Predictive metabolomic signatures of end-stage renal disease: A multivariate analysis of population-based data

The variability of molecular signatures and predictive low molecular weight markers of chronic kidney disease (CKD) in different populations are poorly understood. Thus, in a large sample with 4763 people we compare the molecular signatures and metabolites with diagnostic relevance in plasma and urine of CKD patients of different geographical origins. From an integrated model based on dynamic networks and multivariate statistics, metabolites with predictive value obtained from targeted and untargeted molecular analysis, interactions between metabolic pathways affected by CKD, and the methodological quality of metabolomic studies were analyzed. The metabolites 3-methylhistidine, citrulline, kynurenine, p-cresol sulfate, urea, and citrate presented consistent expression in all population groups. Only increased kynurenine and p-cresol sulfate in plasma samples obtained acceptable scores as CKD biomarkers, independent of geographic origin. Metabolites such as leucine, alanine, isoleucine, serine, histidine, and citrate were nodal points, indicating that protein metabolism pathways are similarly impaired in Asian, European and North American patients. Based on our integrated model, we show that the metabolome of CKD patients exhibits a strong geographic influence, leading to unique metabolic signatures. Contrary to the likelihood of molecular similarities between geographically distinct populations, metabolic convergences in protein metabolism pathways and the molecules kynurenine and p-cresol sulfate were relevant as general predictors of CKD. In general, the quality assessment indicated that the current evidence is based on research models with variable methodological quality, whose limitations described in this study should be considered in the refinement of molecular approaches.