Exact derivation and practical application of a hybrid stochastic simulation algorithm for large gene regulatory networks

TL;DR

This paper introduces a hybrid stochastic simulation algorithm for large gene regulatory networks that combines stochastic and deterministic methods, significantly improving efficiency while maintaining accuracy.

Contribution

The authors derive exact sampling expressions and demonstrate a partitioning approach for simulating large GRNs more efficiently than traditional methods.

Findings

Algorithm is 11 to 445 times faster than Gillespie algorithm.

Accurately simulates gene regulatory networks with reduced computational cost.

Effective for large systems, including multi-cell interactions.

Abstract

We present a highly efficient and accurate hybrid stochastic simulation algorithm (HSSA) for the purpose of simulating a subset of biochemical reactions of large gene regulatory networks (GRN). The algorithm relies on the separability of a GRN into two groups of reactions, A and B, such that the reactions in A can be simulated via a stochastic simulation algorithm (SSA), while those in group B can yield to a deterministic description via ordinary differential equations. First, we derive exact expressions needed to sample the next reaction time and reaction type, and then give two examples of how a GRN can be partitioned. Although the methods presented here can be applied to a variety of different stochastic systems within GRN, we focus on simulating mRNAs in particular. To demonstrate the accuracy and efficiency of this algorithm, we apply it to a three-gene oscillator, first in one cell, and then in an array of cells (up to 64 cells) interacting via molecular diffusion, and compare its performance to the Gillespie algorithm (GA). Depending on the particular numerical values of the system parameters, and the partitioning itself, we show that our algorithm is between 11 and 445 times faster than the GA.