Identification of programmed cell death-related genes and diagnostic biomarkers in endometriosis using a machine learning and Mendelian randomization approach.

Background: Endometriosis (EM) is a prevalent gynecological disorder frequently associated with irregular menstruation and infertility. Programmed cell death (PCD) is pivotal in the pathophysiological mechanisms underlying EM. Despite this, the precise pathogenesis of EM remains poorly understood, leading to diagnostic delays. Consequently, identifying biomarkers associated with PCD is critical for advancing the diagnosis and treatment of EM. Methods: This study used datasets from the Gene Expression Omnibus (GEO) to identify differentially expressed genes (DEGs) following preprocessing. By cross-referencing these DEGs with genes associated with PCD, differentially expressed PCD-related genes (DPGs) were identified. Enrichment analyses for KEGG and GO pathways were conducted on these DPGs. Additionally, Mendelian randomization and machine learning techniques were applied to identify biomarkers strongly associated with EM. Results: The study identified three pivotal biomarkers: TNFSF12, AP3M1, and PDK2, and established a diagnostic model for EM based on these genes. The results revealed a marked upregulation of TNFSF12 and PDK2 in EM samples, coupled with a significant downregulation of AP3M1. Single-cell analysis further underscored the potential of TNFSF12, AP3M1, and PDK2 as biomarkers for EM. Additionally, molecular docking studies demonstrated that these genes exhibit significant binding affinities with drugs currently utilized in clinical practice. Conclusion: This study systematically elucidated the molecular characteristics of PCD in EM and identified TNFSF12, AP3M1, and PDK2 as key biomarkers. These findings provide new directions for the early diagnosis and personalized treatment of EM.