🔬Causal Models Need Causal Data - Xaira’s X-Cell model for Drug Discovery (Bo Wang & Ci Chu, Chief Discovery Officer & Chief AI Scientist)

Xaira Therapeutics is advancing artificial intelligence for drug discovery through a strategic focus on information rich data, led by Chief Discovery Officer Ci Chu and Chief AI Scientist Bo Wang. Traditional virtual cell models, trained on public databases like CELLxGENE, map RNA expression and cell states but struggle to predict how cells will react to specific changes. Because gene expressions are merely correlated, these models cannot determine causal relationships, which limits their performance when scaling model parameters and compute.

To overcome this data information gap, Xaira developed a proprietary dataset called X-Atlas and a corresponding model named X-Cell. X-Atlas was created using parallel CRISPR-based experiments to systematically alter individual genes and observe the resulting upstream and downstream effects across millions of tests. This causal dataset is roughly thirty times more informative than standard observational data, allowing the X-Cell model to successfully scale with parameters and training compute rather than hitting performance walls.

This development matters because it shifts AI-driven biology from merely observing correlations to predicting actual causal outcomes in human cells, such as the effects of drugs or gene edits. By investing heavily in large-scale causal data generation and infrastructure, X-Cell bypasses the limitations of previous RNA expression models, offering a more powerful foundation for virtual cell modeling and therapeutic discovery.