AlphaGenome 2, a model that predicts how a DNA sequence and its variants affect gene activity in a specific human cell type. It is a conditional model: it takes a 1-million-letter DNA sequence and makes predictions for one biological condition at a time, such as a tissue, cell type or cell state. Its readouts cover gene expression, histone modifications, chromatin accessibility and DNA methylation.

DeepMind says the model identifies nearly twice as many variants that change gene expression in cells as the previous state of the art. On the TenK10K benchmark, covering 28 primary immune cell types, it identified 30% of causal variants against 16% for the original AlphaGenome fine-tuned on single-cell data and other specialized tools, and got 43% of directional gene effects against 23%. The post describes training on more than 100 million individual cell profiles across 68 tissues (the body text says over 107 million).

The model is available for non-commercial research through the AlphaGenome API, with commercial research access in private preview on Google Cloud, plus precomputed predictions for trait-associated variants across hundreds of cell and tissue conditions.

Why it matters: Most disease-linked genetic variants sit outside genes and act only in particular cells. A model that resolves effects at the level of individual cell types moves variant interpretation closer to the biology that actually matters for disease.

What it does not show: The figures are DeepMind’s own, from its blog and technical report. The model is for research use only and has not been validated or approved for clinical or diagnostic purposes, and doubling a benchmark recall rate still leaves most causal variants unidentified.