Exploring blood diagnostic markers for diabetic nephropathy through metabolomics and machine learning

Introduction Diabetic nephropathy (DN) is a specific complication of diabetes accompanied, posing a significant challenge to global public health. However, there are still many deficiencies in clinical detection. In this article, we aimed to identify potential biomarkers and explore the mechanisms underlying diabetic nephropathy (DN) by metabolomics studies and machine learning algorithms.

Methods In this study, blood samples were collected from patients with type 2 diabetes mellitus as well as healthy subjects, and the changes in plasma metabolic profiles were investigated based on the LC-MS/MS technique and the machine-learning approach.

ResultsMetabolomics analysis showed that metabolites changed in DN patients at different stages, and pathway enrichment analysis indicated that such changes were mainly in pyrimidine metabolism, fructose and mannose metabolism, and steroid hormone biosynthesis. Combined with machine learning analysis, a total of three potential biomarkers were screened and identified, includingsorbitol, adipic acid and N-Ribosylhistidine, which were all closely associated with urinary albumin creatinine ratio (UACR) and serum creatinine (SCr). Meanwhile, multiple reaction monitoring (MRM) analysis verified the expression level of the candidate marker sorbitol, which was elevated in patients with DN.

Conclusion This study identified three candidate biomarkers and investigated the pathogenesis of DN, offering novel insights into its diagnosis and prevention.