One object of systems biology is to develop mathematical models of signalling pathways and metabolic systems. When the model structure is established using mass balance and/or other physical principles, accurate parameter estimation is expected to obtain unknown parameter values. Due to the high nonlinearity in system models, the large number of parameters involved, the inadequate measurement data in experiments and the noise pollution, etc., parameter estimation has always been a challenging issue in systems biology. In this thesis, the problem of parameter estimation of cell networks has been studied with the TNFαmediated NF-κB signal transduction pathway as the case for study. Some creative ideas and innovative algorithms have been proposed to provide efficient parameter estimation for this type of complex systems. The main contributions can be summarized as follows: 1.For a class of signalling pathways, the minimum states set for parameter identification is proposed based on linear system theory, the Levenberg-Marquardt is also used to estimate unkown parameters of signalling pathways. 2.A new parameter estimation method has been proposed by curve-fitting of model states and applied to a signal transduction model. With this method, the original optimization problem has been transformed into a simple linear or nonlinear regression problem and the computational load is reduced effectively. 3.With states augment method, Extended Kalman filter and Unscented Kalman filter are applied to parameter estimation of signaling pathway models. Based on the simulation of an NF-κB model, comparisons have made for the estimation performance using these two filters. It shows that the Unscented filter-based algorithm provides a better estimation of the states and the parameters. 4.An innovative parameter estimation algorithm has been developed based on the probability density function (PDF) of the measurement states. Two performace functions and there substitutes based on histogram of output are proposed. The method is applied to NF-κB model and the simulation resultes illastrate the effectivness of the method.
修改评论