Research On Prognostic Biomarkers Of Ovarian Cancer Based On Several Biomolecular Networks

Ovarian cancer is a common gynecological malignancy and one of the most fatal female diseases.Due to the inconspicuous early symptoms and a lack of reliable early screening modalities,about 70% of ovarian cancer patients are diagnosed with advanced stages.The survival rates of ovarian cancer patients at different stages vary widely,and over 75% of patients with late-stage ovarian cancer die of the disease.Therefore,there is an urgent need to find new biomarkers to improve the prognosis of ovarian cancer,so as to improve the efficiency of individualized treatment of patients.Based on several biomolecular networks,such as gene co-expression networks and protein-protein interaction networks,this thesis uses ovarian cancer high-throughput data to construct prognostic related biological networks.Novel prognostic biomarkers are identified based on network topology and pathway activity,and reliable prediction models are established.The main work is as follows:(1)A weighted gene co-expression network constructed by ovarian cancer prognostic-associated genes was used to identify prognostic biomarkers and predict patient risk.First,data of 320 patients with ovarian cancer were obtained from the The Cancer Genome Atlas(TCGA)database,and 747 prognostic-associated genes were selected by Cox univariable regression to construct a weighted gene co-expression network.Then,considering the biological significance of the network,the module in co-expression network was re-weighted by integrating the protein-protein interaction data from the module genes.And the module genes were ranked according to the topological properties of the genes in re-weighted network.Finally,the Cox proportional hazards model was employed to construct prognostic models by these topologically important genes.By considering a balance between the model prediction ability and the number of genes,three prognostic biomarkers were identified.Survival analysis showed that the three biomarkers can significantly distinguish patients with different prognosis.(2)In order to explore the prognostic value of alternative splicing events in ovarian cancer,the alternative splicing events were analyzed systematically.First,the data of ovarian cancer patients were obtained from TCGASpliceSeq database,and 290 prognostic-associated alternative splicing events(PASEs)were selected by using univariate Cox regression analysis.Then,the protein-protein interaction data of PASEs were combined to construct a prognostic weighted network for ovarian cancer.At the same time,topological analysis and functional enrichment analysis were performed on the network.Finally,20 alternative splicing events were selected as prognostic features according to the degree of nodes in the network,and the Cox proportional hazard model was used to establish a prognostic model by these features.The results showed that these alternative splicing events can be used to predict the prognosis of ovarian cancer patients,and as potential biomarkers for ovarian cancer prognosis.(3)Identify pathway biomarkers related to the prognosis of ovarian cancer at the level of functional categories.Gene products usually cooperate in the form of functional modules,so compared with study(1)gene biomarkers and study(2)alternative splicing event biomarkers,the biomarkers at the level of functional categories are more stable.Enrichment analysis ofprognostic-associated genes yielded prognostic-associated pathways.A global protein interaction network was constructed,and the topological importance of genes in the network was evaluated by random walks.The weights of genes were adjusted according to the topological importance,and the prognostic-associated pathway activities of ovarian cancer were inferred with the gene expression profile.A Cox model was also constructed to predict survival outcomes.The results showed that the pathway-based method achieves better overall prediction performance than the gene-based method in study(1).Pathway activity accumulates greater discrimination than single gene biomarker and is more robust in predicting survival outcomes.