🧬Google DeepMind Launches AlphaGenome 2, Trained on 100 Million Cell Profiles

Your drug target shortlist just got a lot sharper

Google DeepMind just shipped AlphaGenome 2, and it zooms in where the original model zoomed out. It predicts how genetic variants affect individual cell types, not just broad tissues. That's a big jump for anyone chasing disease mechanisms. If you're in computational biology or building drug discovery pipelines, this is your new variant triage tool. Instead of guessing which cells a population-level variant hits, you can drill down to the cellular level and connect those effects to disease. Specificity matters here because drugs that target particular proteins in particular cell types are the whole game. The numbers: the model was trained on over 100 million individual cell profiles across 68 tissues. It identifies nearly twice as many variants that change gene expression compared to its previous state-of-the-art model. It can also make predictions about new cell types and biological states without retraining. It's available now for non-commercial research via the AlphaGenome API. Commercial teams get a limited private preview. Watch for future versions trained across multiple species, which could feed into the longer-term push toward virtual cells.