3.5. Canonical correlation analysis

3.5. Canonical correlation analysisThe R function cancor was used to calculate the correlation between the NMR signal intensity of soluble metabolites and the content of mineral nutrients in tomato fruits. The canonical vari- ates of each production system were divided into three groups: organic and amino acid, sugar, and aromatic regions. The variates of each group were analysed by Tukey’s HSD test for comparing statistical difference in the canonical correlations among the production systems. The p values of the test are shown in Supple- mental Table 4A–C. The canonical variates of the organic and amino acid group were affected by pesticide application in combination with chemical fertilizers, but not by chemical fertiliz- ers alone or organic fertilizers in combination with pesticides. The fertilizer type affected the canonical variates of sugar groups, but pesticide application had no significant effect on the variates. The canonical correlations between mineral nutrient contents and aromatic substances were significantly influenced by the fertilizer type and the application of pesticides, except in the combined application of chemical fertilizers and pesticides.Fig. 4 shows the correlation between the NMR signal intensity of soluble metabolites and the content of mineral nutrients in tomato fruits. While the application of organic fertilizers resulted in positive correlations between the contents of Mg, P, K, and Fe and the signal intensity of organic and amino acids, the addition of pesticide application lowered the correlations. Conversely, the application of organic fertilizers resulted in negative correlations with the content of Ca, and the addition of pesticide application increased the negative correlations. While the application of chem- ical fertilizers resulted in positive correlations between the con- tents of Mg, P, K, Mn, and Zn and the signal intensity of organic and amino acids, the addition of pesticide application increased the positive correlations with the contents of P, Mn, and Fe, and increased the negative correlations with the content of Ca. The application of organic fertilizers resulted in positive and negative correlations between the Mg, P, and K contents and the signal intensity of sugars, but the addition of pesticide application had almost no effect on these correlations. The correlations between Ca content and the signal intensity of sugars had adverse interac- tions with other minerals. The application of chemical fertilizers and pesticides increased the positive and negative correlations between P, Mn, and Fe contents and the signal intensity of sugars, but had almost no effect on the correlation between Ca and sugars. Large positive correlations between Mn and Fe contents and the signal intensity of each metabolite, except for some sugars (d 3.4–4.0ppm), and negative correlations between Ca and the metabolites were observed in tomatoes cultured with the applica- tion of chemical fertilizers and pesticides.