Unraveling diethyl phthalate-induced prostate carcinogenesis: core targets revealed by integrated network toxicology, machine learning, and structural validation.
Diethyl phthalate (DEP), a widely distributed environmental contaminant, is epidemiologically linked to prostate cancer (PCa). However, its molecular mechanisms beyond endocrine disruption remain poorly defined. We aimed to investigate the core mechanisms potentially underlying DEP-associated prostate carcinogenesis within a genome-exposome interaction framework.
We employed an integrated, multi-level framework combining: (1) Integrated chemical structure-based target prediction; (2) Cross-dataset meta-analysis of PCa transcriptomics (7 GEO datasets) for Differentially Expressed Gene (DEG) identification and Weighted Gene Co-expression Network Analysis (WGCNA); (3) Ensemble machine learning (113 models incorporating RF, XGBoost) for core target screening, augmented by SHAP interpretable to predict potential DEP targets.e AI; and (4) Molecular docking validation (AutoDock Vina, binding free energy assessment).
Integration pinpointed 9 key DEP-PCa targets. Functional enrichment implicated calcium signaling dysregulation, neuroendocrine pathway disruption, and smooth muscle dysfunction as central mechanisms. Machine learning distilled five core regulators: TRPM8, CTSB, CA14, GSTM2, and MYLK. SHAP analysis quantified TRPM8 and CA14 as dominant predictors and revealed critical non-linear interactions: synergistic TRPM8-MYLK co-expression and a CTSB expression threshold effect. Computational validation predicted high-affinity binding of DEP to all five core targets, suggesting potential direct interactions.
Our integrated analysis suggests that DEP may promote prostate carcinogenesis via a multidimensional network centered on calcium signaling perturbation, neuroendocrine dysregulation, and tumor microenvironment acidification, potentially illustrating a genome-exposome interaction mechanism beyond endocrine disruption. We propose that our analytical framework could serve as a reproducible approach for translational exposomics.
We employed an integrated, multi-level framework combining: (1) Integrated chemical structure-based target prediction; (2) Cross-dataset meta-analysis of PCa transcriptomics (7 GEO datasets) for Differentially Expressed Gene (DEG) identification and Weighted Gene Co-expression Network Analysis (WGCNA); (3) Ensemble machine learning (113 models incorporating RF, XGBoost) for core target screening, augmented by SHAP interpretable to predict potential DEP targets.e AI; and (4) Molecular docking validation (AutoDock Vina, binding free energy assessment).
Integration pinpointed 9 key DEP-PCa targets. Functional enrichment implicated calcium signaling dysregulation, neuroendocrine pathway disruption, and smooth muscle dysfunction as central mechanisms. Machine learning distilled five core regulators: TRPM8, CTSB, CA14, GSTM2, and MYLK. SHAP analysis quantified TRPM8 and CA14 as dominant predictors and revealed critical non-linear interactions: synergistic TRPM8-MYLK co-expression and a CTSB expression threshold effect. Computational validation predicted high-affinity binding of DEP to all five core targets, suggesting potential direct interactions.
Our integrated analysis suggests that DEP may promote prostate carcinogenesis via a multidimensional network centered on calcium signaling perturbation, neuroendocrine dysregulation, and tumor microenvironment acidification, potentially illustrating a genome-exposome interaction mechanism beyond endocrine disruption. We propose that our analytical framework could serve as a reproducible approach for translational exposomics.