Partial‐least‐squares and canonical‐correlation analysis of chemical constituents and active ingredients of new types of Chinese mulberries

Abstract Objective To investigate the correlation between chemical constituents and active ingredients of 13 types of Chinese mulberry fruits. Methods Thirteen types mulberry fruits were harvested. The correlation between chemical constituents and active ingredients (primarily anthocyanins and rutins) of 13 new types of Chinese mulberries was assessed using partial‐least‐squares, principle‐component and canonical‐correlation analyses. Results Vitamin C and titratable acid in the mulberry fruits were critical components that affected the active ingredients, especially anthocyanins and rutins. The content of titratable acid content was related to the fruit flavor and maintained the balance of anthocyanins, vitamin C and rutins. Mineral elements, such as Zn and Cu, also played a vital role in these processes. Low contents of sugar, crude protein, crude fat and pectin were significantly correlated with the mineral elements. Conclusion Chemical constituents and mineral elements can mutually affect the concentration. It provides a novel method for any changes in the quality of new types of Chinese mulberries, which can identify the sources of new types of natural antioxidants.

Mulberry foliage is used to feed silkworms and the mulberry fruit is widely consumed due to the chemical composition with high biological activity (Ercisli & Orhan, 2007;Natić et al., 2015).
In recent, different data-based multivariate approaches have been explored for biological data analysis, which can explore dependency relationships between data sets. These methods include multiple linear regression, principal component regression, partial least squares, and canonical correlation analysis. Among these methods, partial least squares and canonical correlation analysis play a dominating role probably because the extracted latent variable may contribute the biological interpretations of the results. Partial least squares is first developed for process monitoring in chemical industry, exploits the co-variation between predictor variables and response variables and explores a new set of latent components which are maximally correlated. Canonical correlation analysis is commonly adopted to seek a pair of linear transformations between two sets of variables, and the data are maximally correlated in the transformed space.
At present, the approaches to enhance the quality and improve the nutritional property of mulberry fruit are relatively limited. The detailed composition and the correlation between chemical constituents, minerals and active ingredients of new types of Chinese mulberries should be investigated to explore novel and effective measures and interventions. This study was designed to quantitatively measure the quantity of chemical components, mineral elements and active ingredients, aiming to elucidate the relationship among these ingredients of new types of Chinese mulberries.  (Table 1). Each cultivar fruit was collected from five trees. All selected trees were picked at the biologically ripe stage, planted for more than 10 years, with high yield and without any pest symptoms. The harvest time was between 20 and 25 May 2015. According to shape and color uniformity, berries were randomly harvested from all cardinally-oriented branches with different directions around the canopy. The picked fruits were stored at −18°C for subsequent chemical component analysis.

| Chemical component analysis
The concentrations of the following components including moisture, crude fat, crude protein, reducing sugar, anthocyanins, titratable acid, pectin, vitamin C, rutin, Cu, Fe, Ca, Mg, Zn, K, Se, and Na (AOAC, 1990) were determined. Three replicates were performed for each measurement. Sample moisture was determined by drying at 105°C to constant weight. Crude fat was quantitatively assayed according to the AOAC method using soxhlet extraction with absolute ether as a solvent. Crude protein was measured based upon the AOAC method using a Kjeldahl apparatus. Titratable acidity was determined by titrimetric method. A portion of 5 ml of mulberry cultivar juice was diluted in 50 mL of distilled water and titrating to pH 8.2 using 0.1 mol/L NaOH.
Anthronecolorimetry was used to quantitatively measure the content of the reducing sugar at a wavelength of 630 nm. A pHdifferential method was adopted to determine the anthocyanin content (the total monomeric anthocyanin content [TMAC]) (Lee, Durst, & Wrolstad, 2005;Souza, Pereira, Queiroz, Borges, & Carneiro, 2012). The absorbance value of the extract was determined at a wavelength of 510 and 700 nm at pH = 1.0 and pH = 4.5. TMAC (expressed as cyanidin-3-glucoside) was calculated using the following equations: where MW is the molecular weight of cyanidin-3-glucoside (449 g/ mol), DF is the dilution factor, VE is the extraction volume, ε is the molar extinction coefficient of cyanidin-3-glucoside (29,600), and M is the quantity of extracted berry.
Pectin was determined using pyridine colorimetry (Pang et al., 2012). Ultraviolet chromatometry was used to determine the content of rutin. The 2,4-Dinitrobenzene hydrazine colorimetry was used to analyze the content of vitamin C. For the mineralsubstance analysis, 1 g sample was added to a digestion tank, supplemented with perchloric acid and nitric acid at a ratio of 1:4.The digestion tank was placed in a dry box at 100°C for 1 hr, and then at 130°C for 2 hr, followed by cooling. The sample solution was transferred to a 50 ml volumetric flask and diluted with high-purity water, supplemented with Ca with 10% strontium chloride. A standard curve was generated based on the standard working solution using Cu, Fe, Ca, Mg, Zn, K, Se, and Na. The concentrations of anthrone, glucose, sulfuric acid, sodium hydroxide, hydrochloric acid, cupric sulfate, potassium sulfate, ammonium sulfate, methyl blue, methyl red, hydrogen peroxide, selenium, orthoboric acid, vitamin, C,2,4-dinitrophenylhydrazine, methanol, ethanol, sodium nitrite, aluminum nitrate, activated carbon, citric acid, potassium biphthalate, rutin reference substance, Cu, Fe, Ca, Mg, Zn, K, and Se were quantitatively analyzed by Sinopharm Chemical Reagent, Shanghai, China.

