Obesity leads to menstrual dysfunction by impacting the "hypothalamic-pituitary-ovarian axis" in women, which can result in polycystic ovary syndrome (PCOS). The differentially expressed genes (DEGs) between the PCOS and control groups were identified using a public database, By intersecting these DEGs with key module genes and obesity related genes (ORGs), we obtained 75 differentially expressed ORGs (DE-ORGs). Further screening using machine learning led to the identification of five potential diagnostic biomarkers: CPT1A, LARS2, GSTP1, TREX1, and PILRB. The expression levels of biomarkers exhibited notable variations between the control and PCOS group, with area under curve (AUC) values exceeding 0.89 for all biomarkers, confirming their role as molecular diagnostic biomarkers for PCOS. The AUC of nomogram achieved 1, indicating its perfect predictive capability for PCOS occurrence. Single-cell analysis highlighted the crucial roles of GSTP1 and epithelial cells in the early stages of PCOS development. This study clarifies the roles of these diagnostic biomarkers, offering a theoretical foundation for the clinical assessment and treament of the disease.