基于甲基化位点的肺腺癌预后预测模型的探索与建立

  目的   利用公共组学数据库挖掘并构建基于DNA甲基化位点的肺腺癌预后模型，并对模型的预测效果进行初步评价。  方法  癌症基因组图谱(the cancer genome atlas, TCGA)及基因表达谱(gene expression omnibus, GEO)数据库用于相关分析和验证，最小绝对值收敛和选择算子法(least absolute shrinkage and selection operator, LASSO) Cox回归模型用于确定与肺腺癌预后风险具有关联的甲基化位点，Cox回归用于甲基化位点与预后风险的关联性分析，Harrell's C统计量用于评价模型的预测效果。  结果  甲基化位点cg02909790和cg19378330被纳入LASSO Cox最优模型。Cox比例风险回归模型结果显示，甲基化位点组合与肺腺癌预后的关联有统计学意义(HR=8.32, 95% CI: 2.41~28.69, P < 0.001)。结合甲基化位点和临床信息建立的肺腺癌预后预测模型预测能力较好，Harrell's C为0.81(95% CI: 0.78~0.83)。  结论  基于甲基化位点cg02909790和cg19378330的肺腺癌预后模型的预测效果良好，可能是潜在的个体化肿瘤分子标志物。

Exploration and establishment of the prediction model for lung adenocarcinoma prognosis based on DNA methylation sites

  Objective  To explore and build a prediction model for the prognosis of lung adenocarcinoma based on DNA methylation sites according to the public omics database and evaluate the prediction efficacy of the prediction model.  Methods  The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases were used for analysis and validation. The least Absolute Shrinkage and Selection Operator (LASSO) Cox regression model was performed to screen DNA methylation sites associated with the prognosis of lung adenocarcinoma. Multivariate Cox regression was used to evaluate the relationship between methylation signature and prognosis of lung adenocarcinoma. The prognostic prediction model of lung adenocarcinoma was estimated using Harrell's C statistics.  Results  Two methylation sites (cg02909790 and cg19378330) associated with the prognosis of lung adenocarcinoma were selected by LASSO Cox regression. Cox regression analysis showed a significant relationship between methylation signature and lung adenocarcinoma prognosis (HR=8.32, 95% CI: 2.41-28.69, P < 0.001). The C statistical value of the prognostic prediction model based on methylation signature for lung adenocarcinoma was 0.81 (95% CI: 0.78-0.83) according to Harrell's C statistical analysis.  Conclusions  The prediction model of lung adenocarcinoma based on cg02909790 and cg19378330 has a good prognostic prediction efficacy and may be potential personalized tumor molecular markers in the development and progression of lung adenocarcinoma.