Bayesian estimation of the rates in a stochastic biochemical reaction network

Dynamics of biological processes is typically specified by a system of coupled biochemical stochastic reactions, whose reaction rates are the unknown parameters. The paper proposes a Bayesian algorithm for estimation of reaction rates of stochastic reactions networks. In the similar vein as the particle MCMC, the parameters (the rates) are estimated in a hierarchical manner: the particle filter is applied to evaluate the likelihood of a proposed parameter vector, while parameter estimation is carried out on the marginal parameter space using an iterative importance sampling scheme. The method is demonstrated with the Lotka-Volterra predator and prey model.