Department of Neurosurgery, West China Hospital Sichuan University, No. 37 Guoxue Lane, Chengdu, Sichuan, China.
Department of Neurosurgery, Zhejiang Provincial People's Hospital, No. 158 Shangtang Road, Hangzhou, Zhejiang, China.
J Transl Med. 2023 Aug 8;21(1):533. doi: 10.1186/s12967-023-04382-2.
Accurately predicting the outcome of isocitrate dehydrogenase (IDH) wild-type glioblastoma (GBM) remains hitherto challenging. This study aims to Construct and Validate a Robust Prognostic Model for IDH wild-type GBM (COVPRIG) for the prediction of overall survival using a novel metric, gene-gene (G × G) interaction, and explore molecular and cellular underpinnings.
Univariate and multivariate Cox regression of four independent trans-ethnic cohorts containing a total of 800 samples. Prediction efficacy was comprehensively evaluated and compared with previous models by a systematic literature review. The molecular underpinnings of COVPRIG were elucidated by integrated analysis of bulk-tumor and single-cell based datasets.
Using a Cox-ph model-based method, six of the 93,961 G × G interactions were screened to form an optimal combination which, together with age, comprised the COVPRIG model. COVPRIG was designed for RNA-seq and microarray, respectively, and effectively identified patients at high risk of mortality. The predictive performance of COVPRIG was satisfactory, with area under the curve (AUC) ranging from 0.56 (CGGA693, RNA-seq, 6-month survival) to 0.79 (TCGA RNAseq, 18-month survival), which can be further validated by decision curves. Nomograms were constructed for individual risk prediction for RNA-seq and microarray-based cohorts, respectively. Besides, the prognostic significance of COVPRIG was also validated in GBM including the IDH mutant samples. Notably, COVPRIG was comprehensively evaluated and externally validated, and a systemic review disclosed that COVPRIG outperformed current validated models with an integrated discrimination improvement (IDI) of 6-16%. Moreover, integrative bioinformatics analysis predicted an essential role of METTL1 neural-progenitor-like (NPC-like) malignant cell in driving unfavorable outcome.
This study provided a powerful tool for the outcome prediction for IDH wild-type GBM, and preliminary molecular underpinnings for future research.
准确预测异柠檬酸脱氢酶（IDH）野生型胶质母细胞瘤（GBM）的预后仍然具有挑战性。本研究旨在使用新的指标——基因-基因（G×G）相互作用构建和验证 IDH 野生型 GBM（COVPRIG）的稳健预后模型，以预测总生存期，并探索其分子和细胞基础。
对包含 800 个样本的四个独立的跨种族队列进行单变量和多变量 Cox 回归分析。通过系统文献回顾，全面评估和比较预测效能，并与以前的模型进行比较。通过对肿瘤和单细胞数据集的综合分析，阐明 COVPRIG 的分子基础。
使用 Cox-ph 模型的方法，从 93961 个 G×G 相互作用中筛选出 6 个相互作用，形成一个最佳组合，与年龄一起构成了 COVPRIG 模型。COVPRIG 分别针对 RNA-seq 和微阵列进行设计，并有效地识别出高死亡率风险的患者。COVPRIG 的预测性能令人满意，曲线下面积（AUC）范围从 0.56（CGGA693，RNA-seq，6 个月生存）到 0.79（TCGA RNAseq，18 个月生存），可以通过决策曲线进一步验证。分别为 RNA-seq 和微阵列队列构建了个体风险预测的列线图。此外，还在包括 IDH 突变样本在内的 GBM 中验证了 COVPRIG 的预后意义。值得注意的是，COVPRIG 进行了全面评估和外部验证，系统评价显示 COVPRIG 的综合判别改善（IDI）为 6-16%，优于目前验证的模型。此外，综合生物信息学分析预测了 METTL1 神经祖细胞样（NPC 样）恶性细胞在驱动不良结局中的重要作用。
本研究为 IDH 野生型 GBM 的预后预测提供了一个强大的工具，并为未来的研究提供了初步的分子基础。
