Abstract:

The advent of microarray technology has made it possible to simultaneously monitor and study expression behavior across entire genomes and is an efficient way of gathering information on genetic functions and pathways. However, the typically large number of genes and the complexity of the underlying biological networks make this a formidable task. A common first step to interpret DNA microarray data is the use of clustering techniques. Classifying genes into clusters can lead to interesting biological insights. Since genes with similar functions cluster together, grouping genes of known functions with poorly characterized ones may provide insights into the functions of the latter. Patterns seen in genome-wide expression data can then give indications about the status of cellular processes and information about unknown biological pathways and gene regulatory networks. It is with a rigorously derived set of clusters that we can then use to uncover insights on the control mechanisms these genes are responsible for in a cell and the ensuing biochemical networks involved. The complexity and size of these interacting components cannot be understood by experiments alone. Instead, the development of computational models and the integration of these models with actual experimental design can then provide valuable insight into these systems-level behaviors. In this dissertation, we present a robust clustering algorithm that iteratively identifies the most biologically coherent data groupings from gene expression data. We then develop optimization-based models to predict the most feasible interactions between these gene clusters and their regulatory transcription factors, as well as the optimal linkages between the upstream cellular metabolites. We use as case studies actual gene expression data and glucose sensing signaling pathways in yeast.