Diagnostic immune-related markers for diabetic kidney disease: a bioinformatics and machine learning approach

Diabetic kidney disease (DKD) is a leading cause of chronic kidney disease, with chronic inflammation driving its progression. This study aimed to identify immune-related diagnostic biomarkers for DKD and explore their association with immune cell infiltration.
Three glomerular transcriptomic datasets (53 DKD, 36 controls) were analyzed via batch-corrected differential expression analysis to screen immune-related differentially expressed genes (DEGs). Machine learning algorithms (least absolute shrinkage and selection operator, support vector machine - recursive feature elimination) prioritized biomarkers, validated by RT-PCR in db/db mice. Immune infiltration was assessed via CIBERSORT and EPIC.
Thirteen immune DEGs were identified, enriched in cytokine signaling and leukocyte chemotaxis. Three biomarkers (albumin (ALB), AP - 1 transcription factor subunit (FOS), and S100 calcium binding protein A9) showed strong correlations with T cell, natural killer cell, and macrophage infiltration, validated by RT-PCR (p < 0.001). Protein-protein interactionnetwork analysis identified ALB and FOS as hub genes (ROC Area Under the Curve: 0.803, 0.795), linking immune dysregulation to glomerular injury.
ALB and FOS serve as novel immunogenetic biomarkers for DKD, highlighting chronic inflammation as a key driver. This framework supports precision immunomodulation, though clinical validation in larger cohorts is needed.