Wei Haotian, Li Xinlong, Feng Peng, He Zhaohui
Department of Neurosurgery, First Affiliated Hospital of Chongqing Medical University, Chongqing, 400016, China.
Lanzhou University Second Hospital, The Second Medical College of Lanzhou University, Lanzhou, 730030, China.
J Mol Neurosci. 2025 Apr 21;75(2):51. doi: 10.1007/s12031-025-02349-0.
Glioma, as one of the most complex and prognostically variable malignant tumors of the central nervous system, poses a significant challenge to clinical decision-making due to its molecular heterogeneity. This study aims to deepen our understanding of glioma molecular subtypes and explore key gene markers with prognostic and diagnostic value. We utilized an angiogenesis-related gene set and employed the Non-negative Matrix Factorization (NMF) algorithm to successfully identify two distinct prognostic subtypes, with subtype one exhibiting more unfavorable prognostic characteristics. To further elucidate the biological functional differences between these two subtypes, we conducted Gene Ontology (GO) functional annotation, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, and Gene Set Enrichment Analysis (GSEA). Building on this, we integrated differentially expressed genes between subtypes with core genes revealed by Weighted Gene Co-expression Network Analysis (WGCNA) through intersection analysis to pinpoint a series of key candidate genes. Subsequently, we constructed a Protein-Protein Interaction (PPI) network to identify genes occupying central nodes within the network. To screen markers with high specificity and sensitivity for prognosis and diagnosis, we adopted a dual-track strategy: on the one hand, we utilized machine learning algorithms such as Lasso regression, Support Vector Machine (SVM), and Random Forest (RF) to select core genes, identifying markers that can accurately predict the subtype with a poor prognosis; on the other hand, we employed a comprehensive evaluation system incorporating 101 machine learning ensemble algorithms to further validate and screen prognosis-related genes. Through cross-validation using these two strategies, we ultimately determined SERPINH1 and CTSZ as dual prognostic and diagnostic markers for glioma. This study not only provides a new perspective and tool for the molecular subclassification of glioma but also, through a rigorous multi-algorithm, multi-dimensional screening process, uncovers SERPINH1 and CTSZ as markers with potential clinical translational value. These findings are expected to offer more precise biomarker support for the early diagnosis and prognostic assessment of glioma, potentially paving new avenues for the development of personalized treatment strategies and improving patient outcomes. This has far-reaching implications for the clinical management of glioma in the field of neurosurgery.
胶质瘤作为中枢神经系统最复杂且预后差异最大的恶性肿瘤之一,因其分子异质性给临床决策带来了重大挑战。本研究旨在加深我们对胶质瘤分子亚型的理解,并探索具有预后和诊断价值的关键基因标志物。我们利用一个血管生成相关基因集,并采用非负矩阵分解(NMF)算法成功识别出两种不同的预后亚型,其中亚型一表现出更不利的预后特征。为了进一步阐明这两种亚型之间的生物学功能差异,我们进行了基因本体论(GO)功能注释、京都基因与基因组百科全书(KEGG)通路分析以及基因集富集分析(GSEA)。在此基础上,我们通过交集分析将亚型之间的差异表达基因与加权基因共表达网络分析(WGCNA)揭示的核心基因整合起来,以确定一系列关键候选基因。随后,我们构建了蛋白质 - 蛋白质相互作用(PPI)网络,以识别在网络中占据中心节点的基因。为了筛选出对预后和诊断具有高特异性和敏感性的标志物,我们采用了双轨策略:一方面,我们利用诸如套索回归、支持向量机(SVM)和随机森林(RF)等机器学习算法来选择核心基因,识别能够准确预测预后不良亚型的标志物;另一方面,我们采用一个包含101种机器学习集成算法的综合评估系统来进一步验证和筛选与预后相关的基因。通过使用这两种策略进行交叉验证,我们最终确定丝氨酸蛋白酶抑制剂H1(SERPINH1)和组织蛋白酶Z(CTSZ)作为胶质瘤的双重预后和诊断标志物。本研究不仅为胶质瘤的分子亚分类提供了新的视角和工具,而且通过严格的多算法、多维度筛选过程,发现SERPINH1和CTSZ作为具有潜在临床转化价值的标志物。这些发现有望为胶质瘤的早期诊断和预后评估提供更精确的生物标志物支持,可能为个性化治疗策略的开发和改善患者预后开辟新途径。这对神经外科领域胶质瘤的临床管理具有深远意义。