Google DeepMind releases AlphaGenome 2 to map genetic variants in specific cell types

Researchers can now link inherited variants to disease-relevant cells more finely, though the model is costlier to run than its predecessor.

Google DeepMind has released AlphaGenome 2, a successor to its genome-predicting model AlphaGenome from last year, which estimates how genetic variants alter gene activity in specific cell types and states rather than in broad tissues. It was trained on more than 100 million individual cell profiles spanning 68 tissues, and the company says it identifies nearly twice as many variants that change gene expression as its previous state-of-the-art model. The model can make predictions about new cell types and biological states without retraining, and can flag gaps in experimental data to guide lab work. For developers of genomics and drug discovery tools, the appeal is a finer-grained link between inherited variants and disease-relevant cells, which matters when targeting particular proteins or cell types. The model is more computationally expensive than its predecessor, so AlphaGenome remains the tool for genome-wide scans, and AlphaGenome 2 is available for non-commercial research through the AlphaGenome API, with a limited private preview for commercial use.