Research On Pattern Mining Method Of Complex Disease Network Based On Multi-omics Data

Complex diseases are often caused by a combination of genetic,environmental,and other factors.Mining the pathogenic patterns has become a hot topic in revealing the genetic mechanism of complex diseases.Among them,constructing and mining the network of multi-omics data provides new insights for detecting pathogenic patterns.Based on the simulation data and multi-omics data(gene expression data,methylation data,copy number variation data,single nucleotide polymorphism data,etc.)of complex diseases,the genetic interaction network construction and network pattern mining methods are used to detect the pathogenesis of complex diseases,which is mainly divided into the following five parts:(1)To solve the problem of constructing the network for multi-omics data,the interactive part mutual information measurement is proposed to effectively quantify the dependence between genes,and construct the genetic interaction network.Then,this method is applied to construct the interaction network of gene expression data for complex diseases,and further extract the characteristic genes as well as important pathways from the gene network of complex diseases.(2)Based on the high-order interaction detection problem between genetic factors of complex diseases,two kinds of nodes are introduced to construct the high-order interaction network: a real node represents a genetic factor,and a virtual node represents the interactions between multiple genetic factors.High-density subgraphs are extracted from the high-order interaction network and further verify the relevant biological significance.(3)To effectively detect the nodes in the genetic interaction network of multi-omics data for complex diseases,the normalized centrality measure is proposed to comprehensively analyze the local and global features of nodes.Specifically,constructing the complex disease network is conducive to fully mining the information contained in the complex disease data.Besides,the characteristic genes and significant pathways are detected to further verify the biological significance of the network.(4)Based on the robust principal component analysis method,a community mining analysis method is proposed to detect the communities of a complex disease network from multi-omics data.Firstly,the robust principal component analysis method is used to select differentially expressed genes of integrated multiple expression matrices from complex diseases.Then,the genetic interaction network is constructed by these differentially expressed genes.The characteristic genes and important communities are further detected from the network,which are related to complex disease.(5)To simultaneously construct and integrate multi-omics data of complex diseases,the non-negative matrix factorization network analysis method is proposed.Firstly,the heterogeneous networks are constructed by the Pearson Correlation Coefficient of multi-omics data.Then,integrating the heterogeneous networks by the proposed method,which effectively reveals the interaction mechanism between each type of characteristics.Finally,core communities and characteristic genes as well as significant pathways are detected.Experiments have shown that our methods are superior to peer methods and can detect more suspicious complex disease-related genetic factors.