Font Size: a A A

Protein-Protein Interaction Network Analysis Based On Spectral Method

Posted on:2008-03-15Degree:MasterType:Thesis
Country:ChinaCandidate:Y Y DongFull Text:PDF
GTID:2120360242998654Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
With the success of human genome project and the finish of model organism sequencing, bioinformatics comes into post-genome era. Proteomics becomes one of the hotspot research fields and the analysis of protein-protein interaction network is becoming a centre task of proteomics. It is important to analyze protein-protein interaction network for the research of proteomics and human disease. What is more, it will be a basic theoretical guide for other research fields of proteomics in post-genome era.On the basis of studying of existing protein-protein interaction network analysis methods, we develop our work on mining and detecting typical characteristic patterns and the structure of communities. We improve current methods, propose new metrics and perform series of experiments for validation. The main contributions of this thesis are summarized as follows:(1) Studying the protein-protein interaction network analysis methods and technologies widely used. A thorough analysis of protein-protein interaction network analysis is given firstly. Then we summarize the difficulties and challenge which protein-protein interaction network analysis is facing now and propose the research scheme of this thesis.(2) Mining typical patterns in protein-protein interaction network. There are many differences between protein-protein interaction network and other networks in real world. Aiming at this problem, we apply spectral analysis method, improve existing model and propose new metrics bipartite coefficient on mining quasi-bipartite patterns in protein-protein interaction network. The results of data experiments indicate that our model can well be used in mining quasi-bipartite patterns in protein-protein interaction network; bipartite coefficient can validate the quality of results efficiently.(3) Detecting the structure of communities in protein-protein interaction network. In order to overcome the disadvantage of traditional spectral bisection method, we improve the spectral bisection method which is based on Normal matrix (i.e. normalized Laplacian matrix) in undirected and unweighted network and propose new metrics community coefficient. The experiments results show that the structure of communities in protein-protein interaction network can be effectively mined by our improved spectral bisection method based on Normal matrix, even if the actual modular structure of protein-protein interaction network seems not so clear intuitively.
Keywords/Search Tags:protein-protein interaction network, protein function annotation, quasi-bipartite pattern, community structure
PDF Full Text Request
Related items