Inference Of Gene Regulatory Networks Using Genetic Programming And Filtering Algorithm

There are behaviors of gene interaction and regulation in the process of gene expression. The expression of individual gene cannot reveal the inherent law of the phenomenon of life, so gene expression should be studied systematically. With the development of high-throughput DNA microarray technology, it is possible to obtain a large number of biological gene expression data in a short time. It provides a data base for constructing gene regulatory network. Biologists can cognize the regulation of gene expression and highly complex life phenomena from the view of the system by constructing gene regulatory network, which is helpful to study the important medical issues, such as occurrence and development of diseases.In recent years, there are more and more mathematical model to construct gene regulatory network, in which differential equation model is more effective to describe the evolution process of biological macromolecules with time.In this paper, we focus on the construction of differential equation model of gene regulatory network and improve the modeling accuracy of differential equation model by proposing new algorithm for regulatory network identification. The main job in this paper as follow:(1) An gene regulation network identification algorithm by genetic programming and normalized subband adaptive filter is proposed. Genetic programming is applied to identify the structure of the model and normalized subband adaptive filter is used to estimate the parameters. This algorithm can accurately identify the model structure and reduce the correlation of the time series data, improve parameter identification accuracy of adaptive filtering algorithm.(2) An gene regulation network identification algorithm by genetic programming and particle filter is proposed. Genetic programming is applied to identify the structure of the model and particle filter is used to estimate the parameters. This algorithm can accurately identify the model structure, the particle filter algorithm is not sensitive to the nonlinear intensity and the noise pattern of the model, and can get more accurate parameter identification result for different nonlinear systems.Two algorithms for differential equation model identification are proposed and they can reduce the influence of noise and the correlation of time series. Compared to the previous algorithm, they have higher accuracy by simulation experimental verification.