Background: Polycystic ovary syndrome (PCOS) is a highly prevalent and complex endocrine-metabolic syndrome, characterized by impaired folliculogenesis and follicular arrest. The marked clinical heterogeneity observed among PCOS patients is hypothesized to reflect variations in its etiopathogenesis, although the underlying mechanisms remain incompletely elucidated. This study was conducted employing individual-patient protein-protein interaction (PPI) network analysis to investigate key candidate genes and their associated signaling pathways in the granulosa cells (GCs) of PCOS patients at a single-patient resolution. The aim of this study was to provide novel insights and deepen the understanding of the pathogenic mechanisms and heterogeneity of PCOS.

Methods: Data were extracted from three mRNA expression datasets derived from high-throughput sequencing (GSE168404, GSE155489, and GSE138518) available in the Gene Expression Omnibus (GEO) database. These datasets included 24 samples, comprising 12 from individuals with PCOS and 12 controls. A gene correlation matrix was generated using control samples to assess global transcriptional alterations induced by data from individual PCOS patients. Individual-patient PPI networks were subsequently constructed to identify key candidate genes and molecular subtypes unique to each patient. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed via KOBAS-i to elucidate the underlying biological functions.

Results: A total of 7752 nodes (1978 unique) and 19,219 edges (13,626 unique) were mapped to construct the individual-patient PPI networks. These networks exhibited considerable variability in terms of nodes (median = 2128, range = 1141–3838) and edges (median = 2369, range = 903–6237). Despite some shared molecular features, the distinct network architectures observed in each patient underscored patient-specific variations in gene expression, thereby reflecting the intrinsic heterogeneity of the PCOS population. To identify key candidate genes, five overlapping nodes (fibronectin 1 (FN1), DNA ligase 3 (LIG3), vesicle associated membrane protein 2 (VAMP2), kinesin family member 23 (KIF23), pescadillo ribosomal biogenesis factor 1 (PES1)) were extracted from these networks, and KIF23, along with protein regulator of cytokinesis 1 (PRC1), were identified as common hub genes when analyses were restricted to more than 50% of patients. Furthermore, four signaling pathways, cyclic Guanosine Monophosphate-Protein Kinase G (cGMP-PKG), Wingless/Int-1 (Wnt), Relaxin, and Apelin, were differentially enriched across all individual-patient PPI networks. These pathways were distinct from those enriched with differentially expressed genes (DEGs) between the PCOS and control groups.

Conclusion: This study revealed patient-specific variations in gene expression by examining distinct network characteristics across individuals, highlighting the molecular heterogeneity inherent in PCOS. In addition to identifying candidate genes and signaling pathways enriched with DEGs, KIF23 emerged as a potential hub gene due to its ubiquitous presence. Additionally, the cGMP-PKG, Wnt, relaxin, and apelin signaling pathways were identified as potential core signaling pathways. However, these findings require further experimental validation through comprehensive in vivo and in vitro studies to establish their biological relevance.