Abstract
BACKGROUND: The pathophysiology of obstructive sleep apnea (OSA) and diabetes mellitus (DM) is still unknown, despite clinical reports linking the two conditions. After investigating potential roles for DM-related genes in the pathophysiology of OSA, our goal is to investigate the molecular significance of the condition. Machine learning is a useful approach to understanding complex gene expression data to find biomarkers for the diagnosis of OSA. METHODS: Differentially expressed analysis for OSA and DM data sets obtained from GEO were carried out firstly. Then four machine algorithms were used to screen candidate biomarkers. The diagnostic model was constructed based on key genes, and the accuracy was verified by ROC curve, calibration curve and decision curve. Finally, the CIBERSORT algorithm was used to explore immune cell infiltration in OSA. RESULTS: There were 32 important genes that were considered to be related both in OSA and DM datasets by differentially expressed analysis. Through enrichment analysis, the majority of these genes are enriched in immunological regulation, oxidative stress response, and nervous system control. When consensus characteristics from all four approaches were used to predict OSA diagnosis, STK17A was thought to have a high degree of accuracy. In addition, the diagnostic model demonstrated strong performance and predictive value. Finally, we explored the immune cells signatures of OSA, and STK17A was strongly linked to invasive immune cells. CONCLUSION: STK17A has been discovered as a gene that can differentiate between individuals with OSA and DM based on four machine learning methods. In addition to offering possible treatment targets for DM-induced OSA, this diagnostic approach can identify high-risk DM patients who also have OSA.