临床肿瘤学杂志

目的 构建结肠腺癌(COAD)转录因子(TFs)预后模型，为评估COAD患者的生存风险提供理论依据。方法 从癌症基因组图谱(TCGA)数据库下载COAD患者的临床信息及RNA-seq数据，筛选差异表达的TFs,并进行基因本体论(GO)和京都基因与基因组百科全书(KEGG)功能富集分析。采用单因素及多因素Cox比例风险回归模型建立具有预后价值的转录因子模型，通过Kaplan-Meier法分析高、低风险组患者的生存状况，利用受试者工作特征(ROC)曲线评估临床病理参数及风险模型对COAD患者预后的预测效能，并随机抽取50%样本进行内部验证。采用单因素与多因素Cox回归分析验证年龄、性别、TNM分期、T分期、N分期及风险评分是否为独立预后因素，并基于多因素Cox回归分析构建包含临床病理参数和风险评分的列线图模型。结果 共鉴定出131个差异表达的TFs。通过单因素及多因素Cox回归分析筛选出3个TFs(FOXS1、HSF4和LBX2),均为影响COAD患者预后的危险因素。多因素Cox回归分析显示，高风险组患者风险评分较高，低风险组评分较低，与内部测试组结果一致。高风险组患者的总生存时间(OS)显著短于低风险组(P<0.001),该结果在内部测试组中也得到验证(P<0.001)。该风险评分模型在1年、3年、5年和10年生存率的ROC曲线下面积(AUC)分别为0.719、0.773、0.773和0.694,内部验证组分别为0.714、0.783、0.783和0.668。在3年生存率的预测中，该模型与TNM分期、年龄、性别、T分期和N分期的AUC值分别为0.783、0.278、0.773、0.655、0.627和0.639,内部验证组分别为0.773、0.339、0.657、0.577、0.609和0.706。多因素Cox回归分析结果显示，风险评分是COAD的预后因素(HR=2.160,95%CI:1.370～3.406,P=0.001)。基于多因素Cox回归分析构建的列线图预测模型的一致性指数为0.878。结论 本研究构建并验证的TFs预后风险模型对区分COAD患者高、低风险亚组具有重要临床意义，可为个体化治疗提供参考。

Objective To obtain the transcription factor prognosis model of colon adenocarcinoma(COAD), providing a theoretical basis for evaluating the survival risk of patients with COAD. Methods the clinical information and RNA seq data of patients with COAD from the Cancer Genome Atlas(TCGA) database was downloaded to obtain differential transcription factor genes and perform GO and KEGG function enrichment analysis. Univariate and multivariate Cox proportional risk regression models were used to develop transcription factor models with prognostic value. Kaplan Meier method was used to analyze the survival status of high-risk and low-risk patients. ROC curve was used to assess the accuracy of risk scores and the sensitivity and specificity of clinical pathological parameters and risk models for predicting the prognosis of colon adenocarcinoma patients. 50% of the cases were randomly selected from the sample for internal validation. Independent prognostic factors such as patient's age, sex, TNM stage, T stage, N stage and risk score were verified by using univariate and multivariate Cox regression analysis, and multivariate Cox regression analysis was used to evaluate the nomograph model of clinical pathological parameters and risk score. Results A total of 131 differentially expressed TFs were identified. Through univariate and multivariate Cox regression analysis, three transcription factors(FOXS1, HSF4, and LBX2) were identified as risk factors affecting the prognosis of COAD patients. The linkage analysis of risk factors using multiple Cox regression showed that high-risk patients had higher risk scores, while low-risk patients had higher risk scores, and similar results were obtained in the internal testing group. The risk score of the model showed that the overall survival time of patients in the high-risk group was shorter than that in the low-risk group(P<0.001), which was also validated in the internal testing group(P<0.001). The area under the ROC curve(AUC value) of this risk scoring model was 0.719, 0.773, 0.773, and 0.694 at 1 year, 3 years, 5 years, and 10 years, respectively. The internal validation group had areas of 0.714, 0.783, 0.783, and 0.668, and the AUC value of clinical pathological features such as TNM stage, age, gender, T stage, and N stage was 0.783, 0.278, 0.773, 0.655, 0.627, and 0.639 at 3-year survival rates, while the internal validation group had areas of 0.773, 0.339, 0.657, 0.577, 0.609, and 0.706 at 3-year survival rates. The results of the multivariate Cox regression analysis showed that the risk score was an independent prognostic factor for COAD(HR=2.160, 95%CI: 1.370-3.406, P=0.001). The concordance index of the nomogram prediction model constructed based on the multivariate Cox regression analysis was 0.878. Conclusion The development and validation of transcription factor prognostic risk models are of great significance for identifying subgroups of patients with high or low risk of COAD.