Effects of different discretisations of the Laplacian upon stochastic simulations of reaction-diffusion systems on both static and growing domains

TL;DR

This study investigates how different spatial discretisations and derivation methods of diffusive jump rates influence stochastic reaction-diffusion simulations, especially for complex systems like Turing pattern formation on static and growing domains.

Contribution

It systematically analyzes the impact of discretisation choices on stochastic model predictions, emphasizing the importance of careful implementation for accurate results.

Findings

Minor differences in simple systems due to discretisation choices

Large discrepancies in complex reaction kinetics predictions

Discretisation impacts are more significant for pattern-forming systems

Abstract

By discretising space into compartments and letting system dynamics be governed by the reaction-diffusion master equation, it is possible to derive and simulate a stochastic model of reaction and diffusion on an arbitrary domain. However, there are many implementation choices involved in this process, such as the choice of discretisation and method of derivation of the diffusive jump rates, and it is not clear a priori how these affect model predictions. To shed light on this issue, in this work we explore how a variety of discretisations and method for derivation of the diffusive jump rates affect the outputs of stochastic simulations of reaction-diffusion models, in particular using Turing's model of pattern formation as a key example. We consider both static and uniformly growing domains and demonstrate that, while only minor differences are observed for simple reaction-diffusion systems, there can be vast differences in model predictions for systems that include complicated reaction kinetics, such as Turing's model of pattern formation. Our work highlights that care must be taken in using the reaction-diffusion master equation to make predictions as to the dynamics of stochastic reaction-diffusion systems.