burak ozyurek*
Gebze Technical University
b.ozyurek2021 [at] gtu.edu.tr
Abstract
Agent-based models (ABMs) are highly effective at simulating multi-scale spatial interactions within biological environments. However, standard ABMs typically lack the capacity to mechanistically simulate intracellular metabolic activity at the single-cell level.
To address this issue, we have developed FluxInside: a novel, high-performance Flux Balance Analysis (FBA) extension designed specifically for agent-based models. Utilising the GNU Linear Programming Kit (GLPK), the module was developed and integrated into the PhysiCell/BioFVM ABM platform through optimised, object-oriented C++ architecture. Our algorithm uses a ‘warm start’ to update localised environmental changes dynamically, rather than rebuilding the FBA object. This reduces execution time to under 5 milliseconds per FBA of a genome-scale metabolic model. This architecture seamlessly integrates cell-type-specific genome-scale metabolic models (GEMs) into local niches of agent-based environments.
To validate this computational framework, we simulated a highly heterogeneous glioblastoma microenvironment comprising over 50 thousand agents and dozens of diffusible substrates. Additionally, we tested it across multiple GEMs belonging to various organisms (E. coli, S. cerevisiae, and H. sapiens). Driven purely by intracellular FBA calculations and without any hard-coding, the agents naturally recapitulated emergent, systems-level behaviours including the Warburg effect, TCA switching, necrotic core formation and hypoxia-induced cellular motility. Subsequent in silico therapeutic trials successfully captured the dynamics of metabolic competition underlying immunotherapy resistance.
FlowsInside provides a highly scalable, bidirectional link between intracellular metabolic flux and tissue-scale continuum dynamics. By enabling metabolically aware simulations for tens of thousands of individual agents, this platform offers a powerful new framework for systems biology research, in silico experimental design, and the optimization of targeted therapies.
Keywords: Mathematical modelling, agent-based modeling
Acknowledgement: This project is generously funded by TÜBİTAK UIDB 1071 ( ERA-NET TRANSCAN-3 , 122N905)