The genetic regulatory network, which is constructed from the time-courses data sets, is always described as highly nonlinear differential equations. Mathematical and computational modeling technologies focus on efficiently identifying the parameters of the nonlinear dynamic biological system. Various derivative-free and derivative-based optimization technologies have been proposed recently to infer the parameters of the S-type genetic regulatory networks (S-systems). The S-system is described as coupled power-law functions. As the involved genes and/or proteins increase, the identification becomes increasingly difficult; multiple attractors exist in the system. How to develop an optimization algorithm to reduce the computation time while keeping the accuracy is necessary. In this study, a gradient-based metaheuristics is proposed. The computational method starts with the hill-climbing optimization, and solves the stagnation phenomenon by using a differential climbing operation and migration synchronous evolution. This method was tested with four biological systems. To show the performance in the solution quality and the computation time, we let the learning be implemented in a wide search space ([0, 100] for rate constants and [-100, 100] for kinetic orders) and initialized all parameters at a bad point (the neighbourhood of 80).

Author Name: Cheng-Tao Wu, Shinq-Jen Wu, Jyh-Yeong Chang

URL: View PDF

Keywords: Parameter Estimation; S-System; Memetic Computation; Genetic Algorithm

ISSN: 2326-5825

EISSN: 2326-5833

EOI/DOI:

Add Citation Views: 1

---