Research On Prediction Methods Of Essential Proteins Based On Dynamic PPI Network

Essential proteins are necessary in the survival and reproduction of organisms,the research on them can help us understand the minimum requirements of the cellular life.The development of high-throughput technology has resulted in large amounts of available protein-protein interaction(Protein-Protein interaction,PPI)data.By taking advantage of the PPI data,Researchers has proposed a lot of methods to predict essential proteins.Most of those methods regard the PPI network as a static graph.But in fact,PPI network changes with the change of time,environment and some other factors.This important inherent dynamics has not been given enough attention.In this paper,we introduce a dynamic PPI network,and integrate the topology of dynamic PPI network and the biological information of a protein to improve the effect of the prediction of essential proteins.Based on that idea,we propose two methods for the prediction of essential proteins.Given that the dynamic PPI network can more really reflect the change of the network,and the edge clustering coefficient can relatively reduce the negative effects of some false positive edge,we use the sum of edge clustering coefficient to measure the topological property of a protein in dynamic PPI network.Since the amount of information conveyed by different proteins is not the same,we take advantage of the dynamic information entropy in complexes of a protein to evaluate the biological significance of proteins.By integrating the two measures mentioned above,we present a method for predicting essential proteins based on dynamic edge clustering coefficient and information entropy,named CDEI.The comparison results obtained from the PPI networks of Saccharomyces cerevisiae which are downloaded from DIP and MIPS database show that CDEI outperforms these five methods DC(Degree Centrality),LAC(Local Average Connectivity),So ECC(Sum of Edge Clustering Coefficient),Pe C and Co EWC(Co-Expression Weighted by Clustering coefficient),and can predict more essential proteins that are overlooked by other methods.Furthermore,when CDEI is applied in the MIPS network,its advantage will be much more outstanding.On the one hand,the essentiality of a protein is closely related to the modularity of PPI network.The local average connectivity just measures the topological status of a protein from the point of modularity.On the other hand,the essentiality of a protein is the product of complex,and different proteins show different importance in complex.Therefore,the dynamic complex centrality is used to evaluate the biological significance of a protein.Considering both aspects mentioned above,we integrate dynamic local average connectivity and complex centrality,and proposed a new method for predicting essential proteins,named CDLC.The experiment results obtained from the PPI networks of Saccharomyces cerevisiae which are downloaded from DIP and MIPS database show that CDLC outperforms the five methods(DC,LAC,So ECC,Pe C and Co EWC)and can predict more essential proteins that are overlooked by other methods.When CDLC is applied in the MIPS network,its advantage will be much more outstanding.In addition,CDLC performs slightly better than CDEI on the prediction of essential proteins.