Integrative network toxicology and single-nucleus transcriptomics reveal molecular overlap between predicted 6PPD-quinone targets and diabetic kidney disease

Abstract Purpose The renal implications of the tire-derived contaminant 6PPD-quinone (6PPD-Q) are not well defined. We conducted a hypothesis-generating computational study to identify molecular overlap between predicted 6PPD-Q activity and diabetic kidney disease (DKD). Methods Predicted human targets of 6PPD-Q were integrated with DKD-associated differentially expressed genes and co-expression modules. Functional enrichment, training-restricted machine learning with SHAP interpretation, immune deconvolution, WGCNA sensitivity analysis, virtual knockout, and single-nucleus transcriptomic analyses were performed. The final model was locked and evaluated once in two external datasets (GSE30528 and GSE104948). Single-nucleus disease-group comparisons were conducted at the donor level, with donors rather than individual nuclei treated as biological replicates. Results Intersection of 1,313 predicted 6PPD-Q targets, 2,326 DKD differentially expressed genes, and 936 genes from positively DKD-associated co-expression modules identified 44 candidate genes enriched in mitochondrial, metabolic, xenobiotic-processing, and inflammatory pathways. Inclusion of the strongly inversely associated brown module in sensitivity analysis added only OGT, while retaining FOS, UQCR10, CTSV, and CXCR4. The locked logistic model achieved AUCs of 0.675 (95% CI, 0.427–0.889) in GSE30528 and 0.815 (95% CI, 0.609–0.962) in GSE104948. Benchmarking showed that the candidate genes did not consistently outperform established DKD-associated genes. CXCR4 exhibited marked immune-cell specificity (Tau = 0.99). Conclusion These analyses identify a hypothesis-generating molecular overlap between predicted 6PPD-Q targets and DKD-associated alterations and nominate a putative CXCR4-associated immune-chemotaxis axis for further investigation.