Mining Microbial High-order Modules Based On Hypergraph Clustering

Microbial communities are ubiquitous and have a huge impact on human health and the quality of the environment.The function of the microbial community is closely related to the community composition,microbial higher-order interactions often control the functionality of microbial communities.The biological network is modular,and the microbial module is very important to the stability and recovery of the system.At present,there are few researches on the relationship between microbial high-order interactions and the mining of microbial high-order modules,so it is an urgent need to explore the microbial high-order modules.In this paper,based on the microbial abundance data in HMP from open source,information entropy is used to construct the high-order logical relationship of microorganisms.The microorganisms that meet certain conditions are formed into a triplet,these microorganisms as the nodes of a hyperedge in the hypergraph,a hypergraph composed of higher order interactions of microorganisms.Secondly,the properties of intra-class scatter matrix are introduced to reconstruct the hyperedge similarity matrix.Finally,the game theory is introduced to transform the hypergraph clustering problem into the non-cooperative multi-player game problem.The main research work is as follows:First,hypergraph clustering based on the intra-class scatter matrix is used to mine the microbial higher-order modules.Most of the modular cluster analysis in previous studies was based on paired interactions between microorganisms.In this paper,based on the human body microbial abundance data,the information entropy is used to calculate the high-order logical relationship between microbial triples in the human body,and constructing hypergraph based on high-order interaction relation network.Based on the hypergraph clustering method,hypergraph clustering algorithm based on the intra-class scatter matrix(HCIS)is proposed to excavate the higher-order modules of microorganisms,and the optimal clustering results are obtained by maximizing the modularity.Experimental results and visual analysis verify the effectiveness and feasibility of HCIS algorithm for higher order module analysis of microorganisms.Second,hypergraph clustering mining microbial high-order modules based on evolutionary game theory.Considering that there is a certain error in selecting the optimal number of clusters by maximizing the degree of modules within the range of the number of clusters,this paper proposes the microbial hypergraph clustering based on evolutionary game theory and the idea of intra-class scatter matrix(HCGI),and introduces the game theory to transform the hypergraph clustering problem into a non-cooperative multi-player game problem.This method can automatically generate clustering without specifying the number of clusters,and eliminate nodes with weak relationship strength in modules.The experimental results and the visual analysis directly prove the reliability of the HCGI method in the analysis of higher-order modules of microorganisms.