Work

Connecting Different Approaches for Cell Cycle Modeling: Learning Ordinary Differential Equations from Individual-Based Models

Year: 2025

Type: article

Abstract: Background and Objectives: We explore the use of cell cycle agent-based models (ABMs) built on the PhysiCell platform alongside their corresponding surrogate ordinary differential equation (ODE) models. We also address issues related to the stochastic nature, computational cost, and accuracy of ABMs in assessing PhysiCell simulations.Methods: Considering the potential benefits of surrogate representations as substitutes for costly computational models, we showcase the use of a sparse identification machine learning tool, namely SINDy-SA, to build surrogate ODE-based representations for ABMs developed on the PhysiCell platform.Results: We show that three key parameters impact both the computational efficiency and accuracy of the PhysiCell simulations: the time step for cell phase transitions, the initial cell count, and the number of replicates. Examining the stochastic characteristics of simulation profiles for cell cycle subpopulation counts and corresponding fractions over time through probability distributions revealed two primary findings: various established probability distributions can appropriately model their dynamics and the convergence to the mean behavior of subpopulation fractions with respect to the number of replicates is faster than that of the subpopulation counts. Finally, the SINDy-SA framework successfully identified the mathematical structures for all employed cell cycle models, even when including apoptosis-induced cell death.Conclusions: We believe that the insights gained from this study can assist researchers in modeling hybrid discrete-continuous multiscale models for tumor growth using the PhysiCell platform and in developing corresponding surrogate representations through the SINDy-SA framework.

Source: IEEE Transactions on Computational Biology and Bioinformatics

DOI: 10.1109/tcbbio.2025.3565543

Authors: Gustavo T. Naozuka, Heber L. Rocha, Thiago J. Pereira, Gustavo B. Libotte, Regina C. Almeida

Institutions: Laboratório Nacional de Computação Científica, Indiana University Bloomington, Indiana University, Universidade Federal Fluminense, Universidade do Estado do Rio de Janeiro

Language: English

Topic: Model Reduction and Neural Networks

Subfield: Statistical and Nonlinear Physics

Field: Physics and Astronomy

Domain: Physical Sciences

Open Access status: hybrid

Estimated APC paid: $2,645

Funder: Breast Cancer Research Foundation