Analysis Of Gene Co-expression Network Based On Heterogeneous Information Network

Nasopharyngeal carcinoma(NPC)is closely related to Epstein Barr virus(EBV),which mostly occurs in southern China and Southeast Asian countries.Although many significant advances have been made in the medical field in recent years for the treatment of nasopharyngeal carcinoma,such as radiotherapy,targeted therapy,immunotherapy,etc.,there is no curative measure at this stage due to the complex etiology of nasopharyngeal carcinoma,which involves the variation of several oncogenes and has problems such as easy local recurrence,easy metastasis and poor prognosis.Therefore there is a great need for effective new biomarkers for the detection and control of nasopharyngeal carcinoma.Since long non-coding RNA(LncRNA)is widely involved in regulating the expression levels of other genes before,during and after transcription,and can also act as a signal,guide,decoy or scaffold molecule to influence the function and localization of proteins and subsequently regulate the transduction pathways of a large number of signals in cells,its status and role in the development of malignant tumors has gradually attracted sufficient attention in related fields and become a novel hotspot for medical research.How to combine LncRNA with the development of nasopharyngeal carcinoma has become a new and valuable research direction.This paper combines the existing theories and technologies of bioinformatics,tumor transcriptomics,gene regulatory networks and heterogeneous information networks,and uses LncRNA and messenger RNA(mRNA)as nodes to construct a gene co-expression network closely related to the development of nasopharyngeal carcinoma,and deeply investigates the expression levels and regulation of LncRNA and mRNA.This will provide a solid foundation for monitoring and treatment of nasopharyngeal carcinoma.First,propose the LncRNA-mRNA gene co-expression network.Specifically,gene expression profiles can be obtained by gene microarray technology and differentially expressed LncRNAs associated with nasopharyngeal carcinoma can be found at first.Then considering the relationship between different types of mRNAs,the relationship between LncRNAs and mRNAs,and combined with the knowledge of gene regulatory network,the LncRNA-mRNA gene co-expression network associated with nasopharyngeal carcinoma can be constructed.Second,a community division algorithm based on genetic algorithm is proposed: the CDGA algorithm.Specifically,the initial population is generated according to the edges in the network,the genetic operators of uniform crossover,basic bit mutation and roulette selection are iteratively run for optimization,and the community is divided by finding the connected graph between the nodes in the network.By analyzing the association partitioning results,the LncRNA and mRNA gene co-expression modules with closer association in the gene co-expression network are identified.Thirdly,a clustering algorithm based on TUCKER decomposition is proposed.During the evolution of nasopharyngeal carcinoma,there may be a promoting or inhibiting relationship between LncRNAs and mRNAs.In order to find this relationship effectively,a clustering algorithm GCEClus based on TUCKER decomposition was designed on the basis of the non-negative matrix decomposition method.Experiments have proved that the proposed clustering algorithm can effectively find LncRNAs and mRNAs that play a core expression role in the evolution of nasopharyngeal carcinoma,help discover LncRNAs and mRNAs that are synergistic or inhibitory to each other,and advance the discovery of co-expression modules of LncRNAs and mRNAs.Finally,a tensor decomposition-based clustering framework for LncRNA-mRNA gene co-expression networks is proposed.Considering the sparsity phenomenon in LncRNAmRNA gene co-expression modules,a more targeted LncRNA-mRNA gene co-expression network clustering model based on CP decomposition and two effective gene co-expression network tensor decomposition algorithms are proposed,in order to screen the key node LncRNAs and further explore their possible role mechanisms in nasopharyngeal epithelial carcinogenesis.