Genomics analysis of large-scale biological networks and their relationships

Biological networks are of great current interest in biology, particularly with the publication of a number of large-scale networks in yeast. Among them, four (expression, interaction, regulatory, and metabolic) networks have special importance, because they describe four processes that are almost sufficient for a cell to manage its basic activities. Additionally, biologists have the most complete data, though still far from enough, about these networks, because large-scale experiments have been performed mainly on them. In this dissertation, I focused my research primarily on genomic analysis of these four types of biological networks. First, since large-scale experiments are expensive and labor-intensive, I developed an in silico mapping method to generate interaction and regulatory networks in a number of organisms, in which experimental information is extremely limited. Second, I built an automated web tool, TopNet, to calculate and compare topological statistics between any given networks robotically, because most topological calculations are highly repetitious and lengthy. Finally, given the networks (generated by many experimental and computational approaches, including my mapping method) and the tool (i.e. TopNet), I began to analyze topological structures of biological networks at three levels: (1) Individual networks. I examined the relationships between protein's essentiality and its topological characteristics within protein-protein interaction networks. I found that proteins with more interaction partners are more likely to be essential. (2) Network pairs. I compared the local structure of expression and regulatory networks. The results show that genes targeted by the same TF tend to be co-expressed, and tend to have similar cellular functions. (3) Global comparison of all four networks. I found that in terms of overall topological correlation---whether neighboring proteins in one network tend to be nearby in another---all four networks are significantly related. However, focusing on the occurrence of hubs, I found that the four networks can be divided into two classes: regulation and action. The later group---comprising metabolic, expression, and interaction networks---share the same scaffold of central hubs, whereas the regulatory network uses distinctly different ones.