Ferroptosis-Related Diagnostic Biomarkers in Diabetic Kidney Disease: A Multi-Omics and Multi-Cohort Study

作者：Li X, Hong H, Li H, Fang Z, Wang J, Wang X, Sui X, Shadekejiang H, Liang M, Gan X, Liu J, Lu C · 发表于：Journal of inflammation research · 年份：2026 · DOI：10.2147/jir.s603085 · 研究领域：DKD、TYRO3、diabetic kidney disease、ferroptosis、machine learning、single-cell RNA sequencing

BACKGROUND: Diabetic kidney disease (DKD) is one of the primary factors leading to end-stage renal disease. Ferroptosis, as a mechanism-related form of programmed cell death, has garnered increasing attention in DKD research. This study aimed to identify and validate ferroptosis-related diagnostic biomarkers for DKD. METHODS: Public DKD transcriptomic datasets were analyzed using differential expression analysis, WGCNA, and multiple machine-learning algorithms to identify key ferroptosis-related genes and construct a logistic regression diagnostic model. Candidate biomarkers were validated in Nephroseq V5 platform, a real-world clinical cohort by ELISA, a human kidney single-cell RNA-seq dataset, and an STZ-induced DKD rat model. RESULTS: Thirty-two ferroptosis-related module-specific differentially expressed genes were identified, from which COL14A1, ACADSB, TYRO3, and ZFP36 were selected as key biomarkers. The four-gene diagnostic model showed strong discriminatory performance in the training dataset and maintained diagnostic value in three independent validation datasets. In the real-world cohort, creatinine-corrected urinary levels of the four corresponding proteins were positively correlated with urinary albumin-to-creatinine ratio and negatively correlated with estimated glomerular filtration rate, and the combined urinary predictor showed high diagnostic performance for DKD. Single-cell analysis revealed cell-type-specific expression patterns, including COL14A1 enric...