Research On Learning Gene Regulatory Network And Metabolic Network Based On Differential Equation

As the number of biology completed the gene sequencing growing rapidly and making a wide of understanding of the gene function, biological network including metabolic network, gene regulatory network, signal transduction network, etc, has been emphasized on in the bioinformatics area. It is realized that various biological components including genes, proteins, metabolites, etc, function through the interaction with other parts, so we understand comprehensively bio-physiological function by making the overall analysis of all relevant components on the system level. It is a difficult problem that how to use the genomic data to analysis the gene regulatory network and metabolic pathway. In recent years, as the development of computer technology, more and more modeling patterns are used in the gene regulatory network and metabolic pathway. With the large amounts of data collection, the model in general is difficult to extract information from the data. Because the system of differential equation is very powerful and flexible, it is in favor of describing the complex relationship of biological networks and used as model to analysis the biological networks by the researchers recently. The evolutionary process of differential equation is still in the infancy, the structure and parameter have large errors, and modeling process needs the excessive calculation. Therefore, there is the need for a new optimization strategy to gain an accurate differential equation, so as to better address practical problems in systems biology. In this paper, we propose an evolutionary method for identifying the gene regulatory network and metabolic network from the observed time series data using a system of ordinary differential equations (ODEs) as a model of network. The structure of ODE is inferred by the Multi Expression Programming (MEP) and the ODE's parameters are optimized by using particle swarm optimization (PSO). And by partitioning the search space of system of ODEs can be reduced significantly. Because the additive tree model is simple in the form and very near to the traditional representation of the system, tree-structure based evolution algorithm and particle swarm optimization are employed to evolve the architecture and parameters of the additive tree models for the system of ODE identification, respectively. And we make some restrictions to the addictive tree model in order to identify the large gene regulatory network. The contents are listed as follow:(1) Overview the basic theory of gene regulatory networks and metabolic pathways, including the concepts, developments and research methods of the gene regulatory networks and metabolic pathways, focus on various model approach of identifying gene regulatory networks, summed up out our thought of identifying gene regulatory networks and metabolic pathways. Overview the concept, application areas and solution of differential equation models, application areas and provided a detailed overview of solution.(2) Introduce multi expression programming and particle swarm optimization. Firstly overview the basic idea of evolutionary algorithms, and secondly emphasize the basic theory of genetic programming and then introduce the basic idea and realization of multi expression programming and particle swarm optimization.(3) Propose a hybrid optimization strategy that the structure of ODE is inferred by the Multi Expression Programming and the ODE's parameters are optimized by using particle swarm optimization, an algorithm that the additive tree model is used to model the system of ordinary differential equation, and make some restrictions to the addictive tree model.(4) Apply the hybrid optimization strategy, additive tree model and restricted additive tree to a large number of gene regulatory networks and metabolic pathways instance. The results show that the hybrid strategy and our models are more effective than the traditional methods and models.