| The recent availability of dozens of completely sequenced genomes provides a unique opportunity to understand life's natural history at the molecular level and permits the search for general principles in biology. A critical role in learning about the many genes that compose a genome is played by comparative genomics. Through comparative genomics, functional links between genes can be identified based upon common phylogenetic distributions, conserved proximity along the chromosomes of multiple genomes, and fusion events of genes to a multidomain gene in another organism. In this thesis, these methods are developed and systematically analyzed to assess quantitatively their reliability as gene function predictive methods. A central theme is the nature of the ranging from physical protein-protein interaction to common biological pathway, or more generally, a functional category. A study is presented of the application of combinations of these links to provide insight into the evolution gene fusions in terms of two competing processes: independent gene fusion events and horizontal gene transfer. By combining the three comparative genomics methods to construct networks, the total set of relationships between the genes of an organism is approximated. These networks capture real biological processes, whose paths correspond to known metabolic pathways and functional systems. A giant cluster characterizes these networks, where eight links separate each pair of genes, on average. The giant clusters, in the microbial genomes examined, are universally small-world scale-free networks. The complete set of comparative genomic links is stored in a web-accessible relational database that acts as a central repository designed to bridge computational and experimental methods, which together will provide new insights into biology. |
