Google DeepMind shifts genomics AI to single cell type resolution with AlphaGenome 2
Early report
Google DeepMind has released AlphaGenome 2, an AI model that predicts genetic variant effects in specific human cell types, shifting from bulk tissue predictions to single-cell resolution.
AlphaGenome 2 is a conditional sequence-to-function model that reads 1 Mb of DNA plus a target cell state, and its predecessor averaged across bulk tissues.
The model resolves closely related cell types and adapts to unseen single-cell datasets without fine-tuning, and is available through the AlphaGenome API for non-commercial research.
In testing, AlphaGenome 2 found nearly twice as many relevant gene variants as its predecessor, according to Google DeepMind.