Zhou Xianqiang, Meng Yufeng, Yang Jie, Wang Hongtao, Zhang Yixin, Jin Zhengjie, Feng Cuiling
Department of Traditional Chinese Medicine, Peking University People's Hospital, Beijing, 100032, China.
Institute of Integrated Traditional Chinese and Western Medicine, Peking University, Beijing, 100871, China.
Inflamm Res. 2025 Apr 17;74(1):66. doi: 10.1007/s00011-025-02025-4.
Chronic obstructive pulmonary disease (COPD) is the leading cause of respiratory system-related mortality worldwide. Although COPD is associated with immune regulation, its underlying mechanisms remain unclear.
Cells from the single-cell RNA sequencing (scRNA-seq) datasets were subjected to clustering analysis and cell type identification to isolate immune cell subgroups specifically expressed in COPD. High-dimensional weighted gene co-expression network analysis (hdWGCNA) was used to identify hub genes related to the immune cell subpopulations. Machine learning algorithms were applied to identify diagnostic genes in the immune cell subpopulations and construct clinical diagnostic models for COPD. In bulk RNA sequencing data, AUC curves were used to assess the stability of the diagnostic models in predicting COPD.
Through 2 rounds of clustering analysis, the macrophage subgroups 1, 2, 7, 11, and 13 which specifically expressed in COPD (COPD_Mφ) were identified. HdWGCNA analysis revealed a hub set of genes closely related to COPD_Mφ from black, blue, yellow, and brown modules. Nonnegative Matrix Factorization (NMF) analysis separated the COPD samples into 2 clusters, with significant increases in the infiltration of Monocytic_lineage, Myeloid_dendritic_cells, and Neutrophils in cluster 1 (P < 0.001). Univariate logistic regression and LASSO regression analyses identified 11 feature genes associated with COPD_Mφ, including CST3, LGALS3, CSTB, S100A10, CYBA, S100A11, ARPC3, FTH1, PFN1, MAN2B1, and RPL39. The RF and convolutional neural network (CNN) models constructed using these feature genes effectively distinguished between normal and COPD patients. Among them, S100A10, RPL39, and FTH1 exhibited differential expression between COPD patients and normal individuals and could serve as potential clinical diagnostic markers for COPD.
The study provides new insights into the immune mechanisms of COPD and lays the theoretical foundation for its future clinical diagnosis and personalized treatment.
慢性阻塞性肺疾病（COPD）是全球呼吸系统相关死亡的主要原因。虽然COPD与免疫调节有关，但其潜在机制仍不清楚。
对单细胞RNA测序（scRNA-seq）数据集的细胞进行聚类分析和细胞类型鉴定，以分离在COPD中特异性表达的免疫细胞亚群。使用高维加权基因共表达网络分析（hdWGCNA）来鉴定与免疫细胞亚群相关的枢纽基因。应用机器学习算法鉴定免疫细胞亚群中的诊断基因，并构建COPD的临床诊断模型。在批量RNA测序数据中，使用AUC曲线评估诊断模型在预测COPD方面的稳定性。
通过2轮聚类分析，鉴定出在COPD中特异性表达的巨噬细胞亚群1、2、7、11和13（COPD_Mφ）。HdWGCNA分析揭示了来自黑色、蓝色、黄色和棕色模块的与COPD_Mφ密切相关的一组枢纽基因。非负矩阵分解（NMF）分析将COPD样本分为2个簇，簇1中单核细胞谱系、髓样树突状细胞和中性粒细胞的浸润显著增加（P < 0.001）。单变量逻辑回归和LASSO回归分析确定了11个与COPD_Mφ相关的特征基因，包括CST3、LGALS3、CSTB、S100A10、CYBA、S100A11、ARPC3、FTH1、PFN1、MAN2B1和RPL39。使用这些特征基因构建的随机森林（RF）和卷积神经网络（CNN）模型能够有效区分正常人和COPD患者。其中，S100A10、RPL39和FTH1在COPD患者和正常个体之间表现出差异表达，可作为COPD潜在的临床诊断标志物。
该研究为COPD的免疫机制提供了新见解，为其未来的临床诊断和个性化治疗奠定了理论基础。
