Identification and validation of immune and cuproptosis - related genes for diabetic nephropathy by WGCNA and machine learning.

BACKGROUND

As the leading cause of chronic kidney disease, diabetic kidney disease (DKD) is an enormous burden for all healthcare systems around the world. However, its early diagnosis has no effective methods.

METHODS

First, gene expression data in GEO database were extracted, and the differential genes of diabetic tubulopathy were obtained. Immune-related genesets were generated by WGCNA and immune cell infiltration analyses. Then, differentially expressed immune-related cuproptosis genes (DEICGs) were derived by the intersection of differential genes and genes related to cuproptosis and immune. To investigate the functions of DEICGs, volcano plots and GO term enrichment analysis was performed. Machine learning and protein-protein interaction (PPI) network analysis helped to finally screen out hub genes. The diagnostic efficacy of them was evaluated by GSEA analysis, receiver operating characteristic (ROC) curve, single-cell RNA sequencing and the Nephroseq website. The expression of hub genes at the animal level by STZ -induced and db/db DKD mouse models was further verified.

RESULTS

Finally, three hub genes, including , and that were up-regulated in both the test set GSE30122 and the validation set GSE30529, were screened. The areas under the curve (AUCs) of ROC curves of hub genes were 0.911, 0.935 and 0.922, respectively, and 0.946 when taking as a whole. Correlation analysis showed that the expression level of three hub genes demonstrated their negative relationship with GFR, while those of displayed a positive correlation with the level of serum creatinine. GSEA was enriched in inflammatory and immune-related pathways. Single-nucleus RNA sequencing indicated the main distribution of in podocyte and mesangial cells, the high expression of in leukocytes and the main localization of in the loop of Henle. In mouse models, all three hub genes were increased in both STZ-induced and db/db DKD models.

CONCLUSION

Machine learning was combined with WGCNA, immune cell infiltration and PPI analyses to identify three hub genes associated with cuproptosis, immunity and diabetic nephropathy, which all have great potential as diagnostic markers for DKD and even predict disease progression.