CellForge: Agentic Design of Virtual Cell Models
Xiangru Tang, Zhuoyun Yu, Jiapeng Chen, Yan Cui, Daniel Shao, Weixu Wang, Fang Wu, Yuchen Zhuang, Wenqi Shi, Zhi Huang, Arman Cohan, Xihong Lin, Fabian Theis, Smita Krishnaswamy, Mark Gerstein
2025-08-05
Summary
This paper talks about CellForge, a smart system that uses multiple AI agents to turn raw biological data from single cells into detailed and optimized virtual cell models that can predict how cells respond to changes like drugs or gene edits.
What's the problem?
The problem is that building computer models of cells is really hard because biological systems are very complex, the data comes in many different forms, and it usually needs experts from many fields to create accurate models.
What's the solution?
CellForge solves this by using a multi-agent framework where different specialized AI agents work together. Some analyze the data and research, others design the best modeling strategies, and another generates the code to create and train the virtual cell model. These agents discuss and improve their plans until they reach the best solution.
Why it matters?
This matters because CellForge helps scientists better understand cells and predict how they react to treatments or changes, speeding up research in biology and medicine while reducing the need for costly experiments.
Abstract
CellForge, an agentic system using a multi-agent framework, transforms raw single-cell multi-omics data into optimized computational models for virtual cells, outperforming state-of-the-art methods in single-cell perturbation prediction.