Static And Dynamic Gene Regulatory Network Construction Method And Analysis

Gene chip technology is an important scientific and technological breakthrough in the post-genome era. It can efficiently provide us with massive and accurate bioinformatics. The scientific research on function research has been promoted from the molecular level to the system level. Many methods about microarray data mining techniques are investigated by researchers, which include gene selection, clustering analysis, and gene network construction etc. The purpose is to identify the gene expression, the relationship between genes, and dig out the mechanisms how genes respond changes in the external environment. It can help us get a better understanding of the phenomenon of life, and then reveal the essence of life.In this dissertation, many problems about construction methods of the dynamic and static gene regulatory network are researched. The research of dynamic network can deeply help us grasp systems'dynamic features, understand the interactive mechanisms between genes and the mechanisms in which genes regulate the functions of cells or tissues. In this work, a multi-source information fusion state-space model(MIF-SSM) is introduced to improve the modelling accuracy and the inferred gene regulatory networks. Variance Bayesian (VB) methods are used to infer the model structure of MIF-SSM and avoid overfitting. The method captures the dynamic nature of the biological system and naturally incorporates other data from transcription factor binding location data into the original gene expression data. We test on a well-established model of T cell activation, and do a comparative analysis of the situations of using expression data only and combining transcription factor binding sites. It's proved that MIF-SSM can dramatically improve the modelling accuracy of gene regulatory networks. Through the analysis of static gene networks, we can distinctly understand the structure of biological systems. Currently, lots of methods have been applied into clustering analysis of the gene expression data, and we select three typical ones to explore yeast cycle gene expression data and mouse'brain gene expression data. By evaluating the clustering results from data structure and biological function, we find the most suitable clustering methods for these species clustering method. With the information of clustering analysis, we can further explore the gene transcription regulatory networks and better grasp the essence of life.