Exploring the potential biomarkers between stroke and obstructive sleep apnea by WGCNA and machine learning.

Abstract

Background: Ischemic stroke (IS) and obstructive sleep apnea (OSA) are highly prevalent disorders with significant societal and individual burdens. OSA exacerbates stroke outcomes, elevates recurrent stroke risk, and impedes functional recovery. Identifying shared biomarkers and elucidating the molecular mechanisms linking IS and OSA have been critical for developing targeted therapies and improving patient prognosis.

Methods: Transcriptomic data for IS and OSA were obtained from the GEO database (GSE58294, GSE135917, GSE38792, and GSE22255). After batch-effect correction, weighted gene co-expression network analysis (WGCNA) and differential expression analysis were performed to identify disease-associated genes. Functional enrichment analysis and a protein-protein interaction network construction were conducted. Advanced machine learning algorithms-Least Absolute Shrinkage and Selection Operator (LASSO) regression and random forests-were applied to screen hub genes, followed by validation of their diagnostic performance. Patients were stratified into high-and low-expression groups based on hub genes levels, and gene set enrichment analysis (GSEA) was performed to characterize pathway activity.

Results: Integration of WGCNA and differential expression analysis revealed 112 shared differentially expressed genes (DEGs) significantly associated with IS and OSA. Enrichment analysis implicated these DEGs in critical processes, including protein ubiquitination, fatty acid metabolism, cell proliferation and apoptosis, autophagy, cyclooxygenase pathway, and chromatin remodeling. Machine learning identified DUSP1 as a central hub gene, with significantly elevated expression in both IS and OSA. Diagnostic validation demonstrated robust performance for DUSP1 (AUCs: 1.000 in GSE58294, 0.885 in GSE135917, 0.718 in GSE22255), though variability was observed in GSE38792 (AUC: 0.487). GSEA highlighted distinct pathway signatures: high DUSP1 expression correlated with activation of ribosome, spliceosome, and nucleocytoplasmic transport pathway, while suppressing Ras/Rap1 signaling, platelet activation, PI3K-AKT signaling, IL-17 signaling, and immune-related pathways (e.g., Fc gamma R-mediated phagocytosis, cytokine-cytokine receptor interaction, and B cell receptor signaling pathway).

Conclusion: Through integrative bioinformatics and machine learning, this study identifies DUSP1 as a novel hub gene linking IS and OSA. Functional annotation reveal its involvement in shared biological pathways, offering mechanistic insights into disease pathogenesis and highlighting DUSP1 as a potential therapeutic target.