Determination of the navigator signals

By performing an ANOVA decomposition^37,38 of each metabolite concentration and pH model, all weighted-important variables (i.e., ¹H NMR chemical shifts) for the construction of each model can be determined, along with their relative statistical importance. According to Fig. 3, the chemical shifts of l-histidine H5 proton (singlet), citrate H3,6 protons (2 doublets of doublets), creatinine H3,6 protons (2 singlets), glycine H_a protons (singlet), l-threonine H4,6 protons (1 doublet and 1 multiplet), 3-methlyhistidine H5 proton (singlet), glycolate H2 protons (singlet), and asparagine H6 protons (2 multiplets) appear to be highly important variables for the construction of almost all concentrations and pH models. Among them, creatinine, citrate, and glycine are the most abundant (Fig. 1b), and their signals are easily identified in a normal urine ¹H NMR profile. The rest of the metabolites signals are less easily detected, as they are frequently overlapped by other signals and/or the variability of their concentrations does not allow for their facile identification (Fig. 1b; Supplementary Table ¹). Therefore, the proton signals of creatinine, glycine, and citrate (five signals in total) were employed as navigator signals.