Integrated transcriptomic and co-expression network analysis identifies immune-metabolic biomarkers of polycystic ovary syndrome in granulosa cells

Fig. 1. Schematic overview of the study workflow. The analysis included transcriptomic data collection and preprocessing, identification of PCOS-related genes through DEG and WGCNA analysis, biomarker selection using LASSO regression, gene prioritization based on GO semantic similarity, and diagnostic modeling with ANN. Subsequent analyses involved functional enrichment, immune-metabolic pathway profiling, integration with disease-related regulatory factors, diagnostic evaluation of key genes, and experimental validation in PCOS-like granulosa cell models

Fig. 2. Transcriptomic analysis and WGCNA identify PCOS-related genes. A and B PCA plots before and after batch effect correction using SVA, showing improved clustering between control and PCOS groups. C Volcano plot of DEGs between PCOS and control samples (|logFC| >0.585, p < 0.05). D Heatmap of 184 DEGs illustrating distinct expression profiles. E and F Scale-free topology fit index and mean connectivity plots for determining soft-threshold power (β = 8). G Cluster dendrogram of gene modules from WGCNA. H Module-trait correlation heatmap identifying the gray60 module as significantly correlated with PCOS. I Venn diagram of intersecting genes between DEGs and the gray60 module, yielding 29 key genes

Fig. 3. Construction of diagnostic models using LASSO and ANN. A and B GO and KEGG enrichment analysis of 29 key genes indicating immune-related functions.C and D LASSO regression model with coefficient profiles and cross-validation error plot. E LASSO-derived coefficients for nine predictive genes. F and G ROC curves demonstrating diagnostic accuracy in training (AUC = 0.9511) and validation (AUC = 0.7188) datasets. H and I Importance ranking of key genes by GOSemSim-based semantic similarity. J Structure of ANN model based on five key genes. K and L ROC curves for ANN model performance in training (AUC = 0.938) and test (AUC = 0.875) cohorts

Fig. 4. Immune landscape of key genes in PCOS. A Pathway correlation matrix showing associations between key genes and immune/metabolic pathways. B-F Pearson correlation heatmaps illustrating relationships between key genes and chemokines (B), receptors (C), MHC molecules (D), immune inhibitors (E), and immune stimulators (F), indicating immunoregulatory roles of these genes in PCOS

Fig. 5. Pathway enrichment and metabolic activity profiling of key genes. A-E GSEA plots of CLDN11, HLA-DMA, TAB3, COLQ, and LYN showing enrichment in immune, metabolic, and cell cycle pathways. F ssGSEA-based heatmap showing differential activity of metabolism-related pathways between PCOS and control groups, with emphasis on amino acid and lipid metabolism

Fig. 6. Regulatory and diagnostic evaluation of key genes. A Gene regulatory network of transcription factors targeting five key genes. B Top transcription factor binding motifs ranked by NES. C Correlation of key genes with PCOS-associated genes from the GeneCards database, including associations with AR and CYP11A1. D-H ROC curves demonstrating diagnostic performance of CLDN11, HLA-DMA, TAB3, COLQ, and LYN

Fig. 7. In vitro validation of key gene expression under PCOS-mimicking conditions. A-D qRT-PCR showing gene expression changes under HA, IR, CLI, and HA + IR + CLI models. E-I Comparative expression of individual genes under all four conditions. J Heatmap showing Z-score-normalized gene expression across all groups, highlighting condition-specific and synergistic responses. K Knockdown of CLDN11 attenuates HA-induced AR expression. L-M Knockdown of LYN reduces CLI-induced IL6 expression and IL-6 secretion. Data are presented as mean ± SD. *p < 0.05. Statistical significance was determined using t-test (A-D) or one-way ANOVA with Tukey’s post hoc test (E-I, K-M); bars with different lowercase letters indicate p < 0.05