Detecting Biomarkers Of Hepatocellular Carcinoma Based On Differential Network Topological Parameters

There are many types of liver cancer.Among them,primary liver cancer is highly prevalent in China,and the main type of it is hepatocellular carcinoma(HCC).HCC is hidden and progresses quickly,so it is often to miss the best time for diagnosis and treatment and thus cause cancer cells to metastasize or enter local severe stage.Therefore,it's difficult to cure and has a poor prognosis,which seriously threatens lives and health.It is a high incidence cancer worldwide.The etiology and exact molecular mechanism of liver cancer are not fully understood,and its occurrence is currently believed to be the result of a combination of multiple pathogenic factors.The early diagnosis of liver cancer is very important,but the current diagnosis of HCC is mainly based on imaging examination.Although China still uses the serum molecular marker alpha-fetoprotein(AFP),the specificity of AFP is not high and false positive results are prone to occur.In recent years,there are some potential HCC biomarkers at the research stage,but most of them have the disadvantages of low specificity and low sensitivity.Therefore,it is urgent to detect new specific and sensitive biomarkers to assist in the clinical diagnosis of HCC,and it has significant social value and economic benefits.Due to the complexity of its pathogenesis,cancer is considered to be a "systemic disease" caused by the interaction of multiple genes,gene products,metabolites,genetic and environmental factors.This indicates the necessity of detecting biomarkers based on the background of complex systems and complex networks.Therefore,this thesis combined gene expression data with the Gene-Gene Interaction(GGI)network to detect biomarkers by studying the significant changes in the topological position of genes and other biomolecules in the interaction network during the disease pathogenesis.The selected candidate markers were verified on multiple aspects.In addition,the method propose in this thesis has been compared with the methods of detecting biomarkers based on other different network components.The expectations and visions for the future research direction of biomarkers were finally proposed.The main research contents of this thesis are as follows:(1)A new model for detecting biomarkers is proposed,which mainly used gene expression data and human gene-gene interaction networks to construct specific gene-gene interaction networks in disease and control state respectively.Sorting and clustering the network topological parameters,removing the redundant parameters and selecting the network topological parameters applicable to this study.Then,based on the difference in the topological positions of genes in the gene-gene interaction networks of disease and control,selecting genes with significant difference in the network topological parameters,clustering the network composed of these genes.The module with the best classification performance in the machine learning model is prioritized.The 33 genes contained in the module are used as candidate biomarkers,called TopMarker.Finally,the functional enrichment analysis of these candidate biomarkers and the validity verifications at other aspects demonstrate that the 33 TopMarker genes have a good classification ability and have close relationships with the pathogenesis of HCC.(2)To further illustrate the rationality and superiority of the method in this thesis,a comparative study of different methods is carried out at the network component level.Listing commonly-used network components in complex network research,such as node,edge,clique and pathway,and detecting biomarkers based on these different network components.The results of different network components are compared with those of TopMarker,which further proves the correctness of the network-based biomarker discovery method and the credibility of the results.