The data were disposed by multivariate analysis tools (PCA, PLS-DA and OPLS-DA). PCA, PLS-DA and OPLS-DA scoring plots and validation plots of the OPLS-DA models were built for the three contrastive groups: 3 min/CK, 5 min/CK and 3 min/5 min (Fig. 1). All the parameters of these models showed in Additional file 1. R2X represents the cumulative interpretation rate of the multivariate statistical analysis modeling, which generally requires R2X > 0.4, indicating that the model is credible. R2 and Q2 are the parameters of response sequencing test, used to measure whether the model is overfitted. External validation generally requires Q2 < 0 to avoid over-fitting. Internal validation generally requires R2 > 0.5; The closer R2 is to 1, the better the model. In our results, all values for Hotelling’s T2 were 95%, Q2 were < 0 (Q2 = − 0.849 to − 0.741), R2X were > 0.4, and R2 were > 0.5, indicating that the model is reliable.
According to the VIP > 1 for the first principal component in the OPLS-DA, and p-value < 0.05 were the criteria for screening differential metabolites, we identified the differential metabolites among the three contrastive groups [see Additional file 2]. The 119 common differential metabolites from the 3 min/CK and 5 min/CK groups were further divided into three sorts represented by A, B and C, as indicated in Additional File 3 based on STC. Furthermore, the metabolites in these three categories were grouped into eight types of compounds (Table 1). Category A included 68 metabolites. These were mainly lipids, fatty acid and carbohydrates. The contents and expression levels of these metabolites are increased gradually. The metabolites of category B comprised 40 compounds, most of which were fatty acids and lipids. The contents and expression levels of these metabolites decreased with increase of the peeled time. There were 11 components in category C, mainly lipids. Interestingly, their contents and expression levels decreased at 3 min, but are increased at 5 min. Furthermore, the contents of chlorogenic acid that associated with browning increased significantly with longer time from peeling. The changes in the contents of these differential metabolites were similar to the expression level changes shown in the heatmap [see Additional file 4].
KEGG pathway analysis
The metabolites in categories A, B and C were mapped, respectively, using the KEGG database onto the KEGG pathways with the following results (Fig. 2).
The metabolites of category A were mainly enriched in 18 metabolic pathways: Linoleic acid metabolism, Pentose phosphate pathway, Nitrogen metabolism, Glycosylphosphatidylinositol (GPI)-anchor biosynthesis, Autophagy-other, Taurine and hypotaurine metabolism, Galactose metabolism, Aminoacyl-tRNA biosynthesis, Arginine biosynthesis, Arginine and proline metabolism, Alanine, aspartate and glutamate metabolism, C5-Branched dibasic acid metabolism, Glutathione metabolism, Butanoate metabolism, Histidine metabolism, ABC transporters, Glyoxylate and dicarboxylate metabolism, and Glycerophospholipid metabolism. Of these, Glycosylphosphatidylinositol (GPI)-anchor biosynthesis, Linoleic acid metabolism, Autophagy-other and Nitrogen metabolism showed extremely significant differences in the 3 min/CK and 5 min/CK two comparative groups at the p < 0.01 level of significance, while the others revealed a significant difference (p < 0.05). There was a remarkable difference (p < 0.05) in the metabolites of category B that were only enriched in one metabolic pathway, Tropane, piperidine and pyridine alkaloid biosynthesis. Finally, the metabolites of category C were enriched in Tyrosine metabolism and had a remarkable difference (p < 0.05) similarly.