基于药物代谢相关基因构建肺癌预后风险预测模型

Construction of lung cancer prognostic risk prediction model based on drug metabolism-related genes

目的 研究基于药物代谢相关基因构建的风险模型预测肺癌患者预后的应用价值。方法 基于Pharma ADME Consortium鉴定的298个药代动力学(ADME)相关基因,通过GEO芯片GSE7670、GSE32863获得ADME相关差异基因(ADME-DEG);将ADME-DEGs进行基因本体(GO)功能富集及蛋白质-蛋白质相互作用(PPI)网络分析;利用一致性聚类、PCA主成分分析,基于18个ADME-DEGs将肺癌患者分为两组:Cluster1(n=315)、Cluster2(n=178);通过LASSO算法获得由10个ADME-DEGs组成的风险模型;通过Kaplan-Meier生存分析、ROC分析及多因素回归分析,基于风险得分建立一个肺癌患者预测Nomogram图,分析风险得分对肺癌患者生存期的预测能力。结果 成功建立由10个ADME-DEGs(SLC22A18、AOX1、CAT、ADH1B、SULF1、PPARG、CYP4B1、FMO2、GPX3、ABCA4)构建的肺癌风险预测模型。Kaplan-Meier生存分析显示,风险得分较高的患者总生存期(OS)较差;ROC分析显示,风险得分能较好地预测肺癌患者的生存期,年龄、性别、风险得分均与肺癌患者OS显著相关;Nomogram图显示,风险得分对于肺癌患者10年内的OS具有良好的预测能力(C-index:0.688)。结论 基于10个ADME-DEGs构建的风险模型具有预测肺癌患者用药后预后情况的潜在作用,为改善患者预后提供了新方法。

Objective To study the application value of a risk model based on drug metabolism-related genes in pre-dicting the prognosis of lung cancer patients.Methods Based on the 298 pharmacokinetics(ADME)-related genes identi-fied by Pharma ADME Consortium,ADME-related differentially-expressed genes(ADME-DEGs)were obtained through GEO chips GSE7670 and GSE32863.ADME-DEGs were analyzed with gene ontology(GO)functional enrichment and pro-tein-protein interaction(PPI)network.By using consistent clustering and PCA principal component analysis,it was found that 18 ADME-DEGs could significantly divide lung cancer patients into two categories:Cluster1(n=315)and Cluster2(n=178).Through the LASSO algorithm,a risk model consisting of 10 ADME-DEGs was finally obtained.Through Kaplan-Meier survival analysis,ROC analysis and multivariate regression analysis,a Nomogram for predicting lung cancer patients was established based on the risk scores,and the predictive ability of the risk scores for the survival time of lung cancer pa-tients was analyzed.Results A lung cancer risk prediction model was successfully established,constructed by 10 ADME-DEGs(SLC22A18,AOX1,CAT,ADH1B,SULF1,PPARG,CYP4B1,FMO2,GPX3,ABCA4).Kaplan-Meier survival anal-ysis showed that patients with higher risk scores had significantly poorer overall survival.ROC analysis showed that risk scores could better predict the survival of lung cancer patients.Age,gender,and risk scores were significantly correlated with the overall survival of lung cancer patients.The Nomogram plot showed that the risk scores had a good predictive abili-ty for the 10-year overall survival of lung cancer patients(C-index:0.688).Conclusion The risk model constructed on the base of 10 ADME-DEGs had a potential role in predicting the prognosis of lung cancer patients after medication,and pro-vided a new method for improving the prognosis of patients.