Presentation Information
[POS-64]A Reinforcement Learning-based Model of Chemotactic Cell Populations Solving Cooperative Games
*Masaki Kato1, Tetsuya J. Kobayashi1 (1. The University of Tokyo (Japan))
Keywords:
Chemotaxis,Keller-Segel model,Pattern formation,Mean Field Games,Reinforcement Learning
Chemotactic cells typically generate their own chemical gradient, resulting in collective behavior such as aggregation or dispersal that are crucial in biological processes including development and immunity [1]. Although these behaviors are functionally significant, classical mathematical models such as Keller-Segel type models typically rely on phenomenological observation [2], lacking an explicit rationale for how the aggregation strategies contribute to functions. Based on an assumption that cells have evolved to behave as if optimizing certain functions, we formulate cellular aggregation problems as cooperative games. We employ Reinforcement Learning and Mean Field Games (MFG) theories [3] to derive mathematical models for collective chemotaxis, in which each cell cooperatively learns an equilibrated strategy by producing or degrading chemicals. In an appropriate limit, our derived models align with a subclass of Keller-Segel type equations, explicitly characterizing rational chemotactic behavior driven by optimization. Our models approach diverse spatial patterns, each corresponding to equilibria from MFG viewpoint. These patterns are not only examined using conventional bifurcation and stability analysis, but also compared regarding their optimality, owing to our optimization-based formulation. Our framework bridges mathematical biology and dynamical optimization theories, providing new insights into chemical-mediated self-organized phenomena and a theoretical foundation for rationalizing cellular aggregation from a functional perspective.
[1] R. H. Insall, P. Paschke, and L. Tweedy, Steering yourself by the bootstraps: how cells create their own gradients for chemotaxis, Trends Cell Biol. 32, 585 (2022).
[2] K. J. Painter, Mathematical models for chemotaxis and their applications in self-organisation phenomena, J. Theor. Biol. 481, 162 (2019).
[3] M. Lauriere, S. Perrin, M. Geist, and O. Pietquin, Learning mean field games: A survey, arXiv preprint arXiv:2205.12944 (2022).
[4] M. Kato, and T. J. Kobayashi, Optimality theory of stigmergic collective information processing by chemotactic cells, arxiv preprint arxiv:2407.15298 (2025).
[1] R. H. Insall, P. Paschke, and L. Tweedy, Steering yourself by the bootstraps: how cells create their own gradients for chemotaxis, Trends Cell Biol. 32, 585 (2022).
[2] K. J. Painter, Mathematical models for chemotaxis and their applications in self-organisation phenomena, J. Theor. Biol. 481, 162 (2019).
[3] M. Lauriere, S. Perrin, M. Geist, and O. Pietquin, Learning mean field games: A survey, arXiv preprint arXiv:2205.12944 (2022).
[4] M. Kato, and T. J. Kobayashi, Optimality theory of stigmergic collective information processing by chemotactic cells, arxiv preprint arxiv:2407.15298 (2025).