基于癌症基因组图谱数据库构建膀胱癌铜死亡相关长链非编码RNA预后模型

Construction of a prognostic model of bladder cancer using cuproptosis-associated long non-coding RNA based on The Cancer Genome Atlas database

目的构建膀胱癌铜死亡相关长链非编码RNA(lncRNA)预后风险模型并检验其预测效能。方法从癌症基因组图谱(TCGA)数据库中下载膀胱癌的RNA表达测序数据和对应样本的临床数据。从已发表的文献中获得17个铜死亡相关关键基因,基于TCGA数据库lncRNA数据,通过相关性分析筛选出与铜死亡关键基因相关的lncRNA,再应用Cox回归和Lasso回归筛选与TCGA数据库膀胱癌患者预后相关的铜死亡lncRNA。将从TCGA数据库筛选的临床信息完整的403例膀胱癌患者分为训练集(203例)和测试集(200例),并依据训练集样本和上述筛选的关键独立预后相关铜死亡lncRNA,构建预后风险预测模型。按照风险评分的中位值,分别将TCGA数据库筛选的膀胱癌全部数据集、测试集和训练集患者分为高风险组和低风险组,应用R语言survival包比较各数据集两组总生存差异。采用主成分分析、受试者工作特征(ROC)曲线验证模型预测效能。采用单因素和多因素Cox回归分析403例膀胱癌患者总生存影响因素,采用ROC曲线分析各因素预测膀胱癌预后的效能。结果经过筛选,共纳入4个具有独立预后意义的铜死亡相关lncRNA,分别为AC104564.3、LINC00649、AL136084.3、AL136295.2,以此构建的预后模型为:风险评分=-0.71342×AC104564.3-0.74494×LINC00649+0.41093×AL136084.3-0.73689×AL136295.2。生存分析显示,全部数据集、测试集和训练集中高风险组的总生存均较低风险组差(均P<0.05),提示高风险评分预示着较差的预后。ROC曲线分析显示,应用预后风险预测模型评分预测TCGA数据库全部403例患者1、3、5年总生存的曲线下面积分别0.665、0.629、0.692。多因素Cox回归分析显示,年龄(≥65岁比<65岁:OR=1.027,95%CI 1.011~1.044,P<0.001)、分期(Ⅳ期比Ⅲ期比Ⅱ期比Ⅰ期比未知分期:OR=1.593,95%CI 1.308~1.939,P<0.001)、风险评分(高比低:OR=1.258,95%CI 1.126~1.406,P<0.001)为患者总生存的独立影响因素。ROC曲线分析显示,年龄、分期、风险评分预测患者5年总生存的曲线下面积分别为0.614、0.685、0.692,提示风险预测模型具有更好的预测效能。结论成功建立了基于4个铜死亡相关lncRNA的膀胱癌患者预后风险预测模型,内部验证该模型有较高的预后预测效能。

Objective To construct a prognostic risk model of bladder cancer using cuproptosis-associated long non-coding RNA(lncRNA)and test its predictive efficacy.Methods RNA expression sequencing data and clinical data of corresponding samples were downloaded from The Cancer Gene Atlas(TCGA)database.The 17 key genes associated with cuproptosis was obtained from the published literature,and then lncRNA of the key genes associated with cuproptosis was screened by correlation analysis based on the lncRNA data from TCGA database.The cuproptosis lncRNA associated with the prognosis of bladder cancer patients were screened by using Cox regression and Lasso regression.A total of 403 bladder cancer patients with complete clinical information screened from TCGA database were divided into a training set(203 cases)and a test set(200 cases),and the prognostic risk prediction model was constructed based on the samples in the training set and the above key independent prognosis-related cuproptosis lncRNA.According to the median value of the risk score,patients in all the datasets,the test set and the training set of bladder cancer screened from TCGA database were divided into high-risk group and low-risk group,and R language survival package was applied to compare the differences in overall survival between the two groups in each dataset.The predictive effect of the model was verified using principal component analysis(PCA)and receiver operating characteristic(ROC)curve.Univariate and multivariate Cox regression analysis were used to analyze the factors affecting overall survival of 403 bladder cancer patients,and ROC curve was used to analyze the efficacy of each factor for predicting the prognosis of bladder cancer.Results After screening,a total of 4 cuproptosis lncRNA with independent prognostic significance were included(AC104564.3,LINC00649,AL136084.3 and AL136295.2),and the prognostic model constructed based on these 4 lncRNA was as follows:risk score=-0.71342×AC104564.3-0.74494×LINC00649+0.41093×AL136084.3-0.73689×AL136295.2.Survival analysis showed that the overall survival of the high-risk group in all datasets,the test set and the training set was poorer than that of the low-risk group(all P<0.05),suggesting that a high risk score predicted poor prognosis.ROC curve analysis showed that the areas under the curve of applying the risk prediction model to predict 1-,3-and 5-year overall survival of all 403 patients in TCGA database were 0.665,0.629 and 0.692.Multivariate Cox regression analysis showed that age(≥65 years old vs.<65 years old:OR=1.027,95%CI 1.011-1.044,P<0.001),stage(stageⅣvs.stageⅢvs.stageⅡvs.stageⅠvs.unknown stage:OR=1.593,95%CI 1.308-1.939,P<0.001)and risk score(high vs.low:OR=1.258,95%CI 1.126-1.406,P<0.001)were the independent influencing factors of patients'overall survival.ROC curve analysis showed that the areas under the curve of age,stage and risk score for predicting the patients'5-year overall survival were 0.614,0.685 and 0.692,suggesting that the risk prediction model had better predictive efficacy.Conclusions A prognosis risk prediction model for bladder cancer patients is constructed based on 4 lncRNA associated with cuproptosis,and the model is internally validated to have a high predictive efficacy.