Open Collections

Knowledge discovery from large-scale biological networks and their relationships.

University of British Columbia

The ultimate aim of postgenomic biomedical research is to understand mechanisms of cellular systems in a systematical way. It is therefore necessary to examine various biomolecular networks and to investigate how the interactions between biomolecules determine biological functions within cellular systems. Rapid advancement in high-throughput techniques provides us with increasing amounts of large-scale datasets that could be transformed into biomolecular networks. Analyzing and integrating these biomolecular networks have become major challenges. I approached these challenges by developing novel methods to extract new knowledge from various types of biomolecular networks. Protein-protein interactions and domain-domain interactions are extremely important in a wide range of biological functions. However, the interaction data are incomplete and inaccurate due to experimental limitations. Therefore, I developed a novel algorithm to predict interactions between membrane proteins in yeast based on the protein interaction network and the domain interaction network. In addition, I also developed a novel algorithm, a gram-based interaction analysis tool (GAIA), to identify interacting domains by integrating the protein primary sequences, the domain annotations and interactions and the structural annotations of proteins. Biological assessment against several metrics indicated that both algorithms were capable of satisfactory performance, facilitating the elucidation of cell interactome. Predicting biological pathways is one of major challenges in systems biology. I proposed a novel integrated approach, called Pandora, which used network topology to predict biological pathways by integrating four types of biological evidence (protein-protein interactions, genetic interactions, domain-domain interactions, and semantic similarity of GO terms). I demonstrated that Pandora achieved better performance compared to other predictive approaches, allowing the reconstruction of biological pathways and the delineation of cellular machinery in a systematic view. Finally, I focused on investigating biological network perturbations in diseases. I developed a novel algorithm to capture highly disturbed sub-networks in the human interactome as the signatures linked to cancer outcomes. This method was applied to breast cancer and yielded improved predictive performance, providing the possibility to predict the outcome of cancers based on “network-based gene signatures”. These methods and tools contributed to the analysis and understanding of a wide variety of biological networks and the relationships between them.

Text

2010-04-12

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/23353

Bioinformatics

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/