Potential therapeutic targets for obstructive sleep Apnea were identified through network pharmacology, WGCNA, machine learning, immune infiltration analysis, and ceRNA network.

BACKGROUND: Our study aims to explore the early genomic diagnostic markers of OSA and the corresponding drug prediction targets using network pharmacology analysis, and to elucidate the etiology and pathogenesis of OSA from the genetic level.

METHODS: We selected OSA-related gene datasets (GSE135917 and GSE38792) from the Gene Expression Omnibus (GEO). Principal component analysis (PCA) was performed to remove outlier samples, and batch correction was applied to the two datasets. The raw expression matrix was log2-transformed, and samples were divided into normal and OSA groups. Differentially expressed genes (DEGs) were identified, and WGCNA analysis was performed on these DEGs to identify mitochondria-associated hub genes, followed by functional enrichment analysis, PPI network construction, core lncRNA-related ceRNA network construction. We then screened core genes with a high risk of OSA. Based on the core genes, we established an easy-to-use nomogram and verified its accuracy in identifying OSA patients then performed a differential expression analysis of the core genes, GSEA and GSVA analyses, and immune infiltration analysis. Finally, we constructed the disease prediction model and predicted the drug targets, thereby obtain a genomic prediction model for OSA.

RESULTS: After PCA and batch correction, an expression matrix comprising 13 normal samples and 19 OSA patient samples was finally included. We identified 1500 differentially expressed genes (DEGs) through differential expression analysis, then screened 61 hub genes by WGCNA analysis, and established an OSA-associated ceRNA network containing 75 predictive miRNAs, 129 lncRNAs and 5 mRNAs. Six robust key genes were identified through PPI network construction: TUFM, CYCS, UQCRC1, COX4I1, TIMM50, and NDUFV1. Finally, after LASSO regression and nomogram validation, a predictive model containing 2 core genes (UQCRC1 and COX4I1) was obtained, and its area under the ROC curve (AUC) was 0.919. Drug target prediction of the core genes showed that 1-Methyl-4-phenyl-2,3-dihydropyridinium CTD 00002003, Cube root extract CTD 00006707, Disodium selenite CTD 00007229, and mitotane CTD 00006344 had good effects.

CONCLUSION: Our current findings provide a rationale for identifying therapeutic targets in the diagnosis and treatment of OSA. In addition, these findings have the potential to facilitate the translation of our study to clinical applications in the future. Insight Box Different from other single-gene predictors of OSA, network pharmacological analysis identified differential genes and explored biomolecular markers of OSA from multi-gene and multi-target perspectives through enrichment analysis, construction of ceRNA gene network, and correlation analysis and explore early genomic diagnostic indicators of OSA and corresponding drug prediction targets using network pharmacological analysis and to elucidate the etiology and pathogenesis of OSA from the genetic level. To provide a more accurate method for the diagnosis, prevention, and treatment of clinical OSA, it is expected to provide a research basis for the accurate diagnosis of clinical OSA and the pathogenesis of OSA.