Accurate Techniques for Parametric Sensitivity of Stochastic Discrete Biochemical Networks

The Chemical Master Equation is a discrete stochastic model of homogeneous biochemical networks. This model has been successfully employed to develop a wide range of biochemical networks that arise in applications. In this thesis, we consider a local sensitivity analysis with respect to the initial conditions for the Chemical Master Equation. Sensitivity is a vital tool for analyzing the robustness of the model output with respect to small fluctuations in its parameters. We utlize some existing finite-difference sensitivity estimators, the Common Random Numbers and the Common Reaction Path methods. Unlike the sensitiviity with respect to the kinetic parameters, the sensitivity considered here can only be applied for integer-valued perturbation parameters. The methods discussed are tested on some models of pracitcal interest of small and large sizes. The results demonstrate that these methods are accurate and reliable.