| Cell functions are regulated by the network of interactions among different components in the cell, such as genes, proteins, and small regulatory RNAs. Although biological and biomedical researchers have successfully identified most of the components and many of the interactions, unfortunately, these collections of components and interactions do not offer convincing concepts to understand the functions and behaviors of biological systems. In this dissertation, we develop computational tools to model and analyze biological networks, and to study biomolecular functions in the context of biological networks. Specifically, we make contributions in two important areas.; First, we develop and apply data mining methods to model gene transcriptional regulatory networks by integrating information from multiple sources. We extensively study the performance and capability of decision-tree methods in reconstructing transcriptional regulatory networks in the budding yeast. We also propose a multivariate regression tree method, called Bi-dimensional Regression Tree (BDTree), which can model gene transcription using cis regulatory elements and transcriptional regulators simultaneously.; Second, we develop computational algorithms for detecting the so-called community structures, which are important structural and functional elements of many complex networks including biological and social networks, and apply the algorithms to relate the structural properties of biological networks to the functions of bimolecules. We show that our algorithms have significantly advanced the state of the art in automatic community discovery. We demonstrate the advantages of our algorithms with several biological applications in analyzing gene expression microarray data and protein-protein interaction networks. |
