Background: Chronic obstructive pulmonary disease (COPD) is a heterogeneous syndrome, resulting in inconsistent findings across studies. Identifying a core set of genes consistently involved in COPD pathogenesis, independent of patient variability, is essential.
Methods: We integrated lung tissue sequencing data from patients with COPD across two centers. We used weighted gene co-expression network analysis and machine learning to identify 13 potential pathogenic genes common to both centers. Additionally, a gene-based model was constructed to distinguish COPD at the molecular level and validated in independent cohorts. Gene expression in specific cell types was analyzed, and Mendelian randomization was used to confirm associations between candidate genes and lung function/COPD. Preliminary in vitro functional validation was performed on prioritized core candidate genes.
Results: Tissue inhibitor of metalloproteinase 4 (TIMP4) was identified as a key pathogenic gene and validated in COPD cohorts. Further analysis using single-cell sequencing from mice and patients with COPD revealed that TIMP4 is involved in ciliated cells. In primary human airway epithelial cells cultured at the air-liquid interface, TIMP4 overexpression reduced ciliated cell numbers.
Conclusions: We developed a 13-gene model for distinguishing COPD at the molecular level and identified TIMP4 as a potential hub pathogenic gene. This finding provides insights into shared disease mechanisms and positions TIMP4 as a promising therapeutic target for further investigation.
Keywords: Chronic obstructive pulmonary disease; Machine learning; Matrix metallopeptidase 4; Mendelian randomization.
© 2025. The Author(s).