源于TARGET、ArrayExpress和GEO数据库构建神经母细胞瘤中铜死亡相关预后模型

Construction and validation of acuproptosis-related prognostic model in neuroblastoma based on TARGET,ArrayExpress and GEO databases

目的基于铜死亡相关基因(cuproptosis-related genes,CRGs)构建神经母细胞瘤(neuroblastoma,NB)预后模型,以改善神经母细胞瘤患者的个体化管理。方法从公开的数据库中收集具有相关完整临床信息的铜死亡基因表达数据。获得来自基因表达综合数据库(gene expression omnibus,GEO)的GSE49711队列、来自开放有效治疗方法的治疗性研究数据库(therapeutically applicable research to generate effective treatments,TARGET)的TARGET-NB队列和来自功能基因组学公共存储库ArrayExpress数据库的E-MTAB-8248队列。排除随访信息缺失的患者,最终纳入968例进行分析。选择GSE49711队列作为“训练集”构建预后模型,其余两个数据集作为“验证集”。使用Log-rank检验筛选有预后意义的变量,再使用LASSO-Cox回归分析建立最佳多基因预后模型。采用ROC曲线、列线图、校准曲线和DCA曲线评估预后模型的准确性。使用RT-qPCR验证风险基因在神经母细胞系中的表达水平,并选择关键风险基因PDHA1进行功能分析。结果通过训练集首次建立了预后模型[风险评分公式为(1.573)×PDHA1+(-0.561)×GLS+(0.320)×LIAS+(0.088)×MTF1+(0.301)×PDHB]。根据风险评分中位值将患者分为高风险组和低风险组。生存分析显示高风险组中的NB患者的生存率明显低于低风险组(P<0.001)。时间依赖性ROC曲线预测3、5、7年生存率的曲线下面积(area under curve,AUC)分别为0.80、0.80和0.81。生存分析表明,在TARGET-NB和E-MTAB-8248队列中,与低风险组比较,高风险组患者的预后患者更差(P=0.011,P=0.0087)。列线图的校准曲线和DCA曲线(C指数为0.736)显示列线图具有良好的临床价值。RT-qPCR和功能丧失实验验证了风险模型中基因的表达水平和关键基因PDHA1的功能。结论基于铜死亡相关基因构建了预测神经母细胞瘤患者生存率的预后模型,并在两个外部数据集中验证了该模型的准确性。

Objective To construct a prognostic model for neuroblastoma(NB)based on cuproptosis-related genes(CRGs)and improve individualized management of neuroblastoma patients.Methods CRGs expression data with complete clinical information were collected from publicly available databases.A total of 3 independent datasets were obtained,including the GSE49711 cohort from Gene Expression Omnibus(GEO)database,the TARGET-NB cohort from Therapeutically Applicable Research to Generate Effective Treatments(TARGET)database,and the E-MTAB-8248 cohort from ArrayExpress database.968 patients were finally included for follow-up data analysis after excluding patients with incomplete follow-up information.The GSE49711 cohort was selected as the training set to construct the prognositic model,and the other two data sets were used as the validation set to verify the accuracy.Log-rank tests were used to screen prognostically significant variables,and the best multi-gene prognostic model was contructed using LASSO-Cox regression.ROC curves,column plots,calibration curves,and DCA curves were used to assess the accuracy of the prognostic models.RT-qPCR was used to validate the expression levels of risk genes in NB cell lines,and the key risk gene PDHA1 was selected for functional analysis.Results A prognostic model was first constructed in the training cohort with a risk score formula of(1.573)×PDHA1+(-0.561)×GLS+(0.320)×LIAS+(0.088)×MTF1+(0.301)×PDHB.According to the risk scores calculated based on the formula,patients were classified into the high-and low-risk subgroups based on the median values.Survival analysis showed that NB patients in the high-risk subgroup had a significantly lower survival rate than that in the low-risk subgroup(P<0.001).The time-dependent ROC curve predicted the area under curve(AUC)of 3-year,5-year and 7-year survival rate was 0.80,0.80,and 0.81,respectively.Survival analysis showed that in the TARGET-NB and E-MTAB-8248 cohorts,the high-risk subgroup was associated with a worse prognosis compared with the low-risk subgroup(P=0.011,P=0.0087).The calibration curve and DCA curve(C-index:0.736)of nomagram showed the good clinical value of nomagram.The expression levels of genes in the risk model and the function of the key gene PDHA1 were verified by RT-qPCR and loss-of-function experiments.Conclusion A prognostic model to predict the survival rate of patients with neuroblastoma is constructed based on cuproptosis-related genes and validated in two external datasets.