Urinary peptidomics and bioinformatics for the detection of diabetic kidney disease

Abstract

AbstractThe aim of this study was to establish a peptidomic profile based on LC-MS/MS and random forest (RF) algorithm to distinguish the urinary peptidomic scenario of type 2 diabetes mellitus (T2DM) patients with different stages of diabetic kidney disease (DKD). Urine from 60 T2DM patients was collected: 22 normal (stage A1), 18 moderately increased (stage A2) and 20 severely increased (stage A3) albuminuria. A total of 1080 naturally occurring peptides were detected, which resulted in the identification of a total of 100 proteins, irrespective of the patients’ renal status. The classification accuracy showed that the most severe DKD (A3) presented a distinct urinary peptidomic pattern. Estimates for peptide importance assessed during RF model training included multiple fragments of collagen and alpha-1 antitrypsin, previously associated to DKD. Proteasix tool predicted 48 proteases potentially involved in the generation of the 60 most important peptides identified in the urine of DM patients, including metallopeptidases, cathepsins, and calpains. Collectively, our study lightened some biomarkers possibly involved in the pathogenic mechanisms of DKD, suggesting that peptidomics is a valuable tool for identifying the molecular mechanisms underpinning the disease and thus novel therapeutic targets.