Figure 1.
Metabolomic profiling of spot urines from SU.VI.MAX2 subjects.
Subjects reported either low or high consumption of coffee, represented by squares and circles respectively. A) One-dimensional OSC-PLS-DA score plot of urinary metabolomes of low and high consumers. B) Loading plot of the OSC-PLS-DA. Circled outlying ions contribute most strongly to the discrimination. C) Model validation assessed by permutation test (n = 100).
Figure 2.
Chemical structures of some identified discriminants.
Table 1.
The strongest contributors to the discrimination of low and high coffee consumers.
Figure 3.
ROC curve analysis of atractyligenin glucuronide and caffeine.
Data for atractyligenin glucuronide are presented in the left-hand column and data for caffeine in the right-hand column. A) Blue curves represent the training set (n = 39 subjects) and pink curves the hold-out set (n = 20 subjects). B) Probabilities of predicted belonging to the high consumer class. Training set, black plots; hold-out set, red plots; filled circles, high consumers; empty circles, low consumers. C) Confusion matrices for the two datasets.
Figure 4.
ROC curve AUCs for single and combination biomarkers.
Error bars represent 95% confidence intervals. cIP, cyclo(isoleucyl-prolyl); MX, 1-methylxanthine; Tr, trigonelline; Atr, atractyligenin glucuronides; Caf, caffeine.