Abstract
BACKGROUND: The causal bridge from environmental exposure to endometriosis (Ems) biology remains incompletely defined. Di(2-ethylhexyl) phthalate (DEHP) is repeatedly implicated in elevated Ems risk, yet actionable molecular anchors linking exposure to phenotype are scarce. METHODS: We established a multi-layered pipeline centered on DEHP. Comprehensive in silico target prediction across ChEMBL, PharmMapper, and SwissTargetPrediction yielded 1364 de-duplicated candidate proteins. Three transcriptomic cohorts (GSE51981, GSE6364, GSE7305) were integrated and analyzed using differential expression and Weighted Gene Co-expression Network Analysis (WGCNA) to derive a 229-gene, high-confidence Ems set. The intersection identified 17 overlapping genes, which were contextualized by protein-protein interaction (PPI) networks and Gene Ontology/Kyoto Encyclopedia of Genes and Genomes (GO/KEGG) enrichment. Interpretable machine learning with SHapley Additive exPlanations (SHAP) prioritized a core signature, followed by molecular docking and 100-ns molecular dynamics (MD) simulations to validate binding feasibility and temporal stability. RESULTS: The 17-gene overlap formed a compact functional subnetwork aligned with a "membrane-lipid homeostasis to vesicular transport to detoxification/de-esterification" axis. Classifiers showed robust discrimination across training and external cohorts (most area under the receiver operating characteristic curve [AUC] > 0.75), while single-gene receiver operating characteristic (ROC) analyses highlighted UGT8 (AUC = 0.869) and EPHX1 (0.853) as highly transferable. SHAP prioritized a seven-gene signature-ELOVL6, LYPLA1, UGT8, SLC1A5, HMGCR, EPHX1, and VAMP2-and revealed non-linear relationships, including ELOVL6-UGT8 synergy, HMGCR-LYPLA1 antagonism, and EPHX1-SLC1A5 context dependence. Docking supported pocket complementarity with ~ 2.2-3.3 Å hydrogen bonds plus extensive hydrophobic/π contacts; MD confirmed stable, compact, and persistent binding for UGT8-DEHP, ELOVL6-DEHP, and HMGCR-DEHP over 100 ns. CONCLUSIONS: This study establishes a comprehensive workflow spanning from chemical exposure identification to target discovery, disease network mapping, interpretable computational modeling, and structural/dynamical validation. We propose a DEHP-Ems regulatory framework underpinned by lipid metabolism, vesicular trafficking, and detoxification pathways. The resulting seven-gene signature provides a clinically applicable panel for diagnostic stratification and highlights potential therapeutic entry points, particularly along the HMGCR axis and via SLC1A5-mediated glutamine uptake. These findings lay the groundwork for future mechanistic studies in primary cell systems, organoid models, in vivo experiments, and prospective clinical validation.