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 J Theis, Smita Krishnaswamy, Mark Gerstein

Keywords: LLMs

Abstract: Virtual cell modeling represents an emerging frontier at the intersection of artificial intelligence and biology, aiming to predict quantities such as responses to diverse perturbations quantitatively. However, autonomously building computational models for virtual cells is challenging due to the complexity of biological systems, the heterogeneity of data modalities, and the need for domain-specific expertise across multiple disciplines. Here, we introduce CELLFORGE, an agentic system that leverages a multi-agent framework that transforms presented biological datasets and research objectives directly into optimized computational models for virtual cells. More specifically, given only raw single-cell multi-omics data and task descriptions as input (e.g., for control and perturbed conditions and a directive to build a model of a new perturbation), CELLFORGE outputs both an optimized model architecture and executable code for training virtual cell models and inference. The framework integrates three core modules: Task Analysis for presented dataset characterization and relevant literature retrieval, Method Design, where specialized agents collaboratively develop optimized modeling strategies, and Experiment Execution for automated generation of code. The agents in the Design module are separated into experts with differing perspectives and a central moderator, and have to collaboratively exchange solutions until they achieve a reasonable consensus. We demonstrate CELLFORGE’s capabilities in single-cell perturbation prediction, using six diverse datasets that encompass gene knockouts, drug treatments, and cytokine stimulations across multiple modalities. CELLFORGE consistently outperforms task-specific state-of-the-art methods, achieving up to 40% reduction in prediction error and 20% improvement in correlation metrics. Overall, CELLFORGE demonstrates how iterative interaction between LLM agents with differing perspectives provides better solutions than directly addressing a modeling challenge

Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)

Submission Number: 9107