Integrative machine learning and experimental validation reveal the molecular mechanisms of 2-bromo-4,6-dinitroaniline nephrotoxicity: CD45 and cyclophilin C as early response biomarkers of renal tubular injury.

2-Bromo-4,6-dinitroaniline (BDNA), a synthetic intermediate of brominated azo dyes, poses potential nephrotoxicity risks; however, its early molecular injury mechanisms remain elusive. This study integrated network toxicology prediction, machine learning algorithms (least absolute shrinkage and selection operator [LASSO], support vector machine recursive feature elimination [SVM-RFE], and random forest), and transcriptomic data mining with a 90-day subchronic male Sprague-Dawley (SD) rat exposure model to identify early biomarkers of BDNA-induced nephrotoxicity. Cross-validation of three algorithms identified CD45 (protein tyrosine phosphatase receptor type C, PTPRC) and cyclophilin C (peptidylprolyl isomerase C, PPIC) as high-confidence core targets (training area under the curve [AUC] = 0.951; validation AUCs = 0.958 and 0.929). Single-gene gene set enrichment analysis (GSEA) revealed PTPRC enrichment in immune-inflammatory pathways and PPIC in protein homeostasis-related pathways. Single-cell analysis confirmed their cell type-specific expression patterns. Subchronic BDNA exposure induced dose-dependent renal tubular epithelial injury with significant upregulation of both targets, without interstitial fibrosis or glomerulosclerosis in male SD rats. PTPRC and PPIC may serve as potential early response biomarkers for BDNA-induced renal tubular injury, involving immune-inflammatory activation and proteostasis imbalance. This study establishes an integrative toxicological framework from computational prediction to experimental validation, while acknowledging that causal roles require further functional verification.