Determining the long-term behavior of cell populations: A new procedure for detecting ergodicity in large stochastic reaction networks

Abstract

A reaction network consists of a finite number of species, which interact through predefined reaction channels. Traditionally such networks were modeled deterministically, but it is now well-established that when reactant copy numbers are small, the random timing of the reactions create internal noise that can significantly affect the macroscopic properties of the system. To understand the role of noise and quantify its effects, stochastic models are necessary. In the stochastic setting, the population is described by a probability distribution, which evolves according to a set of ordinary differential equations known as the Chemical Master Equation (CME). This set is infinite in most cases making the CME practically unsolvable. In many applications, it is important to determine if the solution of a CME has a globally attracting fixed point. This property is called ergodicity and its presence leads to several important insights about the underlying dynamics. The goal of this paper is to present a simple procedure to verify ergodicity in stochastic reaction networks. We provide a set of simple linear-algebraic conditions which are sufficient for the network to be ergodic. In particular, our main condition can be cast as a Linear Feasibility Problem (LFP) which is essentially the problem of determining the existence of a vector satisfying certain linear constraints. The inherent scalability of LFPs make our approach efficient, even for very large networks. We illustrate our procedure through an example from systems biology.