Study Of Gene Regulation Network Construction Algorithm

The core issue discussed in this thesis is on how to construct gene regulatory networks related with different biological issues from a computational perspective. High-throughput biological datasets,such as gene expression microarray and ChIP-chip,set up the informative data foundation for constructing gene regulatory networks.In this thesis,these high-throughput biological data are regarded as the inputs of the mathematical models,and system identification technique and statistical methodology are synthesized to determine significant gene regulatory network elements of specific biological issues.One focus in this thesis is to construct the gene regulatory network across the cell cycle.Identifying significant transcription factors(TFs) that are cell cycle regulators, as well as cell cycle-controlled combinatorial interactions among TFs,provides an access to witnessing the cell cycle-related regulatory circuitry in yeast.In terms of cell cycle-related microarray data and ChIP-chip data,dynamic transcriptional regulatory model and statistical analysis of temporal correlation are combined to identify these network elements.The other focus is to construct gene regulatory networks describing those adaptive mechanisms established in response to various environmental stresses. Based upon the genomic expression patterns,linear regression model is utilized to describe the regulatory relationships between TFs and target genes,and statistical analysis on these regulatory relationships is examined to reveal significant TF-target regulatory contents under various environmental stresses.In this thesis,different methods have been developed to construct gene regulatory networks in different biological backgrounds.The results show that our methods not only provide verified gene regulatory network elements but also reveal potential network elements of statistical significance that are not supported by up-to-date experiments or computational methods.Moreover,exploring the gene regulatory networks from a computational perspective could largely save a great deal of expenses on biological experiments,and better guide further biological experiments.