基于糖代谢相关基因的卵巢癌预后模型的构建及验证  

Construction and validation of an ovarian cancer prognostic model based on glycometabolism-related gene signature

作　　者：刘欣悦 韩妍 姚卉 王君霞[1] 索玉平[1] Liu Xinyue;Han Yan;Yao Hui(Department of Gynecology,Shanxi Provincial People'Hospital,Shanxi Medical University,Taiyuan 030001;Department of Gynecology,Changzhi People's Hospital,Changzhi 046000)

机构地区：[1]山西医科大学附属山西省人民医院妇科,太原030001 [2]长治市人民医院妇科,长治046000

出　　处：《现代妇产科进展》2024年第7期517-524,共8页Progress in Obstetrics and Gynecology

基　　金：国家自然科学基金项目(No:61975105);山西省基础研究计划项目(No:202303021212366)。

摘　　要：目的:建立糖代谢相关卵巢癌预后和药物反应的预测模型,探讨其临床意义。方法:由ICGC数据库和GSE26712数据集获取卵巢癌患者的基因表达谱和临床特征数据,从MSigDB中提取并收集糖代谢相关基因与之取交集得到糖代谢相关基因,使用多种算法,筛选出预后相关基因构建模型。对风险模型进行生存分析、基因功能富集分析和药物反应预测,使用cBioPortal在线工具呈现预后相关基因的遗传信息,运用Cytoscape软件显示预后相关基因和糖代谢共表达基因的网络。在正常卵巢组织细胞与上皮性卵巢癌组织细胞中对预后相关基因表达进行差异验证。结果:得到21个糖代谢相关基因进行LASSO回归分析,进一步进行多变量Cox回归分析,建立了包括LHPP(HR=1.51,95%CI为1.24~1.83,P<0.001)、PCK2(HR=0.72,95%CI为0.57~0.92,P=0.009)、PPP3CA(HR=1.35,95%CI为1.08~1.69,P=0.008)和NADK(HR=0.68,95%CI为0.52~1.89,P=0.005)的最优风险模型。使用cBioPortal在398份卵巢癌样本中探索这4个基因的遗传信息,提示基因结构域的改变可能影响蛋白的功能,Kaplan-Meier生存分析显示高风险组的总生存率较低风险组差(P<0.0001),ROC曲线提示模型区分度良好(2年AUC=0.773、3年AUC=0.839、4年AUC=0.852)。通过GSE26712和GSE9891进一步验证风险评分是卵巢癌患者预后独立危险因素,同时基于GSE9891数据集的临床信息,对风险评分、病理分级、FIGO分期、年龄等因素进行多因素Cox回归分析,发现该风险评分为独立预后因素(P<0.001)。KEGG富集分析提示,高风险评分组介导的生物学功能包括细胞周期、TNF、Hedgehog、DNA修复等通路,其中细胞周期通路显著富集(P<0.001),糖代谢相关基因主要通过细胞周期途径在卵巢癌的发生和进展中发挥关键作用;高风险组中,多数化疗药物敏感性更低,并基于细胞周期检查点抑制剂的研究发现4种药物(CGP-60474、BI-2536、CGP-082996和GW843682X)对高危人群�Objective:To establish a prediction model for the prognosis and drug response of ovarian cancer related to glucose metabolism and explore its clinical significance.Methods:The gene expression profile and clinical characteristics data of ovarian cancer patients were obtained from the ICGC database and GSE26712 data set,and glycometabolism-related genes were extracted and collected from MSigDB and intersected with them to obtain glycometabolism-related genes.An algorithm is used to screen out prognosis-related genes to build a model.Survival analysis,gene function enrichment analysis and drug response prediction were performed on the risk model.The cBioPortal online tool was used to present the genetic information of prognosis-related genes.Cytoscape software was used to display the network of prognosis-related genes and glucose metabolism co-expression genes.Finally,in normal ovarian tissue Differential verification of prognosis-related gene expression in cells and epithelial ovarian cancer tissue cells.Results:21 genes related to glucose metabolism were obtained for LASSO regression analysis,and further multivariable Cox regression analysis was performed to establish a gene including LHPP(HR=1.51,95%CI:1.24~1.83,P<0.001),PCK2(HR=0.72,95%CI:0.57~0.92,P=0.009),PPP3CA(HR=1.35,95%CI:1.08~1.69,P=0.008)and NADK(HR=0.68,95%CI:0.52~1.89,P=0.005)Optimal risk model.Using cBioPortal to explore the genetic information of these 4 genes in 398 ovarian cancer samples suggested that changes in the gene domain may affect the function of the protein.Kaplan-Meier survival analysis showed that the overall survival rate of the high-risk group was lower.Group difference(P<0.0001),the ROC curve indicates that the model has good predictive performance(2-year AUC=0.773,3-year AUC=0.839,4-year AUC=0.852).The risk model is further verified through GSE26712 and GSE9891 to be an independent risk for the prognosis of ovarian cancer patients.Factors,and based on the clinical information of the GSE9891 data set,multi-factor Cox regression analy

关 键 词：卵巢癌 糖代谢 预后模型 药敏预测 生物信息学 

分 类 号：R737.33[医药卫生—肿瘤]