基于公共数据库构建甲状腺癌差异基因Cox比例风险回归模型对患者预后评估价值  

Value of Construction of Cox Proportional Hazard Regression Model for Differential Expressed Gene of Thyroid Cancer Based on Public Databases in Prediction of Prognoses of Patients

作　　者：姜雪丽[1] 赵辛[2] 赵铁铮[3] JIANG Xue-li;ZHAO Xin;ZHAO Tie-zheng(The First Department of Thoracic Surgery,Central Hospital of Handan City,Handan,Hebei 056001,China;Department of Obstetrics and Gynecology,the Fourth Hospital of Shijiazhuang City,Shijiazhuang 050000,China;Department of General Surgery,People's Hospital of Huanghua City,Huanghua,Hebei 061100,China)

机构地区：[1]邯郸市中心医院胸外一科,河北邯郸056001 [2]石家庄市第四医院妇产科,石家庄050000 [3]黄骅市人民医院普外科,河北黄骅061100

出　　处：《临床误诊误治》2021年第11期49-54,共6页Clinical Misdiagnosis & Mistherapy

基　　金：河北省医学科学研究课题计划(20201256)。

摘　　要：目的探讨基于公共数据库构建甲状腺癌差异基因Cox比例风险回归模型对患者预后评估价值。方法对GEO数据库中甲状腺癌数据集GSE138198和GSE50901进行批次校正,联合TCGA数据库,筛选甲状腺癌样本中发生改变的基因。利用R软件将TCGA数据库中甲状腺癌患者随机分为试验组和验证组,在试验组中建立Cox比例风险回归模型,采用Kaplan-meier生存曲线分别在试验组、验证组和整体组(TCGA数据整体)中分析高和低风险组生存状态,采用受试者工作特征(ROC)曲线评估模型预测甲状腺癌患者生存率准确性。分析甲状腺癌高和低风险组差异基因并进行GO和KEGG功能富集分析。采用单因素和多因素Cox回归分析探讨甲状腺癌患者临床病理特征与预后关系。结果单因素Cox回归分析显示,57个差异基因与甲状腺癌患者预后相关(P<0.05)。经LASSO回归分析和多因素Cox回归分析确定PHLDA2、GPR137B、PORCN、MAPK4和TSPYL2共5个基因参与模型构建。在试验组、验证组和整体组中,高风险组生存时间和生存率低于低风险组。高和低风险组差异基因GO功能富集分析发现,受体介导的内吞作用和先天免疫应答激活信号传导等显著富集;KEGG功能富集分析发现,mTOR信号通路、甲状腺激素信号通路和HIF1信号通路等显著富集。多因素Cox回归分析显示,甲状腺癌患者年龄和风险值评分与预后相关(P<0.05或P<0.01)。结论通过GEO和TCGA数据库构建的基于5个差异基因的甲状腺癌Cox比例风险回归模型,具有较高的准确性和可靠性,有助于临床医生判断甲状腺癌患者预后。Objective To explore the value of Cox proportional hazard regression model for differential expressed genes of thyroid cancer based on public databases in prediction of prognoses of patients.Methods The GSE138198 and GSE50901 datasets in the GEO database were batch-corrected,and altered genes in thyroid cancer tumor samples combined with the TCGA database were screened.Thyroid neoplasms patients in the TCGA database were randomly divided into test group and validation group by using R software,and a Cox proportional hazard regression model was established in test group.Survival conditions of patients with high or low risk in test group,verification group and overall group(overall TCGA data)were respectively analyzed by using Kaplan-meier survival curve.Receiver operating characteristic(ROC)curve was used to evaluate the accuracy of the model in predicting the survival probability of thyroid cancer patients.Differential expressed genes between the high and low-risk groups of thyroid cancer patients were analyzed,and GO and KEGG functional enrichment analysis were also performed.Univariate and multivariate Cox regression analyses were used to evaluate the relationships between clinicopathological characteristics with prognoses in patients with thyroid cancer.Results Univariate Cox regression analysis showed that 57 differential expressed genes were associated with prognoses of patients with thyroid neoplasms(P<0.05).LASSO regression analysis and multivariate Cox regression analysis showed that it was determined that PHLDA2,GPR137B,PORCN,MAPK4 and TSPYL2 were involved in the model construction.Among test,validation and overall groups,survival time and survival rate in high-risk group were significantly lower than those in low-risk group.GO functional enrichment analysis for differential expressed genes between high or low risk groups indicated that receptor-mediated endocytosis and innate immune response activation signal transduction were significantly enriched.KEGG function enrichment analysis revealed that mTOR s

关 键 词：甲状腺肿瘤 公共数据库 COX回归分析 预后 

分 类 号：R736.1[医药卫生—肿瘤]