November 11, 2026 Science
Adam Gayoso, Natasha Latysheva, Pushmeet Kohli, and Žiga Avsec

Our next-generation DNA sequence model bridges single-cell biology and human genetics, predicting how genetic variants impact hundreds of human cell types.
Today, we are introducing AlphaGenome 2, our next-generation model that predicts how DNA sequences and their variants affect specific human cell types, giving scientists a powerful new tool to uncover the cellular drivers of disease.
Cells are the fundamental building blocks of life. Every cell in our body carries the same DNA blueprint, yet each cell type and state—such as a heart muscle cell or an immune cell fighting off an infection—performs its role by turning different genes on or off. Global initiatives like the Human Cell Atlas are mapping gene expression across millions of cells, but predicting how subtle DNA variants alter this expression within cells remains an unsolved challenge—one that is a critical driver of human traits and diseases.
Last year, we introduced AlphaGenome to predict how DNA variants affect molecular activity across broad tissues. We then expanded its reach with the AlphaGenome Atlas, which contains predictions for all 9 billion possible single-letter changes in the human genome. These tools have helped researchers prioritize and interpret variant effects, driving discoveries across rare diseases, complex traits, and fundamental biology.
By shifting the focus from broad tissue averages to individual cell types and states, and by redesigning our model architecture, AlphaGenome 2 delivers a step-change in variant prediction performance. Importantly, it enables researchers to predict variant effects in disease-relevant cells that broad tissue models miss, and to make predictions for cell and disease states the model has never seen before.
Compared to our previous state-of-the-art (SOTA) model, AlphaGenome 2 nearly doubles the number of identified variants that alter cellular gene expression. Trained on over 100 million single-cell profiles across 68 tissues, the model transforms static cell atlases into a predictive engine for human genetics. AlphaGenome 2 is available starting today for non-commercial research via the AlphaGenome API, and in private preview for commercial research.
Redesigning AlphaGenome for specific cell types


Comparing how AlphaGenome 1 and 2 process cell types. From a 1-million-base DNA sequence, AlphaGenome generates a fixed number of bulk tissue and cell line outputs in parallel. AlphaGenome 2 is a conditional model that takes a specific biological condition alongside the input DNA sequence—either selected from a catalog of reference cell types or generated on the fly from a scientist’s own single-cell experiments using our cell encoder.
The first-generation AlphaGenome model took a 1-million-base DNA sequence as input and simultaneously predicted properties across thousands of experiments from multiple tissues. While efficient, this parallel approach divided the model’s capacity, making it difficult to capture subtle differences between closely related cell types.
To improve upon this, we fundamentally redesigned how the model handles cell types. Instead of making parallel predictions, AlphaGenome 2 acts as a conditional model, predicting DNA effects for one condition (defined as a tissue, cell type, or state) at a time. Alongside the 1-million-base DNA sequence, it takes the target biological condition as an input at every stage of its architecture. This allows it to learn how the exact same stretch of DNA is read differently across distinct cell types.
Conditional modeling also enables cross-modality imputation, allowing AlphaGenome 2 to predict missing experimental data. Because the model decouples a cell’s biological state from its outputs, if a scientist inputs just a single measurement (such as single-cell gene activity), AlphaGenome 2 can predict unmeasured properties (such as chromatin accessibility). This helps researchers prioritize their next experiments.
We trained AlphaGenome 2 on bulk and single-cell data from human and mouse samples, including over 107 million single-cell profiles. Moving beyond the bulk tissue averages of the original model, AlphaGenome 2 resolves activity within specific cells—such as distinct cardiomyocytes or blood immune cells—helping uncover critical DNA variants that act only in specific cell types.
In addition to predicting the same core DNA functional experimental readouts as AlphaGenome—such as gene expression, histone modifications, and chromatin accessibility—the model adds predictions for a highly requested new readout: DNA methylation.
What is exciting about AlphaGenome 2 is its ability to directly adapt to new s