Department of Neurosurgery, Central Hospital of Weihai, Weihai, China.
Department of Oncology, Central Hospital of Weihai, Weihai, China.
World Neurosurg. 2019 Jun;126:e765-e772. doi: 10.1016/j.wneu.2019.02.147. Epub 2019 Mar 7.
In the study, we aimed to identify key microRNAs (miRNAs) and clinical factors associated with survival time of lower-grade glioma (LGG) and develop an expression-based miRNA signature to provide survival risk prediction for patients with LGG.
We obtained miRNA expression profiles and clinical information of patients with LGG from The Cancer Genome Atlas dataset. All 591 miRNAs were modeled using random Forest Survival, Regression, and Classification to construct a random forest model for survival analysis, and feature selection was performed. We used univariate and multivariate Cox regression analysis to screen differentially expressed miRNAs and clinical factors related to overall survival of patients with LGG.
A total of 591 differentially expressed miRNAs were obtained between LGG and normal tissues. After univariate and multivariate Cox regression analysis, 2 predictive miRNAs (hsa-miR-10b-5p and hsa-miR-15b-5p) and 3 clinical factors (grade, age, and cancer status) were finally screened out to construct a 5-signature, based on which patients in the training dataset were divided into high-risk and low-risk groups. The competitive performance of the 5-signature was revealed by receiver operating characteristic curve analysis. The prognostic value of the 5-signature was successfully validated in the testing and validation dataset.
Our study demonstrated the promising potential of the novel 5-signature as an independent biomarker for survival prediction of patients with LGG.
本研究旨在鉴定与低级别胶质瘤（LGG）患者生存时间相关的关键微小 RNA（miRNA）和临床因素，并构建基于表达谱的 miRNA 特征，为 LGG 患者提供生存风险预测。
我们从癌症基因组图谱（TCGA）数据库中获取了 LGG 患者的 miRNA 表达谱和临床信息。采用随机森林生存、回归和分类模型对所有 591 个 miRNA 进行建模，构建用于生存分析的随机森林模型，并进行特征选择。我们采用单因素和多因素 Cox 回归分析筛选与 LGG 患者总生存期相关的差异表达 miRNA 和临床因素。
在 LGG 和正常组织之间获得了 591 个差异表达的 miRNA。经过单因素和多因素 Cox 回归分析，最终筛选出 2 个预测 miRNA（hsa-miR-10b-5p 和 hsa-miR-15b-5p）和 3 个临床因素（分级、年龄和癌症状态），构建了一个 5 个 miRNA 特征的模型，根据该模型，训练数据集中的患者被分为高危组和低危组。受试者工作特征曲线分析揭示了该 5 个 miRNA 特征的竞争性能。该 5 个 miRNA 特征的预后价值在测试和验证数据集中得到了成功验证。
本研究表明，该新型 5 个 miRNA 特征具有作为 LGG 患者生存预测的独立生物标志物的潜在应用价值。
