Numerical Researches On Protein-Protein Interactions Network

Along with the primary completion of human genome sequence analysis, bioinformat-ics has emerged and developed as a rising interdiscipline, and has become the hot point and leading research area of multidisciplinary research, such as biology, computer science and applied mathematics. It is also now one of the kernel scientific research areas in 21st century. The grand challenge that people are facing in "Post-genome Era" is proteomics, as completed proteome can further reveal the essence of life phenomenon, explore the rules of life procedure based on interactions and functions of proteins. However, effective approaches in this field are not enough so far. Computer science is applied to proteomics, and protein-protein interaction network (PIN) is investigated by using bioinformatics methods in this thesis.With the PSE-BioServer, a Web Service-based Problem Solving Environment (PSE) for Bioinformatics, the similarities of PINs among different species like Yeast, Drosophila and Human are investigated by using across-species network search methods. These similarities can be used to understand the meaning of Protein-Protein Interactions (PPI), predict the functions and interactions of target proteins, and access to the reserved information during species evolution. The information obtained from network similarities can be retained as references for future research and diagnosis.The innovative results in the four aspects are described below.1. Based on thorough studies of known biomolecular network alignment algorithms, such as PathBLAST and MNAligner, Immediate Network Neighbors Preference Method (INPM) is proposed. The INPM is based on the characteristics of PINs and cross-species search of them is implemented. The INPM emphasizes on biological significance of PINs and reduces errors resulting from the lack of original information. Networks found with the INPM have much higher similarities with target networks, compared to those found by NBM, PathBLAST or MNAligner. Moreover, the computing speed of the INPM is faster than other methods along with the augment of target network.2. In order to meet the computing requirements for mass and multi-species data in biomolecular complex network, the parallel algorithm of the INPM is developed, and implemented on cluster of workstation. So the limitation on the size of the target network has been relieved. It is proved that this parallel algorithm has good speedup and scalability. PINs of Yeast and Drosophila are investigated by using the INPM method. 19 reserved PPIs are detected and 5 unfiled PPIs are predicted. Based on Gene Ontology, new functions of 15 proteins are predicted from bioinformatics aspect as well.3. Different from previous studies on single molecular at gene level, we design experiment from proteomics level to study biological experiment data obtained from Drosophila Parkinson Disease (PD) model, and analyze Drosophila PIN related to PD by numerical methods. The pathogenesis at molecular level for human PD is explored by the experiment designed. This is a novel research idea for PD. Known analysis on inducement factors of PD is verified by discussing major subareas in Human and Drosophila PIN related to PD. Furthermore, the function of the new protein CG2233, which may be closely related to PD, is predicted, and 21 proteins that has direct interaction with differentially expressed proteins and alpha-synuclein are listed, which provide reference for filtering PD drug targets.4. Based on Web Service, a Bioinformatics-oriented PSE-BioServer is initially constructed. The PSE-BioServer with PIN search tools integrated will be the base of sharing resources, cooperating work, and offering easy-to-use high performance computing environment. With the PSE-BioServer, researchers in Bioinformatics will effectively solve the problems encountered by traditional methods.