Table_4_Development and Validation of a Novel DNA Methylation-Driven Gene Based Molecular Classification and Predictive Model for Overall Survival and Immunotherapy Response in Patients With Glioblastoma: A Multiomic Analysis.DOCX

PurposeGlioblastoma (GBM) is the most common primary malignant tumor of the central nervous system, with a 5-year overall survival (OS) rate of only 5.6%. This study aimed to develop a novel DNA methylation-driven gene (MDG)-based molecular classification and risk model for individualized prognosis prediction for GBM patients.
MethodsThe DNA methylation profiles (458 samples) and gene expression profiles (376 samples) of patients were enrolled to identify MDGs using the MethylMix algorithm. Unsupervised consensus clustering was performed to develop the MDG-based molecular classification. By performing the univariate, least absolute shrinkage and selection operator (LASSO), and multivariate Cox regression analysis, a MDG-based prognostic model was developed and validated. Then, Bisulfite Amplicon Sequencing (BSAS) and quantitative real-time polymerase chain reaction (qPCR) were performed to verify the methylation and expressions of MDGs in GBM cell lines.
ResultsA total of 199 MDGs were identified, the expression patterns of which enabled TCGA and CGGA GBM patients to be divided into 2 clusters by unsupervised consensus clustering. Cluster 1 patients commonly exhibited a poor prognosis, were older in age, and were more sensitive to immunotherapies. Then, six MDGs (ANKRD10, BMP2, LOXL1, RPL39L, TMEM52, and VILL) were further selected to construct the prognostic risk score model, which was validated in the CGGA cohort. Kaplan-Meier survival analysis demonstrated that high-risk patients had significantly poorer OS than low-risk patients (logrank P = 3.338 × 10-6). Then, a prognostic nomogram was constructed and validated. Calibration plots, receiver operating characteristic curves, and decision curve analysis indicated excellent predictive performance for the nomogram in both the TCGA training and CGGA validation cohorts. Finally, in vitro BSAS and qPCR analysis validated that the expressions of the MDGs were negatively regulated by methylations of target genes, especially promoter region methylation.
ConclusionThe MDG-based prognostic model could serve as a promising prognostic indicator and potential therapeutic target to facilitate individualized survival prediction and better treatment options for GBM patients.

Purpose 胶质母细胞瘤（Glioblastoma, GBM）是中枢神经系统最常见的原发性恶性肿瘤，5年总生存（overall survival, OS）率仅为5.6%。本研究旨在构建一种基于DNA甲基化驱动基因（methylation-driven gene, MDG）的新型分子分型与风险模型，用于胶质母细胞瘤患者的个体化预后预测。
Methods 研究纳入患者的DNA甲基化谱（458例样本）与基因表达谱（376例样本），采用MethylMix算法鉴定甲基化驱动基因。通过无监督共识聚类构建基于甲基化驱动基因的分子分型。进一步通过单因素、最小绝对收缩和选择算子（least absolute shrinkage and selection operator, LASSO）及多因素Cox回归分析，构建并验证了基于甲基化驱动基因的预后模型。随后采用亚硫酸氢盐扩增子测序（Bisulfite Amplicon Sequencing, BSAS）与实时定量聚合酶链反应（quantitative real-time polymerase chain reaction, qPCR）验证胶质母细胞瘤细胞系中甲基化驱动基因的甲基化水平与表达情况。
Results 共鉴定得到199个甲基化驱动基因，通过无监督共识聚类可将TCGA与CGGA队列中的胶质母细胞瘤患者分为2个亚型。亚型1患者通常预后较差、年龄更大，且对免疫治疗更为敏感。随后筛选出6个甲基化驱动基因（ANKRD10、BMP2、LOXL1、RPL39L、TMEM52及VILL）构建预后风险评分模型，并在CGGA队列中完成验证。Kaplan-Meier生存分析显示，高风险组患者的总生存显著劣于低风险组（log-rank P = 3.338 × 10^-6）。此外，构建并验证了预后列线图，校准曲线、受试者工作特征（receiver operating characteristic, ROC）曲线及决策曲线分析均表明，该列线图在TCGA训练队列与CGGA验证队列中均表现出优异的预测性能。最后，体外BSAS与qPCR分析证实，甲基化驱动基因的表达受靶基因甲基化（尤其是启动子区域甲基化）的负向调控。
Conclusion 基于甲基化驱动基因的预后模型可作为颇具前景的预后指标与潜在治疗靶点，助力胶质母细胞瘤患者的个体化生存预测与优化治疗方案选择。