食管鳞癌淋巴结转移关键基因鉴定及预后分析  

Prognostic Analysis and Identification of Key Genes Related to the Lymph Node Metastasis of Esophageal Squamous Cell Carcinoma

作　　者：李孟祥 程维刚 陈攀[2] 冯笑山 高社干[2] 齐义军[2] LI Meng-xiang;CHENG Wei-gang;CHEN Pan;FENG Xiao-shan;GAO She-gan;QI Yi-jun(School of Information Engineering,College of Clinical Medicine,Henan University of Science and Technology,Luoyang 471023,China;The First Affiliated Hospital,College of Clinical Medicine,Henan University of Science and Technology,Luoyang 471023,China)

机构地区：[1]河南科技大学信息工程学院,中国洛阳471023 [2]河南科技大学临床医学院第一附属医院,中国洛阳471023

出　　处：《湖南师范大学自然科学学报》2021年第6期92-100,共9页Journal of Natural Science of Hunan Normal University

基　　金：国家自然科学基金资助项目(U1604191)。

摘　　要：目的:应用食管鳞癌淋巴结转移的关键mRNA分子构建最佳淋巴结转移预测模型。方法:获取GSE53624和GSE53622中食管鳞癌转录组数据,XGBoost算法gain值确定食管鳞癌淋巴结转移关键mRNA分子,构建基于XGBoost,LR及SVM的淋巴结转移预测模型并评价其效能(AUC和K-S法),生存分析和GO富集分析关键mRNA分子的预后意义及生物学功能。结果:159个食管鳞癌差异表达mRNA分子中关键mRNA分子有18个;3种淋巴结转移预测模型中,XGB-18 mRNA模型的AUC和K-S值最优,并且其预测值是食管鳞癌预后的独立危险因素;18个mRNA分子中的4个低表达分子富集于上皮细胞角质化过程,在淋巴结转移阳性食管鳞癌中表达进一步降低。结论:XGB-18 mRNA模型诊断食管鳞癌淋巴结转移的效能最优,并且是食管鳞癌独立预后危险因素之一。Objective:The key genes associated with the lymph node metastasis in esophageal squamous cell carcinoma(ESCC)were used to construct the optimal lymph node metastasis prediction model.Methods:After downloading the mRNA transcriptome data of GSE53624 and GSE53622,the gain score from XGBoost algorithm was adopted to ascertain the key genes related to the lymph node metastasis in ESCC.These key genes were employed for building the lymph node metastasis prediction model based on the XGBoost,LR and SVM algorithms,followed by the evaluation of prediction accuracy using AUC and K-S.Furthermore,the survival analysis and Gene Ontology(GO)enrichment analysis were performed to assess the prognostic implication and biological functions.Results:Eighteen genes among 159 differentially expressed genes were ascertained to be the key genes related to the lymph node metastasis in ESCC.These 18 mRNAs as input features were used to construct the lymph node metastasis prediction model based on the XGB-/LR-/SVM-18 mRNA.The prediction accuracy of XGB-18 mRNA was superior to those of LR-18 mRNA and SVM-18 mRNA with greatest values of AUC and K-S.The multivariate survival analysis demonstrated that the XGB-18 mRNA score was an independent risk factor for the prognosis of ESCC.Four mRNAs among the 18 key mRNAs were enriched in keratinization during the epithelial differentiation and downregulated maximally in ESCC with lymph node metastasis.Conclusions:The XGB-18 mRNA classifier predicts well the lymph node metastasis of ESCC and is also an independent prognostic factor.

关 键 词：食管鳞癌 淋巴结转移 mRNA转录组 机器学习 XGBoost 

分 类 号：R735.1[医药卫生—肿瘤]