To investigate the potential molecular links between nonylphenol (NP) exposure and osteonecrosis of the femoral head (ONFH), we employed an integrated strategy combining network toxicology, transcriptomic analysis, weighted gene co-expression network analysis (WGCNA), machine learning, immune infiltration analysis, molecular docking, molecular dynamics simulation, and independent dataset validation. Based on the GSE123568 discovery dataset of steroid-induced ONFH and steroid-exposed non-ONFH controls, 391 differentially expressed genes were identified, including 123 upregulated and 268 downregulated genes. Intersection of DEGs with WGCNA-derived module genes yielded 342 SONFH-associated candidate genes for subsequent analysis. NP-related targets were collected from CTD, SwissTargetPrediction, and Target Net databases, and 40 overlapping genes were identified as potential NP-related ONFH targets. PPI network analysis highlighted 10 key genes, while LASSO regression, random forest, and SVM-RFE identified PECAM1, CXCR2, and MAP2K3 as common core feature genes. Integrated candidate genes were mainly associated with inflammation, oxidative stress, vascular regulation, extracellular matrix remodeling, endocrine-lipid metabolic disturbance, and bone metabolism. ROC analysis showed that the candidate gene set discriminated SONFH samples from steroid-exposed non-ONFH controls in the discovery dataset, with an AUC of 0.967. Immune infiltration analysis revealed that PECAM1 was significantly correlated with multiple immune cell subsets. Molecular docking and a 100 ns molecular dynamics simulation provided exploratory structural support for a potential NP-PECAM1 interaction. In summary, PECAM1 may represent a prioritized candidate node linking NP-related molecular perturbations with SONFH/ONFH-associated endothelial-immune dysregulation.