基于RNA结合蛋白转录组数据构建食管鳞癌预后预测模型  

Construction of a prognostic prediction model of esophageal squamous cell carcinoma based on RNA binding protein transcriptome data

作　　者：胡力文 郭梓鑫 宋丛宽 汪育锦 王青雯 李炫飞 胡卫东[1] Hu Liwen;Guo Zixin;Song Congkuan;Wang Yujin;Wang Qingwen;Li Xuanfei;Hu Weidong(Thoracic Surgery,Zhongnan Hospital of Wuhan University Hubei Provincial Cancer Clinical Research Center&Hubei Key Laboratory of Biological Behavior of Tumor,Wuhan 430071,China;Gastrointestinal Surgery,Zhongnan Hospital of Wuhan University,Wuhan 430071,China)

机构地区：[1]武汉大学中南医院胸外科湖北省肿瘤临床研究中心肿瘤生物学行为湖北重点实验室,430071 [2]武汉大学中南医院胃肠外科,430071

出　　处：《中华实验外科杂志》2022年第11期2211-2214,共4页Chinese Journal of Experimental Surgery

基　　金：国家自然科学基金(82070302);武汉市腹膜癌临床医学研究中心资助项目(2015060911020462);武汉大学中南医院医学科技创新平台建设支撑项目(PTXM2021019);武汉大学中南医院转化医学与交叉学科研究联合基金(ZNJC202015)。

摘　　要：目的基于美国癌症基因组图谱(TCGA)数据库的RNA结合蛋白转录组数据构建食管鳞癌预后预测模型。方法从TCGA数据库下载80例食管鳞癌组织及11例正常组织的转录组数据及临床资料。应用倍数差异法筛选队列中差异表达基因,分析其KEGG通路、GO生物功能。使用survival程序包对食管癌数据RBPs进行单变量Cox分析,进一步筛选出重要的候选基因。利用LASSO-Cox回归分析,筛选出食管癌预后相关标志物,与食管癌预后相关的临床特征相结合,构建多变量Cox回归模型。利用Kaplan-Meier分析、ROC分析验证所构建的模型的预测性能。采用实时荧光定量PCR技术检测10例食管癌组织和10例癌旁组织中特征基因的表达,两组间比较采用t检验。结果在食管鳞癌组织和正常食管组织中进行差异基因分析筛选得到差异表达的38个RBPs,其中16个在肿瘤组织中上调、22个下调。经Cox回归分析后,3个RBPs(TRIT1、PIWIL4、ANG)被纳入模型的构建。根据中位风险评分生存分析,将患者分为低危组和高危组,结果表明,无论是在TCGA训练集还是测试集,高危亚组患者总体生存时间低于低危亚组患者,其结果具有统计学意义(P<0.05)。训练集中模型预测3年生存率的ROC曲线下面积为0.749,在测试集则为0.738,提示模型准确率较高。在独立预后分析中,单因素Cox、多变量Cox结果表明,风险评分与患者的OS相关(P<0.05),说明该预后模型可以作为独立的预后因子预测食管鳞癌患者的生存。最后通过qRT-PCR验证食管鳞癌患者的癌组织和癌旁组织中特征基因的表达,结果显示,食管鳞癌患者肿瘤组织ANG、TRIT1基因表达显著高于癌旁组织(13.661±10.354比2.500±2.333、24.551±22.518比1.752±2.057),差异有统计学意义(t=2.782、2.470,P<0.05),PIWIL4基因表达显著低于癌旁组织(0.229±0.341比1.511±1.180),差异有统计学意义(t=-2.763,P<0.05)。结论本研究构建的基于3个RBPs的�Objective To construct a prognostic prediction model for esophageal squamous cell carcinoma(ESCC)based on RNA binding protein(RBP)transcriptome data from the American cancer genome atlas(TCGA)database.Methods The transcriptome data and clinical data of 80 esophageal squamous carcinoma tissues and 11 normal tissues were downloaded from TCGA database.Multiple difference method was used to screen and find differentially expressed genes in the cohort,and analyze their KEGG pathway and GO biological function.The survival package was used to perform univariate Cox analysis on RBPs of esophageal cancer data to further screen out important candidate genes.Using LASSO Cox regression analysis,we screened the prognostic related markers of esophageal cancer,combined with the clinical characteristics related to the prognosis of esophageal cancer,and built a multivariate Cox regression model.Kaplan Meier analysis and ROC analysis were used to verify the prediction performance of the model.Finally,We used real-time fluorescent quantitative PCR to detect the expression of characteristic genes in 10 esophageal cancer tissues and 10 adjacent tissues,T-test was used for comparison between the two groups.Results 38 differentially expressed RBPs were screened by differential gene analysis in esophageal squamous cell carcinoma tissues and normal esophageal tissues,of which 16 were up-regulated and 22 were down-regulated in tumor tissues.After the Cox regression analysis,the three RBPs(TRIT1,PIWIL4,and ANG)were included in the model construction.According to the median risk score survival analysis,the patients were divided into low risk group and high risk group.The results showed that no matter in TCGA training set or test set,the overall survival time of patients in high risk subgroup was lower than that in low risk subgroup(P<0.05).The area under the ROC curve of the training set model to predict the 3-year survival rate was 0.749,and that in the test set was 0.738,indicating that the accuracy of the model was higher.In independent

关 键 词：食管癌 RNA结合蛋白 预后风险评分模型 美国癌症基因组图谱 

分 类 号：R735.1[医药卫生—肿瘤]