Google DeepMind has released AlphaGenome 2, a model designed to predict how DNA variants affect particular human cell types and states. The narrower view matters because a genetic change can alter activity in one disease-relevant cell while remaining hidden in an average across an entire tissue.
The model takes a DNA sequence of up to one million letters plus a selected biological condition, such as a cell type, and predicts molecular effects for that condition. DeepMind says it trained the system on more than 100 million individual cell profiles across 68 tissues. In its tests, AlphaGenome 2 identified nearly twice as many variants that change gene expression as the company’s previous model.
Researchers can also provide data from their own single-cell experiments, allowing the model to estimate unmeasured properties such as chromatin accessibility. Those predictions can help prioritize laboratory work, but they do not replace experiments or clinical validation. AlphaGenome 2 is available now through an API for non-commercial research, while commercial research access is in private preview.