A multi-chip integrated analysis employing machine learning techniques identifies biomarkers for diagnosing PCOS and its comorbidity with T2D

Polycystic ovary syndrome (PCOS) is a common endocrine disorder closely linked to type 2 diabetes (T2D). Although linked, potential biomarkers for PCOS have not yet been sufficiently studied. This study aimed to identify diagnostic biomarkers linked to PCOS through an integrative analysis of multi-chip data and to examine their relationship with T2D. We used multi-chip datasets from public databases, conducted differential expression analysis to screen genes, and performed cross-disease gene overlap analysis using Venn diagrams. We conducted functional enrichment analyses, including gene ontology and Kyoto Encyclopedia of Genes and Genomes analyses, to investigate the biological significance of the intersecting genes. Based on these findings, we developed a protein-protein interaction network and identified key hub genes. We utilized a Least Absolute Shrinkage and Selection Operator regression model with 10-fold cross-validation for modeling. Subsequently, univariate and multivariate logistic regression analyses were performed to assess T2D risk. We also assessed the model’s ability to distinguish T2D using a diagnostic nomogram. Our results identified PRPF31, HABP4, CTPS2, and GFM1 as core hub genes with diagnostic potential. The effectiveness of these biomarkers was validated using receiver operating characteristic curve analysis across multiple datasets. Logistic regression analyses revealed that PRPF31 serves as an independent predictor of T2D, with reduced expression significantly correlating with increased T2D risk. The C-index from the diagnostic nomogram analysis indicated that the model possessed a strong discriminative ability for T2D. These findings enhance our understanding of the molecular mechanisms linking PCOS and T2D, laying the groundwork for clinical applications in the early detection and management of PCOS and its associated conditions.