Source-linked AI summary
A practical guide to stochastic simulations of reaction-diffusion processes
Radek Erban, Jonathan Chapman, Philip Maini
TL;DR
Stochastic reaction-diffusion modelling requires simulation methods when molecular fluctuations matter, while deterministic descriptions can miss behaviours such as random switching. The paper provides an accessible, example-based treatment spanning chemical reactions, diffusion, and reaction-diffusion systems, with basic theory for analysis. It also discusses bounded-domain diffusion, biological pattern formation, and extensions involving second-order reactions.
Problem
Small molecular abundances and biologically relevant stochastic fluctuations can make deterministic reaction-diffusion models inaccurate or inapplicable.
Method
The paper teaches stochastic simulation algorithms through examples, covering the Gillespie SSA, chemical master equation, diffusion SSAs, and basic reaction-diffusion methods.
Results
Stochastic simulations capture random switching and oscillatory behaviour that the corresponding deterministic models miss in the presented examples.
Takeaways & Limitations
The paper provides a practical entry point for modelling reaction-diffusion processes and interpreting differences between stochastic simulations and deterministic models.
Takeaways & Limitations
The paper focuses on one-dimensional diffusion problems and does not present the nontrivial generalisation of its SSA to second-order reactions.
Abstract
from arXiv · showhide
A practical introduction to stochastic modelling of reaction-diffusion processes is presented. No prior knowledge of stochastic simulations is assumed. The methods are explained using illustrative examples. The article starts with the classical Gillespie algorithm for the stochastic modelling of chemical reactions. Then stochastic algorithms for modelling molecular diffusion are given. Finally, basic stochastic reaction-diffusion methods are presented. The connections between stochastic simulations and deterministic models are explained and basic mathematical tools (e.g. chemical master equation) are presented. The article concludes with an overview of more advanced methods and problems.
1. Introduction.
The paper offers an accessible, example-based introduction to stochastic reaction-diffusion modelling, combining simulation methods with theoretical tools for analysis. It progresses from chemical reactions to diffusion and reaction-diffusion systems.
- The introduction assumes basic differential-equation knowledge but no prior advanced probability, stochastic analysis, or programming-language knowledge.Examples can be implemented in any computer language, with optional Matlab and Fortran codes available.
- The paper explains stochastic methods through illustrative examples and introduces theory when it provides insight into the example under study.The authors present this example-based approach as more accessible than introducing the complete theory first.
- It begins with single-reaction stochastic simulation and the chemical master equation, then develops the Gillespie SSA for systems of chemical reactions.The chemical master equation is presented as a mathematical equation suitable for analysing the basic stochastic simulation algorithm.
- The paper next studies stochastic simulation algorithms for molecular diffusion and later applies these methods to reaction-diffusion models.The diffusion section focuses on models subsequently used for stochastic reaction-diffusion simulations.
2. Stochastic simulation of chemical reactions.
The section introduces stochastic simulation of chemical reactions through the Gillespie SSA, chemical master equations, and illustrative reaction systems. It shows how stochastic means, fluctuations, and stationary distributions relate to deterministic descriptions, while noting limits for more complex systems.
- Stochastic simulation of degradation: The naive SSA advances in small time steps, using a uniform random number to decide whether degradation occurs with probability A(t)k∆t.The method is correct when ∆t is small.
- Stochastic simulation of degradation: Averaging many stochastic realizations produces a reproducible stochastic mean, whereas individual trajectories depend on the random-number sequence.For the degradation example, the stochastic mean equals the corresponding deterministic ODE solution.
- Stochastic simulation of production and degradation: For production and degradation with k1 = 0.1 sec−1 and k2 = 1 sec−1, the stationary mean and variance are both 10, with fluctuations around the mean.Long-time SSA histograms compare well with the analytically derived Poisson stationary distribution.
- Gillespie algorithm: The Gillespie SSA instead samples the exact time to the next reaction and updates the molecule count using one random number.For multiple reactions, it also selects which reaction occurs using probabilities α_i/α_0.
- Gillespie algorithm: For the two-species system, long-time simulations estimate average molecule numbers of 9.6 for A and 12.2 for B, differing from the deterministic steady states of 10.The stationary values of A and B are correlated across realizations.
- Gillespie algorithm: The two-species chemical master equation cannot be solved analytically and does not yield closed evolution equations for stochastic means and variances.The previous analytical approach therefore does not work for this system.
3. Diffusion.
The paper presents stochastic models for molecular diffusion, covering individual-molecule trajectories and compartment-based simulations, with links to deterministic diffusion equations. Reflective boundaries and the relation d = D/h2 connect the stochastic algorithms to continuum descriptions.
- Individual-molecule diffusion: Diffusion is introduced through stochastic molecular motion, with one-dimensional bounded domains used to incorporate boundary conditions without technicalities.The focus is the interval [0, L], while higher-dimensional problems can be treated similarly.
