Google DeepMind releases AlphaGenome 2 to predict DNA variant effects in specific cell types

Google DeepMind has released AlphaGenome 2, a model that predicts how DNA sequences and their variations affect gene activity in specific human cell types, the lab announced on 11 November. It is available for non-commercial research through the AlphaGenome API and in private preview for commercial research.

The first AlphaGenome predicted the effects of DNA variants averaged across broad tissues. AlphaGenome 2 instead takes one condition at a time, such as a tissue, cell type or cell state, alongside a sequence of up to a million DNA letters, so it can capture how the same DNA is read differently in, for example, heart muscle cells and immune cells. DeepMind says it identifies nearly twice as many variants that change gene expression in cells as its previous model. It was trained on bulk and single-cell data from human and mouse samples, including more than 107 million individual cell profiles across 68 tissues, and adds DNA methylation to its predictions.

Researchers can also give it a small single-cell sample of a new cell state, such as cells activated by a vaccine, and get predictions without retraining. DeepMind has precomputed predictions for trait-associated variants across hundreds of cell and tissue conditions. Genomics lead Žiga Avsec told The Verge that the detailed predictions are "more expensive" to generate, so the new model will sit alongside the original for genome-wide scans, TechRepublic reports. The model is for research only and is not approved for clinical or diagnostic use.