Abstract
We present a maximum likelihood method for inferring kinetics of stochastic systems of chemical reactions, given discrete time-course observations of the abundance of either some or all of the molecular species and a BlenX model of the system. BlenX is a process calculus providing a tool and algebraic laws for a high-level description of interactions, communications, and synchronizations between processes representing the biomolecules. BlenX offers an efficient alternative to differential equations, but it poses different challenges to the model calibration. The main difficulty is the sampling of the reaction pathways between two observed states. We define a maximum likelihood function in terms of reaction propensities and we estimate it by sampling the intermediate pathways from the transition system of a BlenX. The method of sampling the transition system is inspired to the elementary mode analysis. Our method is illustrated with the example of a BlenX model of chaperone-assisted protein folding.