- Individual-molecule diffusion: The Smoluchowski-based SSA advances each molecule using normally distributed random displacements and reflects positions that cross either boundary.The reflection implements zero-flux boundary conditions when boundaries do not chemically interact with molecules.
- Individual-molecule diffusion: For 1000 molecules, the simulation replaces individual trajectories with a spatial histogram containing molecule counts in 40 bins of length 25 µm at t = 4 min.Ten representative trajectories illustrate motion, whereas the histogram summarizes the ensemble spatial distribution.
- Compartment-based diffusion: The compartment-based SSA models diffusion as molecules moving between neighboring compartments, with rate constant d = D/h2.For interior compartments the mean obeys a discrete diffusion equation, while boundary equations encode the zero-flux condition.
- Compartment-based diffusion: The compartment model yields steady-state mean and variance N/K in every compartment, while its deterministic counterpart is the discretized diffusion equation.The stochastic mean is averaged over many SSA realizations, and the continuum comparison uses the corresponding Fokker–Planck equation.
4. Stochastic reaction-diffusion models.
The paper builds reaction-diffusion models by combining compartmental diffusion with production and degradation, and also describes an individual-molecule formulation. These models are compared with deterministic reaction-diffusion equations, while nonlinear reactions expose important scope limits.
- 4.1. Compartment-based reaction-diffusion SSA: The compartment-based reaction-diffusion model divides [0, L] into K = 40 compartments of length h = 25 µm and represents diffusion, degradation, and localized production as chemical reactions.Diffusion has rate d = D/h2, degradation has rate k1, and production occurs in the first K/5 compartments with k2 = kph.
- 4.1. Compartment-based reaction-diffusion SSA: The Gillespie SSA simulates these reactions using propensities Ai(t)d for diffusion, Ai(t)k1 for degradation, and k2 for production.One realization is plotted at two times and compared with the numerical solution of the deterministic equations.
- 4.1. Compartment-based reaction-diffusion SSA: The stochastic mean satisfies discrete reaction-diffusion equations whose continuum form is a deterministic reaction-diffusion equation with zero-flux boundaries.The parameters D, kp, and k1 are independent of h, and solving the continuum equation is equivalent to solving its discretized form.
- 4.2. Smoluchowski-based reaction-diffusion SSA: An individual-molecule SSA advances positions by the diffusion model, removes molecules probabilistically through degradation, and creates molecules uniformly in the production region.The finite time step must be sufficiently small so that k1∆t ≪ 1 and the corresponding production probability remains appropriately approximated.
- 4.3. Nonlinear chemical kinetics: For nonlinear reaction-diffusion kinetics, the Gillespie SSA is presented as straightforward, whereas extending the individual-molecule SSA is described as nontrivial and omitted.The paper points to external work for the generalized individual-molecule method.
- 4.3. Nonlinear chemical kinetics: For a two-species nonlinear example, the deterministic equations provide a reasonable description compared with one SSA realization but are not equal to the stochastic mean.This distinction separates agreement with a sample realization from equality with an ensemble-average quantity.
5. Two important remarks.
The paper shows why stochastic reaction-diffusion modelling matters when fluctuations produce behaviour that deterministic ODE/PDE descriptions miss, including switching, oscillations, and noisy biological patterns.
- Deterministic modelling can fail when stochastic fluctuations determine biologically observed phenomena or molecular abundances are small.
- 5.1. Deterministic vs. stochastic modelling.: Stochastic simulations reveal random switching between favourable states that deterministic models miss and are necessary to estimate mean switching times.
- 5.1. Deterministic vs. stochastic modelling.: In the Schnakenberg example, deterministic trajectories converge to a steady state while intrinsic noise enables stochastic oscillations by allowing trajectories to leave a stable nullcline.
- 5.2. Biological pattern formation.: The French flag example uses an A concentration gradient and thresholds to distinguish three cell fates, but stochastic realizations produce noisy, variable patterns.
- 5.2. Biological pattern formation.: Turing patterns arise without prepatterning when unequal diffusion destabilizes a constant steady state; stochastic models can switch between patterns with different numbers of peaks.
- The paper presents stochastic reaction, diffusion, and reaction-diffusion simulation methods through simple examples rather than focusing on computer implementation.
6. Discussion.
The discussion places the presented methods within broader stochastic modelling frameworks and identifies boundaries involving boundary interactions and nonlinear chemical kinetics.
- The discussion notes connections among the chemical master equation, Fokker-Planck equation, chemical Langevin equation, and Smoluchowski description.
- The paper presents two diffusion models: a compartment-based reaction-chain model and a Smoluchowski position-jump process.
- The compartment-based diffusion model couples straightforwardly with chemical reaction modelling, while velocity-jump processes provide an alternative for discontinuous molecular motion.
- Reflective boundary conditions apply when molecules do not chemically interact with domain boundaries, but biological interfaces may require reflection and adsorption.
- The chemical-master-equation analysis derives means and variances for illustrative systems without second-order reactions; nonlinear kinetics make analysis significantly harder.