1.中南大学湘雅医院，甲状腺外科，湖南长沙 410013;2.中南大学湘雅医院，国际医疗部，湖南长沙 410013;3.中南大学湘雅医院，中南大学湘雅医学院，湖南长沙 410013
彭婀敏， Email: firstname.lastname@example.org
1.Department of Thyroid Surgery, Central South University, Changsha 410013, China;2.International Medical Department, Xiangya Hospital, Central South University, Changsha 410008, China, Central South University, Changsha 410013, China;3.Xiangya School of Medicine, Central South University, Changsha 410013, China
背景与目的 甲状腺癌发病率逐年升高，但其发病机制仍不清楚，阐述甲状腺癌的发病机制对改善甲状腺癌患者的预后至关重要。研究表明，m6A甲基化调控因子高度参与癌症发生发展，具有良好的潜在预后价值。因此，本研究通过生物信息学方法分析甲状腺癌中m6A甲基化调控因子的表达并构建基于m6A甲基化调控因子的甲状腺癌预后模型。方法 从TCGA数据库下载甲状腺癌m6A甲基化调控因子的表达数据和相应的临床病理资料，通过Wilcoxon检验分析20个m6A甲基化调控因子在肿瘤和正常组织的差异表达；用一致性聚类分析将甲状腺癌患者分为两个聚类，比较两个聚类患者临床病理因素和总体生存率的差异；Lasso Cox回归分析构建风险预测模型并用ROC曲线下面积（AUC）评估模型的预测能力。结果 19个m6A甲基化调控因子在甲状腺癌和正常组织表达具有统计学差异（均P<0.05），其中HNRNPC、IGF2BP2、FMR1在甲状腺癌组织中明显高表达，而其余表达下调。聚类分析示，cluster 1生存期低于cluster 2（P<0.05），颈淋巴结转移发生率明显高于cluster 2（P<0.01）。基于Lasso Cox回归分析筛选的4个基因（IGF2BP2、RBM15、YTHDF1、YTHDF3）构建风险评估模型，相比低风险组患者，高风险患者生存期明显缩短（P=0.007）；ROC曲线示，该模型可以预测甲状腺癌患者的预后（AUC=0.731）。结论 m6A甲基化调控因子在甲状腺癌中存在差异表达，基于甲状腺癌中关键m6A甲基化调控因子所构建的预后风险模型具有较好的预测能力，可为临床决策提供一定的依据。
Background and Aims The incidence of thyroid carcinoma has been steadily increasing and the molecular mechanism underlying the tumorigenesis is still unclear. Thus, exploring the underlying mechanism is of great importance for improving the prognosis of PTC patients. Studies demonstrated that m6A methylation regulators are widely involved in the occurrence and development of cancers and have superior prognostic value. Therefore, this study was conducted to investigate the expressions of m6A methylation regulators in thyroid cancer and construct a prognostic risk model for thyroid cancer based on m6A methylation regulators using bioinformatics approaches.Methods The gene expression profiles of m6A RNA methylation regulators and the corresponding clinical information were downloaded from The Cancer Genome Atlas (TCGA). The differential expressions of 20 m6A methylation regulators between tumor and normal tissues were analyzed by Wilcoxon test. The thyroid cancer patients were divided into two clusters by consensus clustering, and the differences in clinicopathologic factors and overall survival rate between the two clusters of patients were compared. Subsequently, the risk model was constructed by Lasso Cox regression analysis and its predictive value was evaluated by the area under the ROC (AUC).Results There were 19 m6A methylation regulators that showed significantly different expressions between thyroid cancer and normal tissues (all P<0.05), in which, the HNRNPC, IGF2BP2, FMR1 were remarkably upregulated, while the remaining were down-regulated in thyroid cancer tissue. The clustering analysis showed that the overall survival of cluster 1 was poorer than that of cluster 2 (P<0.05), while the incidence of cervical lymph node metastasis was significantly higher than that of cluster 2 (P<0.01). A prognostic risk model was constructed based on 4 genes (IGF2BP2, RBM15, YTHDF1 and YTHDF3) that were screened by Lasso Cox regression analysis, in which, the high-risk patients had a worse prognosis than that of low-risk patients (P=0.007). The ROC analysis indicated a reliable prediction performance of the model for patients with thyroid cancer (AUC=0.731).Conclusion There are differential expressions in m6A methylation regulators in thyroid cancer, and the constructed prognostic risk model based on the hub m6A methylation regulators in thyroid cancer has better predictive ability, which will provide certain recommendations for clinical decision-making.