A Predictive Model for Diabetic Kidney Disease Based on Inflammatory Gene Signatures and Its Regulatory Network

Background: Diabetic kidney disease (DKD) is a leading cause of end-stage renal disease (ESRD) globally, characterized by increased albuminuria and reduced glomerular filtration rate. Recent evidence points to inflammation as a vital contributor to the development and progression of DKD, involving interactions among immune cells, cytokines, and chemokines. Our study focused on uncovering inflammation-related genes in DKD to understand its mechanisms and developed an inflammation-centric predictive model. We aimed to bridge molecular insights with immune interactions, paving the way for innovative treatments. Methods: This study involves comprehensive data collection from gene expression omnibus (GEO) datasets (GSE1009 and GSE30528) to identify differentially expressed genes (DEGs) between patients with DKD and healthy controls (HC). Using the ComBat method for batch effect removal, R package Limma for DEGs identification, and Metascape for enrichment analysis, we focused on the interplay between inflammation-associated genes and immune cell infiltration. We developed a predictive model for DKD using the least absolute shrinkage and selection operator (LASSO) regression, centered on six potential candidate genes: chitinase-3-like protein 1 ( CHI3L1 ), coagulation factor V ( F5 ), decay-accelerating factor ( CD55 ), insulin-like growth factor 1 ( IGF1 ), vascular endothelial growth factor A ( VEGFA ), and 15-hydroxyprostaglandin dehydrogenase ( HPGD ), within a training cohort. This model was subsequently validated in a test cohort utilizing data extracted from the GEO dataset GSE96804. Immune cell infiltration was determined using CIBERSORT, followed by Pearson correlation analysis to elucidate the interactions between hub genes, immune cells, and chemokines. Results: We identified 349 DEGs, including 99 upregulated and 250 downregulated genes, highlighting the significant role of inflammation in DKD. Through weighted gene co-expression network analysis (WGCNA), a module consisting of 784 genes strongly associated with DKD was identified. Within this module, six inflammatory-related genes were identified as crucial for the predictive model, achieving an area under the receiver operating characteristic curve (AUC) of 1 in training and 0.76 in validation. Analysis of immune cells revealed significant differences between DKD patients and controls, while Pearson correlation analysis highlighted key associations with immune infiltration and regulation. Conclusions: Our study provides novel insights into the genetic and inflammatory landscape of DKD, establishing a predictive model with high accuracy compared to existing models. We pinpoint significant correlations between hub genes and immune cell dynamics, potentially opening avenues for new therapeutic strategies. Our findings underscore the promise of precision medicine in diagnosing and treating DKD.