基于WGCNA与机器学习算法挖掘肾透明细胞癌M2型巨噬细胞相关关键预后基因

Mining Key M2 Macrophage-Related Prognostic Genes in Clear Cell Renal Cell Carcinoma Based on WGCNA and Machine Learning Algorithms

摘要: 目的：挖掘肾透明细胞癌(ccRCC)中与M2型巨噬细胞浸润密切相关的关键基因，并评估其临床预后价值及免疫微环境特征。方法：基于TCGA-KIRC队列转录组数据，利用CIBERSORT算法量化免疫细胞浸润，结合加权基因共表达网络分析(WGCNA)筛选与M2型巨噬细胞高度相关的基因模块，并与InnateDB数据库IRIS基因集取交集。应用单因素与多因素Cox回归及最小绝对收缩和选择算子(LASSO)回归剔除冗余特征，鉴定核心预后基因，并在GSE167573外部队列中进行验证。利用单样本基因集富集分析(ssGSEA)及相关性分析探讨关键基因与M2型巨噬细胞的关联。结果：成功筛选出CST2、HAS2及C1QTNF1为核心预后基因。这三个基因在ccRCC癌组织中呈显著高表达，且高表达与晚期病理分期及不良总体生存期(OS)显著相关。外部独立队列验证证实了其预后预测的稳健性。免疫相关性分析显示，CST2、HAS2及C1QTNF1的表达水平与M2型巨噬细胞浸润丰度均呈显著正相关。结论：本研究利用机器学习与生信方法筛选出CST2、HAS2及C1QTNF1，发现其在ccRCC中与M2型巨噬细胞浸润显著相关。这些基因可作为潜在的预后标志物，并可能参与免疫微环境重塑过程。鉴于本研究主要揭示了相关性，上述基因作为潜在候选靶点的生物学作用仍有待后续实验与临床研究进一步验证。

Abstract: Objective: To identify key genes closely related to M2 macrophage infiltration in clear cell renal cell carcinoma (ccRCC) and evaluate their clinical prognostic value and immune microenvironment characteristics. Methods: Based on transcriptomic data from the TCGA-KIRC cohort, CIBERSORT algorithm was used to quantify immune cell infiltration. Weighted Gene Co-expression Network Analysis (WGCNA) was combined to screen gene modules highly correlated with M2 macrophages, which were then intersected with the IRIS gene set from the InnateDB database. Univariate and multivariate Cox regression analyses combined with Least Absolute Shrinkage and Selection Operator (LASSO) regression were applied to eliminate redundant features and identify core prognostic genes, followed by validation in the external GSE167573 cohort. Single-sample Gene Set Enrichment Analysis (ssGSEA) and correlation analysis were used to explore the association between key genes and M2 macrophages. Results: Three core prognostic genes, CST2, HAS2, and C1QTNF1, were successfully identified. These genes were significantly highly expressed in ccRCC tumor tissues, and their high expression was associated with advanced pathological stages and poor overall survival (OS). Validation in an independent external cohort confirmed the robustness of their prognostic prediction. Immune correlation analysis showed a significant positive correlation between the expression levels of CST2, HAS2, and C1QTNF1 and the abundance of M2 macrophage infiltration. Conclusion: Using machine learning and bioinformatics approaches, this study identified CST2, HAS2, and C1QTNF1, revealing a significant correlation with M2 macrophage infiltration in ccRCC. These genes may serve as potential prognostic biomarkers and might be involved in immune microenvironment remodeling. Given that these findings primarily present correlational evidence, their biological roles as potential candidate targets require further validation through subsequent experimental and clinical studies.