Zhenghua Cao,1 Shengkun Zhao,1 Shaodan Hu,2 Tong Wu,3 Feng Sun,2 LI Shi2 

1Changchun University of Traditional Chinese Medicine, Changchun, Jilin, People’s Republic of China; 2Affiliated Hospital of Changchun University of Traditional Chinese Medicine, Changchun, Jilin, People’s Republic of China; 3Geriatric Department, Suzhou Hospital of Integrated Traditional Chinese and Western Medicine, Suzhou, Jiangsu, People’s Republic of China

Correspondence: LI Shi, Email shili0648@163.com

Purpose: To employ bioinformatics and machine learning to predict the characteristics of immune cells and genes associated with the inflammatory response and ferroptosis in chronic obstructive pulmonary disease (COPD) patients and to aid in the development of targeted traditional Chinese medicine (TCM). Mendelian randomization analysis elucidates the causal relationships among immune cells, genes, and COPD, offering novel insights for the early diagnosis, prevention, and treatment of COPD. This approach also provides a fresh perspective on the use of traditional Chinese medicine for treating COPD.
Methods: R software was used to extract COPD-related data from the Gene Expression Omnibus (GEO) database, differentially expressed genes were identified for enrichment analysis, and WGCNA was used to pinpoint genes within relevant modules associated with COPD. This analysis included determining genes linked to the inflammatory response in COPD patients and analyzing their correlation with ferroptosis. Further steps involved filtering core genes, constructing TF-miRNA‒mRNA network diagrams, and employing three types of machine learning to predict the core miRNAs, key immune cells, and characteristic genes of COPD patients. This process also delves into their correlations, single-gene GSEA, and diagnostic model predictions. Reverse inference complemented by molecular docking was used to predict compounds and traditional Chinese medicines for treating COPD; Mendelian randomization was applied to explore the causal relationships among immune cells, genes, and COPD.
Results: We identified 2443 differential genes associated with COPD through the GEO database, along with 8435 genes relevant to WGCNA and 1226 inflammation-related genes. A total of 141 genes related to the inflammatory response in COPD patients were identified, and 37 core genes related to ferroptosis were selected for further enrichment analysis and analysis. The core miRNAs predicted for COPD include hsa-miR-543, hsa-miR-181c, and hsa-miR-200a, among others. The key immune cells identified were plasma cells, activated memory CD4 T cells, gamma delta T cells, activated NK cells, M2 macrophages, and eosinophils. Characteristic genes included EGF, PLG, PTPN22, and NR4A1. A total of 78 compounds and 437 traditional Chinese medicines were predicted. Mendelian randomization analysis revealed a causal relationship between 36 types of immune cells and COPD, whereas no causal relationship was found between the core genes and COPD.
Conclusion: A definitive causal relationship exists between immune cells and COPD, while the prediction of core miRNAs, key immune cells, characteristic genes, and targeted traditional Chinese medicines offers novel insights for the early diagnosis, prevention, and treatment of COPD.

Keywords: bioinformatics analysis, Mendelian randomization, machine learning, COPD, characteristic genes, targeted traditional Chinese medicine, early diagnosis