StoryResearch
Arc Institute releases State, a virtual cell model trained on 270 million cells
Arc Institute released State, a virtual cell model that predicts how cells respond to genetic and chemical perturbations. Trained on data from nearly 170 million observed cells and over 100 million perturbed cells across 70 human cell types, State is the largest single-cell perturbation model released to date.
- Trained on 270M+ cells across 70 human cell types with chemical, genetic and cytokine perturbations.
- Uses two-module architecture: State Embedding (SE) for cross-batch normalization, State Transition (ST) for perturbation modeling.
- Enables drug discovery by predicting cell responses before experiments, reducing development time.
- Released for noncommercial use; commercialization path through licensing arrangements.