FIGURE 1.

Identification of putative biomarkers by using metabolite profiling and multivariate analysis. A: Study design for the dietary intervention study (n = 8). B: Representative ¹H nuclear magnetic resonance (NMR) spectra of urine specimens in response to fruit consumption (red) compared with the standard (STD) meal (black). Apparent differences are highlighted (dashed rectangles). C: Partial least-squares discriminant analysis (PLS-DA) scores plot of urine specimens 0–24 h after fruit challenge, which shows a clear separation of the fruit and STD meals. All urine specimens from the morning of day 1 to the evening of day 2 were allocated to the STD diet, and all urine specimens collected after consumption of the fruit meal (bed time of day 2 until evening of day 3) were allocated to the fruit class. D: Loading plots of the fruit challenge compared with the STD meal indicated the following putative biomarkers for fruit consumption: hippuric acid (δ 2.97d, 7.55t, 7.64t, and 7.84d), proline betaine (δ 2.18m, δ 2.30m, δ 2.50m, δ 3.11, δ 3.31, and δ 3.54), tartaric acid (δ 4.34s), and unknown (δ 7.74d and δ 6.98d). The P value of proline betaine before fruit consumption compared with after fruit consumption was <0.0001. ppm, parts per million; a.u., arbitrary units; Y, response variable (classification identifier); R2Y, variation of Y modeled; Q2Y, cross-validated variation of Y predicted; T[1], first predictive PLS scores vector; Tyosc [1], first orthogonal PLS score vector.