Almost every cell in the body carries the same DNA, but heart, skin and immune cells switch different genes on and off, and those patterns shift as a cell changes state [2]. Most variants that population studies link to disease sit outside genes, in regulatory stretches of DNA that act like dimmer switches for nearby or distant genes. The same variant can therefore do nothing in one cell type and change a gene’s output sharply in another [2].
The first AlphaGenome, previewed in June 2025 and published in Nature in January 2026, produced fixed parallel outputs for bulk tissues and cell lines [1][3]. DeepMind says that design “spread the model’s capacity thin” [1]. AlphaGenome 2 instead takes a biological condition as an input alongside the DNA sequence and predicts for one condition at a time, using that condition at every stage of the architecture [1]. Its readouts are those of the first model (gene expression, histone modifications and chromatin accessibility) plus DNA methylation [1].
A cell encoder lets scientists build a condition from their own single-cell experiment. DeepMind says that with a few dozen cells of single-cell RNA data and no retraining, the model can match or beat models trained or fine-tuned directly on the new dataset [1]. Given one measurement, such as single-cell gene activity, it can also predict unmeasured properties such as chromatin accessibility [1].