A lactylation-related transcriptional signature as candidate biomarkers for the early detection of polycystic ovary syndrome: an integrated genomic analysis

Abstract

Polycystic Ovary Syndrome (PCOS) is a prevalent endocrine disorder that negatively impacts women’s reproductive and metabolic health, leading to complications such as irregular menstruation, infertility, obesity, and metabolic syndrome. Current diagnostic methods, which primarily rely on clinical symptoms and hormone levels, lack specificity in early detection. This study aimed to identify lactylation-related transcriptional signatures as candidate biomarkers for the early detection of PCOS. We conducted a comprehensive bioinformatic analysis using transcriptomic data from granulosa cells of PCOS patients in the GEO database. Patients were categorized into two molecular subtypes via consensus clustering based on the expression of lactylation-related genes. Subsequently, diagnostic models were constructed and validated to evaluate the potential of these transcriptional signatures as early diagnostic markers. External validation was performed using an independent dataset, and selected candidate genes were further examined in DHEA-treated KGN cells. We identified nine key genes (HK3, SDC3, TGFBI, ZYX, LSP1, HMGA1, LCP1, B3GAT1, and ZNF280C) that were significantly differentially expressed in PCOS granulosa cell transcriptomic datasets. Pathway enrichment analyses revealed their involvement in energy metabolism, cell proliferation and apoptosis, which highlights their potential roles in PCOS pathogenesis. A random forest-based diagnostic model showed high internal predictive performance, with an ROC AUC of 0.993; however, external validation yielded a lower AUC of 0.628, indicating limited generalizability and the need for further validation. These findings suggest that lactylation-related transcriptional signatures may serve as potential candidate biomarkers for the early detection of PCOS. However, because this study was based primarily on transcriptomic inference and did not directly measure lactylation modifications, the results should be interpreted as hypothesis-generating. Future studies should validate these biomarkers in larger clinical cohorts, primary granulosa cells, and through direct lactylation or proteomic assays.