Identification Of Essential Proteins Based On Multi-Layer Protein-Protein Interaction Networks

Protein is the material basis of life.All life activities of living things depend on the interaction between a series of proteins.The protein interaction network(PPI network)is a biological network that describes the interaction between proteins.The essential protein is an important node in the PPI network that undertakes important functions.Therefore,detecting and identifying essential proteins from a network perspective is an important research content of bioinformatics.At present,the research of essential protein recognition based on network(graph structure)mainly focuses on two aspects: 1)use a variety of information fusion to build a more reliable PPI network;2)graph structure based centrality measurement method or machine learning method.However,these studies are basically carried out for single-layer networks.It is rare to construct PPI networks from the perspective of multi-layer sequential networks and study their properties.This is exactly the research goal and task of this article.The main work of this article is as follows:(1)A generalized 3Sigma threshold methodThe 3Sigma method is a commonly used threshold method to determine protein activity status based on gene expression level data,and plays an important role in the construction of active PPI networks.This method adjusts the degree of deviation of the threshold from the mean through the k-value coefficient and the F function.In order to further improve the recognition rate of essential proteins in the PPI network,we added an h parameter to the F function.By adjusting the parameter h,we can effectively adjust the threshold setting of gene expression data of different discrete degrees.When h = 2,revert to the 3Sigma method.Experiments show that compared with the 3Sigma method,when the parameter h is set to 0 ? 1,the PPI network constructed by the generalized 3Sigma method has a higher recognition rate of key proteins.(2)A construction method for multi-layer active PPI networksAt present,the construction of active PPI network mainly uses the threshold method to determine the protein's active state at each observation time point,and then calculates the active interaction set,and the active PPI network is derived from the static PPI network from the set.A multi-layer active PPI network construction methodis proposed.First,the threshold method is used to calculate the active protein set at each observation time point,and the active PPI network at each time point is derived from the static PPI network from each set,and finally the multi-layer active PPI network is formed.Experiments show that the number of active nodes and active interactions at the T=8 network layer is the highest.Compared with other sequential layers,the three central methods have the highest recognition rate of essential proteins in the T = 8 sequential layer.Among them,the highest recognition number is 85 in Top100,at least 3.66% higher than single-layer active PPI network,and the highest recognition number in Top600 is 346,at least 26.74% higher than single-layer active PPI network.(3)A multi-layer weighted average centrality methodBased on the multi-layer network,a multi-layer weighted average centrality measurement method is proposed to solve the problem of identifying essential proteins in the multi-layer network.The method first measures the centrality of the active proteins at each network layer,and then weights and sums the centrality values of each layer of each protein according to the given layer weight coefficients,and divides by the number of active layers of the protein to get the final centrality value of the protein.Finally,the ranking method is used to calculate the recognition rate of essential proteins.Experiments show that the multi-layer PPI network has a higher recognition rate of essential proteins.The highest recognition number of Top100 is 88,which is 7.32% higher than that of single-layer active PPI network,the highest recognition number of Top600 is 376,which is 16.05% higher than single-layer active PPI network.In this paper,from the perspective of multi-layer networks,relevant researches are carried out in terms of threshold methods,multi-layer network construction methods,multi-layer network centrality measurement methods,etc.Proposed generalized 3Sigma method,multi-layer active PPI network construction method,multi-layer weighted average centrality method.It provides an exploratory approach for studying protein interaction properties and essential protein detection in multi-layer networks.