DNA methylation profiling to predict overall survival risk in gastric cancer: development and validation of a nomogram to optimize clinical management
DNA methylation profiling to predict overall survival risk in gastric cancer: development and validation of a nomogram to optimize clinical management
复制标题
DNA 甲基化分析预测胃癌总体生存风险：开发和验证列线图以优化临床管理
DOI：
10.7150/jca.44436
复制
发表时间：
2020-01-01
影响因子：
3.9
通讯作者：
Tao, Kaixiong
中科院分区：
文献类型：
--
作者：
Ma, Xianxiong;Chen, Hengyu;Tao, Kaixiong
据报道，DNA甲基化在胃癌的发生和发展中起着重要作用。我们的目的是结合胃癌的临床和甲基化因素，系统地建立一个个性化的生存风险预测模型。采用单因素Cox比例风险回归分析，根据早期和晚期胃癌组甲基化位点的差异表达，确定与预后相关的甲基化位点，然后应用最小绝对收缩和选择算子（LASSO） Cox回归模型筛选候选甲基化位点。随后，根据候选位点进行多变量Cox比例风险回归分析，确定预测特征。相对工作特征曲线（ROC）分析表明，11-甲基化特征对1、3、5年生存事件具有很高的预测效率。根据基于11-甲基化特征的风险评分进行分类的低危组患者的生存率明显高于高危组。结合甲基化风险评分和其他临床因素进行Cox回归分析，发现11-甲基化特征是独立的危险因素。在测试数据集中验证了风险评分的预测值。此外，构建了nomogram， ROC和标定图分析证明了nomogram的良好性能和临床应用价值。综上所述，该结果提示11-DNA甲基化特征可能是潜在的独立预后标志物，并可作为指导胃癌患者总生存期临床预测的重要工具。
DNA methylation has been reported to serve an important role in the carcinogenesis and development of gastric cancer. Our aim was to systematically develop an individualized prediction model of the survival risk combing clinical and methylation factors in gastric cancer. A univariate Cox proportional risk regression analysis was used to identify the prognosis-associated methylation sites based on the differentially expressed methylation sites between early and advanced gastric cancer group, then we applied least absolute shrinkage and selection operator (LASSO) Cox regression model to screen candidate methylation sites. Subsequently, multivariate Cox proportional risk regression analysis was conducted to identify predictive signature according to the candidate sites. Relative operating characteristic curve (ROC) analysis manifested that an 11-methylation signature exhibited great predictive efficiency for 1-, 3-, 5-year survival events. Patients in the low-risk group classified according to 11-methylation signature-based risk score yield significantly better survival than that in high-risk group. Moreover, Cox regression analysis combing methylation-based risk score and other clinical factors indicated that 11-methylation signature served as an independent risk factor. The predictive value of risk score was validated in the testing dataset. In addition, a nomogram was constructed and the ROC as well as calibration plots analysis demonstrated the good performance and clinical application of the nomogram. In conclusion, the result suggested the 11-DNA methylation signature may be potentially independent prognostic marker and functioned as a significant tool for guiding the clinical prediction of gastric cancer patients' overall survival.