Wang Sheng, Xu Xia
Zhejiang Jinhua Guangfu Hospital, Jinhua, China.
Department of General Medicine, Xiangya Hospital, Central South University, Changsha, China.
Front Oncol. 2021 Mar 30;11:564960. doi: 10.3389/fonc.2021.564960. eCollection 2021.
Glioblastoma (GBM) is the frequently occurring and most aggressive form of brain tumors. In the study, we constructed an immune-related gene pairs (IRGPs) signature to predict overall survival (OS) in patients with GBM. We established IRGPs with immune-related gene (IRG) matrix from The Cancer Genome Atlas (TCGA) database (Training cohort). After screened by the univariate regression analysis and least absolute shrinkage and selection operator (LASSO) regression analysis, IRGPs were subjected to the multivariable Cox regression to develop an IRGP signature. Then, the predicting accuracy of the signature was assessed with the area under the receiver operating characteristic curve (AUC) and validated the result using the Chinese Glioma Genome Atlas (CGGA) database (Validation cohorts 1 and 2). A 10-IRGP signature was established for predicting the OS of patients with GBM. The AUC for predicting 1-, 3-, and 5-year OS in Training cohort was 0.801, 0.901, and 0.964, respectively, in line with the AUC of Validation cohorts 1 and 2 [Validation cohort 1 (1 year: 0.763; 3 years: 0.786; and 5 years: 0.884); Validation cohort 2 (1 year: 0.745; 3 years: 0.989; and 5 years: 0.987)]. Moreover, survival analysis in three cohorts suggested that patients with low-risk GBM had better clinical outcomes than patients with high-risk GBM. The univariate and multivariable Cox regression demonstrated that the IRGPs signature was an independent prognostic factor. We developed a novel IRGPs signature for predicting OS in patients with GBM.
胶质母细胞瘤（GBM）是最常见且侵袭性最强的脑肿瘤形式。在本研究中，我们构建了一个免疫相关基因对（IRGPs）特征模型，用于预测GBM患者的总生存期（OS）。我们从癌症基因组图谱（TCGA）数据库（训练队列）的免疫相关基因（IRG）矩阵中建立了IRGPs。经过单变量回归分析和最小绝对收缩和选择算子（LASSO）回归分析筛选后，将IRGPs进行多变量Cox回归以建立IRGP特征模型。然后，使用受试者操作特征曲线（AUC）下的面积评估该特征模型的预测准确性，并使用中国胶质瘤基因组图谱（CGGA）数据库（验证队列1和2）验证结果。建立了一个包含10个IRGP的特征模型，用于预测GBM患者的OS。训练队列中预测1年、3年和5年OS的AUC分别为0.801、0.901和0.964，与验证队列1和2的AUC一致[验证队列1（1年：0.763；3年：0.786；5年：0.884）；验证队列2（1年：0.745；3年：0.989；5年：0.987）]。此外，三个队列的生存分析表明，低风险GBM患者的临床结局优于高风险GBM患者。单变量和多变量Cox回归表明，IRGPs特征模型是一个独立的预后因素。我们开发了一种用于预测GBM患者OS的新型IRGPs特征模型。
