基于TCGA构建口腔鳞癌自噬相关基因风险预测模型及验证

Construction and validation of risk prediction model for autophagy related genes in oral squamous cell carcinoma based on TCGA

目的 利用癌症基因组图谱(TCGA)数据库构建口腔鳞状细胞癌(OSCC)的自噬相关基因(ARG)风险预测模型。方法 基于TCGA数据库筛选OSCC中差异表达ARG。依次行单因素Cox分析、Lasso回归分析和多因素Cox分析构建ARG风险预测模型,并根据模型风险评分(RS)的中位数将患者分为高、低风险组,组间生存差异通过Kaplan-Meier进行评估。单因素和多因素Cox分析检验模型在预测预后中的作用,受试者工作特征曲线(ROC)评估模型的准确性,最后分析模型与临床病理特征的相关性。Western-blot验证模型中基因BAK1在OSCC中的表达。结果 OSCC中共筛选出37个差异表达ARG(FDR<0.05,|logFC|>1)。行单因素Cox分析、Lasso回归分析及多因素Cox分析筛选出FADD、NKX2-3、BAK1作为预后相关差异表达ARG,用于风险预测模型的构建,即RS=(1.558 5×FADD)+(-0.557 8×NKX2-3)+(1.547 1×BAK1)。Kaplan-Meier生存分析表明,高风险组的5年生存率低于低风险组(P<0.001)。单因素和多因素Cox分析表明,预测模型有效且独立于其他临床因素(P<0.001)。ROC曲线结果显示,该模型诊断OSCC预后的曲线下面积为0.614。在较高T分期和较大年龄患者中可发现模型的RS增高(P<0.05)。Western-blot结果表明,在OSCC癌组织中BAK1蛋白表达高于癌旁组织(t=3.197,P=0.033)。结论 基于FADD、NKX2-3、BAK1等3个ARG构建的风险预测模型可作为OSCC患者预测预后的生物标志物,从而有助于对不同风险的患者进行个体化诊疗。

Objective To construct an autophagy related gene(ARG) risk prediction model for oral squamous cell carcinoma(OSCC) using the cancer genome atlas(TCGA) database. Methods Screening ARG differentially expressed in OSCC based on TCGA database. The single factor Cox analysis, Lasso regression analysis and multi factor Cox analysis were conducted to construct the ARG risk prediction model. The ARG risk prediction model was divided into high and low risk groups according to the median of the model risk score(RS). The survival difference between groups was evaluated by Kaplan Meier. Univariate and multivariate Cox analysis examined the role of the model in predicting prognosis, the accuracy of the model was evaluated by the receiver operating characteristic curve(ROC), and finally the correlation between the model and clinical pathological characteristics was analyzed. Western blot was used to verify the expression of gene BAK1 in OSCC. Results A total of 37 differentially expressed ARGs were screened by OSCC(FDR<0.05, | logFC |>1). Single factor Cox analysis, Lasso regression analysis and multivariate Cox analysis screened FADD, NKX2-3, and BAK1 as prognostic related differences to express ARG for the construction of risk prediction model, that is, RS=(1.558 5×FADD)+(-0.557 8 × NKX2-3)+(1.547 1 × BAK1). Kaplan-Meier survival analysis showed that the 5-year survival rate of high-risk group was lower than that of low-risk group(P<0.001). Cox analysis of single factor and multiple factors showed that the predictive model was effective and independent of other clinical factors(P<0.001). The ROC curve results showed that the area under the curve for diagnosing the prognosis of OSCC was 0.614. The RS of the model was increased in higher T stage and older age(P<0.05). Western blot results showed that the expression of BAK1 protein in OSCC cancer tissues was higher than that in adjacent tissues(t=3.197, P=0.033). Conclusion The risk prediction model based on FADD, NKX2-3, BAK1 and other three ARGs can be used as a biomarker to predict the prognosis of OSCC patients, which is helpful for individualized diagnosis and treatment of patients with different risks.