Identification of biomarkers related to oxidative stress in polycystic ovary syndrome based on bioinformatics

Abstract Background Polycystic ovary syndrome (PCOS) is a prevalent disorder that impacts reproductive and endocrine function in women of childbearing age, affecting approximately 20–25% of this population.Given the inherent ambiguity and inconsistency in diagnostic procedures for PCOS, coupled with limited treatment options, there is an urgent imperative to explore novel approaches that can enhance both clinical diagnosis and therapeutic interventions for PCOS. Methods Differentially expressed genes(DEGs) were obtained by comparing the difference of gene expression level between PCOS patients and control samples.Construct co-expression networks and recognition modules, and record the genes in each module.Intersection genes of oxidative stress-related genes(OSRGs), key genes screened by WGCNA and DEGs were screened.Enrichment of GO and KEGG pathways was analyzed and the PPI network of key OSRGs for PCOS was constructed in subsequent studies.LASSO regression, SVM-RFE algorithm and random forest (RF) algorithm were used to screen PCOS biomarkers."pROC" was used to assess the predictive ability of the biomarker and its nomogram for PCOS.Immunoinfiltration analysis was performed, and the correlation between biomarker prognostic genes and immune cells was calculated by Spearman method.After the TF-mRNA-miRNA regulatory network was constructed, drug prediction and animal modeling were performed.Finally, the PCOS mouse model was constructed and verified by vaginal epithelial staining and HE staining, and the differential expression of biomarkers in PCOS and control mice was detected by RT-qPCR. Results We first screened 364 down-regulated genes and 397 up-regulated genes.We constructed 25 co-expression modules, in which salmon and greenyellow modules were positively correlated with PCOS.The salmon and greenyellow modules comprised 145 and 166 genes, respectively.KEGG and GO enrichment analyses elucidated the signaling regulatory pathways and biological functions of key genes.Five genes (LTA, PLA2G7, TNFSF10, NCF2, and BCL2A1) were then identified as prognostic markers for PCOS. Compared with normal patients, LTA, PLA2G7, TNFSF10, NCF2 and BCL2A1 expressions were significantly up-regulated in PCOS patients.There were significant differences in Activated CD4 T cell and Memory B cell between PCOS patients and normal samples.Drug prediction results indicated that FENRETINIDE, Doxorubicin Hydrochloride and Demecolcine may be key drugs for the treatment of PCOS.Finally, the results of vaginal epithelial staining and HE staining showed that the PCOS mice were successfully constructed, and RT-qPCR results showed that the expressions of LTA, PLA2G7, TNFSF10, NCF2, and BCL2A1 were statistically different between the control group and the PCOS group. Conclusion We identified a total of 23 important OSRGs based on bioinformatics analysis. Subsequently, employing machine learning techniques, we successfully pinpointed five potential diagnostic markers (LTA, PLA2G7, TNFSF10, NCF2, and BCL2A1) that may play crucial roles in the progression of PCOS. This novel approach enhances the reliability and accuracy of PCOS diagnosis.