Efficient Finite Difference Method for Computing Sensitivities of Biochemical Reactions

Sensitivity analysis of biochemical reactions aims at quantifying the dependence of the reaction dynamics on the reaction rates. The computation of the parameter sensitivities poses many computational challenges when taking stochastic noise into account. This paper proposes a new efficient finite difference method for computing parameter sensitivities of biochemical reactions. We employ propensity bounds of reactions to simulate the nominal and perturbed processes and build the estimator. The exactness of the simulation is reserved by applying the rejection-based mechanism. Our approach reduces the variance of the estimator by exploiting the positive correlation of these processes and improves the performance by skipping the propensity updates. Furthermore, by using propensity bounds to couple processes, our approach allows simultaneous perturbation of many reaction rates in computing the sensitivities, which further improves the efficiency of the estimator. We benchmark our method on reaction models to prove its applicability and efficiency.