Harnessing Agent-Based Modeling in CellAgentChat to Unravel Cell-Cell Interactions from Single-Cell Data

Abstract

Deciphering cell-cell interactions is crucial yet challenging, given the inherent intricacy and heterogeneity of cellular dynamics within biological systems. Existing methodologies, primarily those grounded in statistical or machine-learning approaches, often demand intricate mathematical representations that limit their efficacy in biological settings. Herein, we introduce CellAgentChat, a pioneering application of agent-based modeling (ABM) devised for unraveling cell-cell interactions from single-cell RNA sequencing and spatial transcriptomics data. Our approach encapsulates biological systems as collections of autonomous agents, subject to clear, decipherable rules, and adeptly sidesteps conventional restrictions. Tested against three diverse single-cell datasets, CellAgentChat proved its robustness in delineating intricate signaling events spanning varied cell populations. A unique attribute of our model lies in its ability to yield animated visualizations of single-cell interactions and the flexibility to modify agent behavior rules, thereby enabling comprehensive exploration of both proximal and remote cellular communications. Furthermore, by leveraging the inherent ABM characteristics of CellAgentChat, we can perform intuitive in silico perturbations of cellular interactions through agent rule adjustments, thus creating opportunities for novel intervention strategies. This cutting-edge ABM approach constitutes a significant stride towards a holistic understanding of cellular signaling interactions across a broad spectrum of biological contexts.