Alphagenome 2 advances genetic modeling by linking dna variation to specific cell types

Researchers have unveiled AlphaGenome 2, a new DNA sequence model that predicts how genetic variations influence gene expression across hundreds of human cell types. This next-generation tool integrates single-cell biology with human genetics. It was trained on over 100 million individual cell profiles across 68 tissues, identifying nearly double the number of variants that alter gene expression compared to the prior model. Unlike its predecessor, AlphaGenome 2 is a conditional model, accepting a specific biological condition (like a cell type) as input. It also supports cross-modality imputation, allowing predictions of unmeasured properties such as chromatin accessibility from single-cell gene activity. The model can adapt to new single-cell data without full retraining, demonstrated by its ability to distinguish vaccine-activated immune T cells from resting ones in COVID-19 data. The platform is available today for non-commercial research via the AlphaGenome API, with a private preview for commercial use. The team also precomputed predictions for trait-associated variants across numerous conditions to aid in bridging the gap between genome-wide association studies and drug discovery.