Integrative Bioinformatics, Experimental Validation, and Interpretable Machine Learning Reveal...

Cadmium (Cd) is a toxic heavy metal strongly implicated in lung adenocarcinoma (LUAD) through mechanisms involving oxidative stress, epigenetic dysregulation, and chronic inflammation. This study aimed to identify Cd-responsive genes associated with LUAD and to evaluate the protective effects of oxyresveratrol (O-RES) against Cd-induced molecular alterations. Using an integrated bioinformatics approach across six GEO datasets, key differentially expressed genes (DEGs) were identified and subsequently validated in silico and in vivo using a Cd-induced rat lung injury model. DEG analysis revealed four hub genes: Cbx2, Cdh3, Crabp2, and Slc15a3, linked to chromatin remodeling, cell adhesion, retinoid signaling, and immune regulation. Cd exposure significantly dysregulated these genes and increased pro-inflammatory cytokine expression, whereas O-RES treatment dose-dependently restored gene expression and attenuated inflammation. Molecular docking further supported favorable interactions between O-RES and the hub proteins. In addition, machine learning-based regression models were applied to integrate transcriptional responses across experimental groups. A Random Forest model achieved high predictive accuracy for a Cd-O-RES exposure index (R² = 0.90), while SHAP analysis identified Egln3 as the dominant context-dependent contributor, followed by Cbx2 and Crabp2. Complementary Elastic Net regression supported these findings through consistent linear associations. Overall, integrating interpretable machine learning with experimental evidence enhances mechanistic insight into Cd-induced transcriptional reprogramming and supports the protective role of O-RES.