Fig 1.
Representation of all elements of the experimental design.
The tensor for the complete design consisted of four dimensions: the experimental cases (total 24 –the shaded areas in the figure are for cases 1 and 24), time points (6, hourly), interventions (4, including the vehicle only) and spectral bins (344, following data pretreatment). The total number of 198 144 data points (= 24 x 6 x 4 x 344) thus required bioinformatics analysis to uncover the information from the four interventions. All 24 experimental subjects agreed to participate in each of the four interventions (i.e. a cross-over design), indicated by 1 (vehicle only), 2, 3 and 4. For each of the 24 cases, urine samples (from which the NMR spectral bins were generated) were collected at one hour prior to the intervention (time –1), just before the intervention (time 0) and then at hourly intervals for 4 hours (times 1–4). The results reported in this paper apply to only one of the treatments (consumption of flavored water), which yielded 49 536 data points.
Fig 2.
Flow diagram indicating the main lines of activity following data generation, identification and quantification of important metabolites on the intervention towards the proposed biological interpretation.
VIP refers to variables important in projection, based on a PLS-DA; SSL refers to the sum of the squared loadings of the ASCA model; p refers to the RM-ANOVA p-values for the time effect.
Fig 3.
500 MHz 1H-NMR spectra of urine.
Top spectrum taken from QC sample; numbers indicate the following metabolites: 1: TSP; 2: ethanol; 3: lactic acid; 4: 2-hydroxyisobutyric acid; 5: alanine; 6: acetic acid; 7: pyruvic acid; 8: succinic acid; 9: citric acid; 10: dimethylamine; 11: creatine; 12: trimethylamine-N-oxide (TMAO); 13: creatinine; 14: glycine; 15: guanidinoacetic acid (GAA); 16: glycolic acid; 17: hippuric acid; 18: urea; 19: formic acid. The six important metabolites (high VIP values) are highlighted below to qualitatively show differences over time 0 (black), 1 (blue), 2 (green), 3 (red) and 4 (orange)–scaled according to creatinine. Apart from citric acid and hippuric acid, the detoxification product of benzoic acid, no other constituents from the vehicle appeared to be detectable in the NMR analysis.
Fig 4.
Group separation among experimental groups through dendrograms and Volcano plots based on equidistant binning spectral data.
(a–c) Dendrograms from cluster analysis of 21 cases are shown. The dendrograms were constructed based on subsets of the data representing time 0 along with the data obtained one hour before the intervention (time –1); one hour; and four hours following the intervention (times 1 and 4), respectively. Data for all cases at time 0 are shown as black dots and at the other time slots as red dots. Time points from the same individuals that clustered close together are indicated by squared brackets, linking them. (d–f) Volcano plots for the same time points as in a–c indicate the distribution of the individual bins, based on FC and WC p-values. Demarcation of important bins is shown by the horizontal and vertical dotted lines. The number of influential bins is shown as red dots in the rectangles of the upper left and right segments of each Volcano plot.
Fig 5.
PC1 and PC2 for all individual cases (colored dots) and indication of the 95% confidence ellipsoids for scores of PC1 and PC2 for times –1, 1 and 4 hours. The centroids for these three clusters are indicated as red squares. The trajectories of the spectral profiles of three individuals are also shown. The trajectories of three cases (numbers 9, 7 and 19 are shown as purple, blue and green lines, respectively) illustrate the inter-individual responses to the intervention. The direction of the trajectories is indicated by the short black arrow, starting from time 0, for clarity.
Fig 6.
Application of ASCA to the 344 NMR spectral bins for 21 individuals over the period of the intervention.
Plots of the sum of the ASCA effects and projected residuals are color coded according to time following the intervention (–1, 0, 1, 2, 3 and 4 hours shown in red, black, blue, light blue, green and pink, respectively), with the arrow showing the time-dependent trend, using the same discriminating color sequence.
Fig 7.
Bins which are common to any two statistical approaches are indicated by the lightly shaded areas, and the 18 bins that are common to all three are shown in the central, heavily shaded area.
Table 1.
Quantified data on metabolically important metabolites.
All quantified values included are from NMR-determined urine analyses; the references are all from the Human Metabolome Database, and are expressed as μmoles metabolite/mmole creatinine. WC p-values are based on the comparison of the respective metabolite concentrations from time = 1 hour to 4 hours, relative to time zero.
Fig 8.
Metabolite profile of benzoic acid biotransformation with hippuric acid as outcome.
The metabolites of the primary biotransformation reactions are shown in red. Metabolites of metabolic reactions proposed as being associated with the biotransformation reactions are shown in black. [Metabolites shown in green are presumed to decrease due to increased glycine demand and synthesis of glycine through the reversed glycine cleavage system.] Enzyme nomenclature (names accepted by the IUBMB): EC 1.4.1.2: glutamate dehydrogenase; EC 2.1.1.2: guanidinoacetate N-methyltransferase; EC 2.1.2.1: glycine hydroxymethyltransferase (alt.: serine hydroxymethyltransferase); EC 2.1.2.10: aminomethyltransferase (a glycine synthetase); EC 2.1.3.3: ornithine carbamoyltransferase; EC 2.1.4.1: glycine amidinotransferase; EC 2.3.1.1: amino-acid N-acetyltransferase; EC 2.3.1.71: glycine N-benzoyltransferase (a GLYAT); EC 2.7.3.2: creatine kinase; EC 3.5.3.1: arginase; EC 4.3.2.1: argininosuccinate lyase; EC 6.2.1.2: butyrate-CoA ligase (a medium-chain acyl-CoA ligase); EC 6.3.4.5: argininosuccinate synthase; EC 6.3.5.5: carbamoyl-phosphate synthase (glutamine-hydrolyzing)