3.3. Principal components analysis

3.3. Principal components analysis (PCA)PCA is a clustering method that reduces the dimensionality of multivariate data while preserving most of its variance (Eriksson et al., 2006). The mineral nutrients in tomato fruits cultured using organic and chemical fertilizers with and without pesticides were subjected to PCA to outline the differences between the culture sys- tems. As shown in Fig. 2A, the contents of mineral nutrients were generally separated between organic and chemical fertilizer appli- cation by both principal component 1 (PC1) and PC2. This separation occurred in the first two principal components, which cumulatively accounted for 88.06% of the total variation. There was no indication of a separation between pesticide application and no-pesticide application using either organic or chemical fertilizers.Fig. 2B shows the loading scores of the mineral nutrients. The contents of K and Ca in tomato fruits were positively discriminat- ing components of the PC1 scores and negatively discriminating components of the PC2 scores, respectively.Fig. 2C shows the PCA results for soluble metabolites between organic and chemical fertilizer applications with and without pesti- cide applications. The graph (Fig. 2C), which plots PC1 scores versus PC2 scores, clearly discriminates between the metabolic profiles of tomatoes grown with organic versus chemical fertilizer, but no dif- ference was found between pesticide application and no-pesticide application. It also shows the contribution of the soluble metabo- lites to the scores, because the separation only occurred along PC2.The low field regions (d 5.3–9.2 ppm) of the H NMR spectra were extracted from the whole spectra and analysed by PCA, gaining more chemical information in this area (Fig. 2D). Fig. 2E shows the variables with more than 0.1 negative and positive PC2 loading scores of both the whole and low field regions in PCA. Aliphatic pro- tons from organic acids, amino acids, and sugars contributed to the discrimination. A part of sugar region (d 3.6–4.0 ppm) contributed to negative PC2 scores, whereas the amino acid and organic acid regions (d 3.5 ppm) and phenolics regions (d 6.0 ppm) contrib- uted to positive PC2 scores. PC2 scores greater than ±0.2 corre- sponded to d 3.818, 3.386, 7.31, 7.346 and 7.418 ppm.3.4. H– C HSQC spectraTo identify the signals contributing to PC2 scores, 1H–13C HSQC spectra were obtained. Fig. 3A–C show the HSQC spectra of the region of d 5.3–8.2 ppm, the sugar region (d 3.1–5.2 ppm), and the organic and amino acid region (d 3.1 ppm). The signals were assigned by using the PRIMe website (http://prime.psc.riken.jp) and the Human Metabolome Database website (http:// www.hmdb.ca) in Supplemental Table 3. A summary of the identi- fied signals contributing to PC2 loading scores greater than ±0.1 is presented in Table 2. Monosaccharides (glucose and fructose) and disaccharides (sucrose and maltose) affected the PC2 scores of chemical fertilizer applied tomatoes, while amino acids (phenylal- anine, histidine, tyrosine, glutamine, aspartate) and citrate affected the PC2 scores of organic fertilizer applied tomatoes. Chlorogenate,glucuronate and sorbitol also influenced PC2 scores of organic fer- tilizer applied tomatoes.