Unveiling urinary diagnostic biomarkers for diabetic kidney disease using metabolomics and machine learning approaches.

Aims: Diabetic kidney disease (DKD) is a specific complication of diabetes that poses a major challenge to global public health. However, clinical detection of DKD still has notable limitations. This study aimed to identify potential biomarkers and explore the underlying mechanisms of DKD.

Materials and methods: Urine samples were collected from patients with type 2 diabetes mellitus and healthy subjects. Changes in urine metabolic profiles were analysed via liquid chromatography-tandem mass spectrometry combined with a machine learning approach.

Results: Metabolomics revealed characteristic metabolite alterations at different stages of DKD progression, and pathway enrichment analysis revealed significant changes in pathways such as biotin metabolism and taurine and hypotaurine metabolism, among which biotin and taurine are the key regulatory molecules of these pathways. Combined screening with two machine learning algorithms finally identified five differentially expressed metabolites: hypoxanthine, N-acetyl-DL-histidine, cortisol, tetrahydrobiopterin and L-kynurenine. Correlation analysis coupled with receiver operating characteristic curve validation showed these seven biomarkers were significantly correlated with clinical indicators (urinary albumin creatinine ratio, serum creatinine) and had early diagnostic value. Notably, multiple reaction monitoring validation revealed taurine and hypoxanthine expression exhibited DKD stage-dependent characteristics.

Conclusion: This study initially identified seven DKD biomarkers with excellent diagnostic performance and further clarified their potential role in mediating DKD progression via mechanisms involving biotin metabolism, taurine and hypotaurine metabolism and steroid hormone biosynthesis. These findings provide important insights for the early precise diagnosis and mechanistic exploration of DKD.