Evolving Additive Tree Model For Inferring Gene Regulatory Networks

With the development of bioinformatics and computer technologies, more and more researchers lay their emphasis on the study of gene regulatory networks(GRNs).At the same time, a huge amount of gene expression data has produced which has built a solid foundation for the research of gene regulatory networks. In recent years, system biology has developed rapidly. It also provides some good theories and methods to study the interactions between different genes from the system level. It is realized that we could have a better and wide understanding of biology based on the systematic analysis of all substances of the system from the viewpoint of the system.There are many methods and models to infer gene regulatory networks. The process of the reconstruction of gene regulatory networks is that basically through using the theory and methods of the bioinformatics and gene expression data, researchers build some appropriate models and then construct gene regulatory networks based on the deduction of the interactions between genes. But it is becoming more and more difficult to find the useful information as choosing the general modals with the increment of gene expression data. The system of ordinary differential equations(ODEs) has some good dynamic properties. It is convenient to describe the complex relationships of the biology networks. Therefore more and more researchers prefer the system of ordinary differential equations as the model of constructing their gene regulatory networks. There are still many difficulties in the evolution of the system of ordinary differential equations, so are the improvements. In this paper we propose an evolutionary method for constructing gene regulatory networks from the time series data using the combination of the system of ordinary differential equations and evolving additive tree as the model. The structure of the system of ordinary differential equations is inferred by evolving additive tree model and its’ parameters are optimized by using the genetic algorithm. And the evolving process of the system of ordinary differential equations can be reduced significantly by partitioning method. We still adopt the parallel computing method to reduce the evolving time.The main research contents are as followed:(1) Overview of the basic theory of gene regulatory networks including background and development status and construction algorithms and models. After knowing the basic background, we emphasized on different approaches of identifying gene regulatory networks. Finally we summarize all their developments and applications.(2) Overview of the basic theories of computational intelligent algorithms including floating-point genetic algorithm and genetic programming algorithm.(3) Introduction of the additive-structure tree model. Firstly, we introduce the constructing approaches and evolution operators of the additive tree model. Then we code the system of ordinary differential equations by using the additive tree model combined with the optimization approach of the equations’ parameters. Lastly we introduce the whole evolving process of the system of ordinary differential equations.(4) Experiments of our revolutionary method for identifying gene regulatory networks. We propose four different experiments to verify the feasibility and effectiveness of the approach and finally analyze the experiment results.