Predicting cellular responses to perturbation across diverse contexts with State.

State predicts cellular responses to perturbations across diverse contexts with over 30% improved accuracy on large datasets.

• Why it matters: Understanding how cells respond to various perturbations is crucial for advancing personalized medicine and biological research, but current models lack generalization across different cellular environments.
• What they did: The team trained State using single-cell gene expression data from 167 million cells to predict effects of genetic, signaling, and chemical perturbations, and developed Cell-Eval for model assessment.
• The result: State significantly enhances the ability to identify differential gene expression and strong perturbations in unseen cellular contexts, enabling scalable AI modeling of cell states.