Establishment of gene diagnosis model of diabetic kidney disease based on bioinformatics / 国际生物医学工程杂志

ABSTRACT

Objective:To explore the potential biomarkers and pathogenesis of diabetic kidney disease (DKD) using bioinformatics techniques.Methods:Three DKD glomerular tissue datasets, GSE30528, GSE96804, and GSE104948, were downloaded from the gene expression database. The bioinformatics techniques, combined with machine learning, were applied to screen DKD biomarkers and establish a DKD gene diagnostic model. The CIBERSORT algorithm was applied to analyze immune cell infiltration in DKD glomerular tissue. The correlation between biomarkers and immune cells, model risk scores, and the glomerular filtration rate (eGFR) of patients was investigated.Results:A three-gene model consisting of G6PC, CDH10 and TPPP3 was screened and established from 26 differentially expressed genes, which showed good diagnostic efficacy in both the training set (AUC=0.984) and validation set (AUC=0.992). Analysis of infiltrating immune cells in DKD glomerular tissue showed that γδ T cells, activated NK cells, M2 macrophages, resting dendritic cells, resting mast cells, activated mast cells, and neutrophils may be involved in the DKD process, and that G6PC, CDH10 and TPPP3 were associated with multiple immune cell infiltrations. Using the model to calculate the risk score of DKD patients, the higher the score, the lower the eGFR level, and the risk score was significantly negatively correlated with the eGFR level. Conclusions:G6PC, CDH10 and TPPP3 can be good diagnostic markers for DKD, correlate with DKD immune cell infiltration, and be used as targets for DKD diagnosis and treatment.