Analysis Of Clinicopathological Characteristics Of Ovarian Cancer And Construction Of Diagnostic Prediction Model Based On Multi-omics Data

Objective:To compare the clinicopathological data and survival of patients with ovarian high-grade serous carcinoma and ovarian clear cell carcinoma,to establish a diagnosis model based on methylation by means of bioinformatics analysis,and to verify our diagnosis model in the clinical database.Methods:We collected demographic and clinicopathological data of patients with ovarian high-grade serous carcinoma and ovarian clear cell carcinoma from the Peking Union Medical College Hospital.The clinical data and survival time were compared by t-test.Then we downloaded the database of methylation sequencing of ovarian carcinoma from the Internet,and screened out the different methylation genes of tumor tissue and adjacent tissue.Then,the samples were randomly divided into training set and validating set equally,and the diagnosis model of ovarian carcinoma was constructed using the differential methylation gene in the training set by the logistic regression model.Finally,the diagnostic model was validated in the data set,the score was calculated,the ROC curve was established,and the area under ROC curve was calculated.Results:We collected 120 patients(45.6%)with high-grade serous ovarian carcinoma and 143 patients(54.4%)with clear cell ovarian carcinoma.The proportion of patients with high-grade serous ovarian carcinoma in the late FIGO stage was significantly higher than that of patients with clear cell ovarian carcinoma(P=0.046).The progression free survival time(P<0.001)and the total survival time(P=0.002)of patients with high-grade serous ovarian carcinoma were significantly lower than that in patients with ovarian clear cell carcinoma.Four databases,TCGA,gse26989,gse51688 and gse40105,were downloaded.Five differential methylation sites,cg20131968,cg01214847,cg16986846,cg14992108 and cg11639651 were obtained by differential methylation analysis.In the training set,the diagnostic risk model of ovarian carcinoma was constructed by logistic regression:Y=17.960-15.540*cg20131968-7.171*cg01214847+4.519*cg16986846-6.251*cg14992108-6.887*cg11639651.Then in the downloaded database,the risk scores of the diagnosis model classified the two samples well both in the training set and the verification set.The area under the ROC curve reached 0.994 in the training set and 0.985 in the verification set.In the four data sets,the risk scores of ovarian carcinoma samples were significantly higher than that of normal samples.However,there were no significant differences in the expression of these five genes in different stages of ovarian carcinoma samples.Conclusion:This study compared the differences between ovarian high-grade serous carcinoma patients and ovarian clear cell carcinoma patients.We found that more patients with ovarian high-grade serous carcinoma were in the late stage,and the survival rate of patients with ovarian high-grade serous carcinoma was significantly lower than that of patients with ovarian clear cell carcinoma.Therefore,the establishment of a diagnostic model for ovarian carcinoma has a high clinical value.Then,using bioinformatics method,we analyzed the genes with methylation difference between ovarian carcinoma and normal tissue,and constructed a logistic regression model to diagnose ovarian carcinoma,which can provide reference for the diagnosis of ovarian carcinoma patients.