AI model finally unlocks how tiny DNA changes cause disease in specific cells

DeepMind has released AlphaGenome 2, a new AI model that predicts how subtle variations in DNA affect gene expression in specific human cell types. This represents a significant leap from their previous version, which analyzed broad tissue averages, to now focusing on individual cell types and states like heart muscle cells or immune cells fighting infections.

The model, trained on over 100 million individual cell profiles across 68 tissues, identifies nearly twice as many disease-relevant genetic variants compared to prior state-of-the-art tools. For researchers, this means better understanding of how specific DNA changes drive diseases in particular cell types that would be missed in broader analyses. The breakthrough transforms static cell atlases into predictive engines for human genetics, offering a powerful new tool for uncovering cellular drivers of rare diseases and complex traits.