Computationally comparing biological networks and reconstructing their evolution

摘要: Biological networks, such as protein-protein interaction, regulatory, or metabolic provide information about biological function, beyond what can be gleaned from sequence alone. Unfortunately, most computational problems associated with these networks are NP-hard. In this dissertation, we develop algorithms to tackle numerous fundamental in the study of networks. First, present a system for classifying binding affinity peptides diverse array immunoglobulin antibodies. Computational approaches problem integral virtual screening and modern drug discovery. Our is based on an ensemble support vector machines exhibits state-of-the-art performance. It placed 1st 2010 DREAM5 competition. Second, investigate network alignment. Aligning different species allows discovery shared structures conserved pathways. We introduce original procedure alignment novel topological node signature. The pairwise global alignments produced by our procedure, when evaluated under multiple metrics, both more accurate robust noise than those previous work. Next, explore ancestral reconstruction. Knowing state us examine how pathways have evolved, extant diverged that their common ancestor. describe framework representing evolutionary histories efficient reconstructing either single parsimonious history, near-optimal histories. Under models evolution, effective at inferring interactions. Additionally, approach noisy input, used impute missing interactions experimental data. Finally, framework, GrowCode, learning growth models. While work focuses developing manually, procedures parameters existing models, GrowCode learns fundamentally new match target flexible user-defined way. show learned produce whose properties real-world closely