Modeling Genetic Regulatory Networks by Sigmoidal Functions: A Joint Genetic Algorithm and Kalman Filtering Approach

In this paper, the problem of genetic regulatory network inference from time series microarray experiment data is considered. A noisy sigmoidal model is proposed to include both system noise and measurement noise. In order to solve this nonlinear identification problem (with noise), a joint genetic algorithm and Kalman filtering approach is proposed. Genetic algorithm is applied to minimize the fitness function and Kalman filter is employed to estimate the weight parameters in each iteration. The effectiveness of the proposed method is demonstrated by using both synthetic data and microarray measurements.