Integrated network toxicology and machine learning reveal potential TDCPP-related hub genes in endometrial cancer

Endometrial cancer (EC) is a common gynecological malignancy with increasing global incidence. Although hormonal dysregulation, metabolic disorders, and genetic alterations are known contributors, the role of environmental contaminants remains unclear. Tris(1,3-dichloro-2-propyl) phosphate (TDCPP), an organophosphate flame retardant with endocrine-disrupting potential, has been linked to oxidative stress and cell proliferation, but its potential molecular links to EC remain unexplored. We integrated bioinformatics and machine learning to explore TDCPP–EC associations. TDCPP targets were collected from ChEMBL and CTD. EC transcriptomic data were obtained from GEO and TCGA. Differentially expressed genes were intersected with TDCPP targets and EC-related genes. Functional enrichment, PPI networks, and CytoHubba identified candidate hub genes, which were further refined using LASSO, SVM-RFE, and random forest. SHAP analysis, ROC curves, immune infiltration, and molecular docking were used for model interpretation and validation. We identified 352 TDCPP-related targets and 113 TDCPP-related DEGs. Enriched pathways included endocrine resistance, oxidative stress, and cellular senescence. Fourteen hub genes were narrowed to six (ATF2, AURKA, AXIN2, FGFR2, FGFR3, KPNA2) by all three machine learning methods. SHAP highlighted ATF2, FGFR3, and AXIN2 as influential features. The six-gene signature showed strong diagnostic performance in both GEO (training) and TCGA (validation) cohorts. Immune infiltration linked these genes to T cells, NK cells, and neutrophils. Molecular docking predicted favorable binding of TDCPP to all six proteins, especially FGFR3 and FGFR2. This study provides a hypothesis-generating bioinformatics framework suggesting possible associations between TDCPP targets and EC and identifies six potential hub genes involved in oxidative stress, receptor signaling, and immune regulation. These findings offer candidate biomarkers and mechanistic insights, but experimental validation is required.