Exploring the biological functions of PCOS: identifying hub androgen-related genes through bioinformatics

Polycystic ovary syndrome (PCOS) is a prevalent reproductive endocrine disorder in women. While the role of androgens in PCOS is well-recognized, the underlying mechanisms warrant further investigation. In this study, we identified potential hub androgen-related genes (ARGs) in PCOS and established diagnostic and classification models to find novel biomarkers for PCOS therapy. Five datasets (GSE34526, GSE80432, GSE95728, GSE124226, and GSE137684) were retrieved from GEO, followed by data normalization and batch effect removal. To identify hub genes, a PPI network was constructed based on differentially expressed ARGs. Subsequently, we employed the least absolute shrinkage and selection operator (LASSO) regression analysis and random forest (RF) algorithm to screen hub ARGs. Besides, hub ARGs associated with PCOS were determined by integrating the results of the three algorithms. Additionally, a nomogram was constructed using these hub ARGs to predict the risk of PCOS development. We also investigated the classification of ARG molecular subtypes and assessed immune characteristics and gene expression profiles in different subtypes. Lastly, RT-qPCR was utilized to validate the reliability of the hub genes. A total of 91 ARGs were retrieved from the GSEA website. This study included 26 healthy and 34 PCOS samples. Using the LASSO identified 13 key ARGs, RF identified 10 crucial ARGs, and PPI identified 19 pivotal ARGs. Integration of three methods identified four hub ARGs (ALDH1A1, DHRS9, PRKCB, and SGPL1). And a nomogram was constructed to predict the risk of PCOS occurrence. Notably, we validated the expression levels of the 4 hub ARGs in ovarian tissues from PCOS mice using RT-qPCR. The results showed that the expression levels of DHRS9, SGPL1, and ALDH1A1 were significantly downregulated, while PRKCB was significantly upregulated, which was consistent with our data analysis findings. Furthermore, samples were divided into two distinct ARG patterns and further explored the relationship between immune cell infiltration and these patterns. ARG scores were significantly higher in cluster A or gene cluster A compared to cluster B or gene cluster B. Finally, we evaluated the expression levels of PCOS-related genes in distinct clusters. In summary, our results may further elucidate the mechanisms of PCOS pathogenesis and offer novel ideas for PCOS diagnosis and treatment.