厄洛替尼耐药预测模型构建的研究

Study on the construction of an erlotinib resistance prediction model

目的 用R语言从数据库中挖掘数据,用LASSO法构建厄洛替尼耐药预测模型,为临床用药提供指导。方法 从癌症细胞系百科全书和癌症依赖关系图下载肿瘤细胞系的详细信息,包括各抗癌药物半数抑制浓度(IC_(50))值、细胞系来源、mRNA表达,根据厄洛替尼IC_(50)值的大小将细胞系分为耐药组和敏感组,通过R语言包edgeR分析2组间mRNA表达差异,从癌症基因组图谱数据库中下载用厄洛替尼进行治疗的患者的临床数据,以LASSO法筛选出与厄洛替尼耐药相关的mRNA,构建厄洛替尼耐药预测模型,并分析与厄洛替尼治疗患者的预后关系。结果 厄洛替尼耐药组和敏感组的mRNA表达有显著性差异,按P值由小到大选取100个对应基因进行LASSO分析,得到6个基因与厄洛替尼耐药相关,分别是含脱氢酶结构域2(ABHD2)、ERBB受体反馈抑制因子1(ERRFI1)、具有序列相似性的家族107成员B(FAM107B)、多同源同系物2(PHC2)、核糖体蛋白S6激酶A1(RPS6KA1)和微管蛋白2AⅡa类(TUBB2A),再通过交叉验证后使用广义线性模型功能建立耐药预测模型。TUBB2A高表达不仅与厄洛替尼耐药性相关,还与厄洛替尼治疗非小细胞肺癌不良预后相关。结论 ABHD2、ERRFI1、FAM107B、PHC2、RPS6KA1和TUBB2A等6个基因与厄洛替尼耐药相关,其中TUBB2A高表达不仅与厄洛替尼耐药相关,还与接受厄洛替尼化疗患者的不良预后相关。

Objective To construct an erlotinib resistance prediction model with the LASSO method by using the R language to mine data from databases,providing guidance for clinic.Methods Detailed information of tumor cell lines,including 50%inhibiting concentration(IC_(50))values of various anticancer drugs,cell lines,and mRNA expression,was downloaded from Cancer Cell Line Encyclopedia and Dependency map.The cell lines were divided into resistant and sensitive groups of erlotinib by IC_(50) values.The mRNA expression differences between the two groups were analyzed using the edgeR package of R language.The erlotinib-resistant associated mRNAs were selected and used to construct an erlotinib resistance prediction model by LASSO.The clinical data of patients treated with erlotinib were obtained from the the Cancer Genome Atlas database.The model was analyzed to determine its relationship with the prognosis of patients treated with erlotinib.Results Significant difference was found in the mRNA expression between erlotinib-resistant and sensitive groups.Out of 100 corresponding genes were selected in ascending order of P-values,6 genes including abhydrolase domain containing 2(ABHD2),ERBB receptor feedback inhibitor 1 Gene(ERRFI1),family with sequence similarity 107 member B Gene(FAM107B),polyhomeotic homolog 2 Gene(PHC2),ribosomal protein S6 kinase A1 Gene(RPS6KA1)and tubulin beta 2A class II a Gene(TUBB2A)were identified as associated with erlotinib resistance by LASSO analysis.After cross-validation,a resistance prediction model was established using the general linear module function.Interestingly,high expression of TUBB2A was not only associated with erlotinib resistance but also with poor prognosis of non-small cell lung cancer patients treated with erlotinib.Conclusion The 6 genes(ABHD2,ERRFI1,FAM107B,PHC2,RPS6KA1 and TUBB2A)were found to be associated with erlotinib resistance,of which TUBB2A can predict prognosis of patients receiving erlotinib chemotherapy.