Searching Disease-related Sub-networks In Biological Networks

With the rapid development of biological technique in recent years, amount of biological data is flourishing. As the accomplishment of the whole genome sequencing and development of functional genomics, genome, protome, and interactome data are increasing everyday. The traditional one to one analysis is not adequate for the data scale and biological questions. So the methods of local and global analysis of network appeared. Research on biological networks based on the theory of complex networks has become a hotspot of systems biology. The key point is to study different biological networks in the perspective of local and global characteristics in order to solve biological questions.Facing the more and more complex biological networks, the methods of extracting disease related sub-modules and sub-networks have important implications on reducing the complexity of network analysis and understanding inheritance mechanisms of biological networks and diseases. It also becomes one of the hotspots of bioinformatics. In this thesis, I systematically illustrated the problem in the following two ways: the method of searching sub-networks in complex biological networks and the sub-networks' biological meanings analysis. I elaborated it from the following two aspects:i. Searching method for disease related protein sub-networks combined with expressions of gene chips: we use greedy algorithm to search sub-networks related with Bipolar Disorder by combining protein-protein interaction network with biochip data. The sub-networks we found have good correlations with known disease proteins, and are able to inspect and verify unknown protein functions. This method can successfully extract disease related proteins in the network.ii. Alignment of protein-protein interaction network and disease network: Modularity is an important property of cellular networks. It is generally agreed that proteins or genes in the same module often have the same or similar function. Similar disease phenotypes are begotten by the modular nature of gene network, thus we suppose that all human disease phenotypesalso correlate with genes in a modular style. We constructed a networkrepresentation of phenotypes using the dataset of phenotypic similarity andcompared this network with protein-protein interaction network and extractedBi-modules from the two networks. We introduced an approach ofcomparative analysis of PPI networks and disease networks to address theproblem of finding the optimal global alignment between the two networks,aiming to find a correspondence between nodes and edges of the inputnetworks that maximizes the overall "matchness" between the networks. Themethod is effective in both artificial networks and real networks.In all, we explored the approach of searching sub-networks in biologicalnetworks from different ways and verified its validity. It will help us to understanddifferent types of biological networks in both local and global level and solvepractical problems.