Limits for Stochastic Reaction Networks

Reaction systems have been introduced in the 70s to model biochemical systems. Nowadays their range of applications has increased and they are fruitfully used in different fields. The concept is simple: some chemical species react, the set of chemical reactions form a graph and a rate function is associated with each reaction. Such functions describe the speed of the different reactions, or their propensities. Two modelling regimes are then available: the evolution of the different species concentrations can be deterministically modelled through a system of ODE, while the counts of the different species at a certain time are stochastically modelled by means of a continuous-time Markov chain. Our work concerns primarily stochastic reaction systems, and their asymptotic properties. In Paper I, we consider a reaction system with intermediate species, i.e. species that are produced and fast degraded along a path of reactions. Let the rates of degradation of the intermediate species be functions of a parameter N that tends to infinity. We consider a reduced system where the intermediate species have been eliminated, and find conditions on the degradation rate of the intermediates such that the behaviour of the reduced network tends to that of the original one. In particular, we prove a uniform punctual convergence in distribution and weak convergence of the integrals of continuous functions along the paths of the two models. Under some extra conditions, we also prove weak convergence of the two processes. The result is stated in the setting of multiscale reaction systems: the amounts of all the species and the rates of all the reactions of the original model can scale as powers of N. A similar result also holds for the deterministic case, as shown in Appendix IA. In Paper II, we focus on the stationary distributions of the stochastic reaction systems. Specifically, we build a theory for stochastic reaction systems that is parallel to the deficiency zero theory for deterministic systems, which dates back to the 70s. A deficiency theory for stochastic reaction systems was missing, and few results connecting deficiency and stochastic reaction systems were known. The theory we build connects special form of product-form stationary distributions with structural properties of the reaction graph of the system. In Paper III, a special class of reaction systems is considered, namely systems exhibiting absolute concentration robust species. Such species, in the deterministic modelling regime, assume always the same value at any positive steady state. In the stochastic setting, we prove that, if the initial condition is a point in the basin of attraction of a positive steady state of the corresponding deterministic model and tends to infinity, then up to a fixed time T the counts of the species exhibiting absolute concentration robustness are, on average, near to their equilibrium value. The result is not obvious because when the counts of some species tend to infinity, so do some rate functions, and the study of the system may become hard. Moreover, the result states a substantial concordance between the paths of the stochastic and the deterministic models.