Screening key genes for diabetic kidney disease and neutrophil extracellular traps using bioinformatics and machine learning

Objective To screen for characteristic genes associated with diabetic kidney disease (DKD or diabetic nephropathy, DN) and neutrophil extracellular traps (NETs) using bioinformatics and machine learning, and to construct a diagnostic model to provide new insights for the clinical diagnosis of DKD.Methods DKD datasets from the Gene Expression Omnibus (GEO) database were integrated. After data standardization and batch effect correction using R 4.3.2 software, differentially expressed genes were screened and intersected with NETs-related genes from the Genecard database and relevant literature. Key genes were screened using the least absolute shrinkage and selection operator (LASSO) regression, random forest (RF), and support vector machine (SVM) algorithms. Cross-analysis of candidate genes was performed to identify core diagnostic biomarkers, which were validated for their expression and diagnostic performance in independent datasets (GSE157657 and GSE198048).Results A total of 19 DKD and NETs-related differentially expressed genes were identified. After screening by machine learning algorithms, NFIL3, TGFBI, CHI3L1, EGR1, PTGS2, ALB, and VCAN were determined to be core feature genes, which were significantly enriched in cytokine-cytokine receptor pathway and fatty acid metabolism pathways. The diagnostic model constructed achieved an area under the curve (AUC) of 0.996 in the training set and performed excellently in the validation sets GSE104948 (AUC = 1.000) and GSE104954 (AUC = 1.000).Conclusion The neutrophil trap-associated genes NFIL3, TGFBI, CHI3L1, EGR1, PTGS2, ALB, and VCAN identified in this study have diagnostic potential for DKD. They not only provide novel candidate biomarkers for DKD diagnosis but also lay the foundation for further elucidating the molecular mechanisms of NETs in DKD pathogenesis.