Dynamic Protein Complexes Detection And Conserved Complexes Evolution Research

Protein complex comprising of multiple densely connected proteins performs specific biological functions. Dynamic protein-protein interaction (PPI) network can reflect the dynamic evolving process of PPI in a certain extent, which is helpful to reveal the formation and development of the disease. Comparative analysis shows that PPI networks are often conservative, and the homology group retains the mirror function among species.It is of great importance to mine dynamic protein complexes, study the conserved and specific components of protein complexes. At present, the dynamic PPI network and the complex mining algorithm are still needed to be improved, and the dynamic characteristics of the complex are often ignored. In this thesis, we focus on the construction of dynamic protein network by the incorporation of temporal gene expression data, the identification of dynamic protein complex, and the analysis of conserved and specific components of protein complexes across species. Below elaborates the main content and innovation in this paper:(1) A method for the identification of protein activity based on the deviation degree of gene's expression curve is designed, and then the Time-Evolving Protein Interaction Network (referred as TEPIN) is built. Because of the different degree of interaction between the proteins, this thesis proposes a weighted PPI method based on gene expression and connected affinity to measure the degree of interaction between proteins. In order to verify the validity of TEPIN and its weighting method, three classical algorithms are used to recognize protein complexes from TEPIN network and existing classical dynamic networks. The experimental results show that the algorithms in the TEPIN on the specificity, sensitivity, comprehensive score, matching degree and enrichment properties are obviously better than the rest networks; in addition, the weighted network can better reflect the biological properties in protein interaction networks.(2) Based on the topological properties and biological characteristics of dynamic protein interaction network and protein complex, an algorithm (Dynamic Core-Attachment,referred as the DCA algorithm) for mining protein complexes based on dynamic core-attachment is proposed. In this thesis, DCA algorithm is adopted to capture the dynamic features within the core and attachment from the dynamic protein-protein interaction network. The comparative results show that the DCA algorithm performs significantly better than the existing DPC, TS-OCD, CAMSE, ClusterONE, CoreAttach, CPM,MCODE, SPICI and COACH algorithms in multiple evaluation index such as accuracy,hF-measure and functional enrichment analysis, which means it significantly improve the accuracy and the biological significance of predicted protein complex.(3) We study the functional enrichment of the newly added components in the evolution of three species of mice, humans and fruit flies by comparing the protein complexes detected from conservative PPI network and species PPI networks using CAMSE algorithm. Experimental results show that the specific components enriches significantly special biological functions, and indeed they play an important role within the species, thus help to study the evolution mechanism of the species at the protein level.