| Statistical analysis
The average values were obtained from three parallel experiments for each type of mulberry fruit. The results are expressed as means ± standard deviation. Data were subject to standardization. The partial-least-squares (PLS) and principal component analysis (PCA) were employed using the Unscrambler software package (Version 9.7; CAMO, Trondheim, Norway).The PLS was used to detect cause-effect relationship, and the correlation coefficient (R 2 ) and root-mean-square error of cross validation (RMSECV) were used to establish a model to evaluate the effect of total phenols, anthocyanins and other related components.
Canonical-correlation analysis was used to analyze the correlation between the mineral elements and chemical components (DPS7.0 software).

| Analysis of main chemical components and mineral elements
As illustrated in Table 2 Table 3.

| Regression analysis of anthocyanins and rutins with other chemical components
The relationship between each chemical component and the rutin  crude fat, and pectin were negatively correlated with anthocyanin.
The quantity of K and Se was weakly negatively correlated with reducing sugar (Figure 4).

| PCA of chemical components
Principal components of eight indices were evaluated (Table 4).
Kaiser-Meyer-Olkin value was 0.691 and the level of significance was 0.093, indicating that these data could be analyzed by PCA.
Based on the value-greater-than-1.0 rule, three principal components were identified by the varimax rotation method, which ex-

| PCA of mineral elements
Principal components of seven indices were analyzed in Table 5.
Kaiser-Meyer-Olkin value was 0.662 and the level of significance was 0.000, suggesting that these data could be analyzed by PCA. Based on the eigenvalue-greater-than-1.0 rule, two principal components were identified by the varimax rotation method, which accounted for 86.871% of the general data of the sample. The first principal com-

| Canonical-correlation analysis of chemical components and mineral elements
From the PCA, three chemical principal components including the activity factor (f1),the reducing sugar and crude protein (f2) and the crude fat and pectin (f3) were identified. Two mineral principal components of Ca plus three other mineral elements (f4) and Cu plus two other mineral elements (f5) were obtained. The canonical-correlation analysis between chemical components and mineral elements of these five principal components was performed (Table 6).

| Correlation coefficient matrix
By maximizing variance varimax rotation in the PCA, the correlation index of each principal ingredient from the same group was zero, suggesting that no principal ingredients could be substituted. The principal f1 of chemical components was negatively correlated with that of mineral elements, indicating that the mineral elements exerted a weak inhibitory effect on vitaminC, anthocyanin, rutins and titratable acid. The f2/f3 was positively correlated with f4, suggesting that reducing sugar, crude protein, crude fat, and pectin were significantly correlated with mineral elements (Table 6).

| Canonical-correlation-coefficient analysis
The first canonical variable was qualified the significance test (α = 0.01) with a canonical-correlation coefficient of 0.859, whereas the second canonical variable failed to achieve the significance test (α = 0.01) with a canonical correlation of 0.401. A canonical-correlation coefficient existed between chemical components and mineral elements, which was validated by analyzing the first canonical variables (Table 7).

| Canonical-correlation-structure analysis
To analyze the relative effects of principal components between two groups when they formed a canonical variable, it was necessary to observe the first canonical variable: m1 = 0.113f1 + 0.640f2 + 0.247f3.

| D ISCUSS I ON
In this investigation, the new type of Chinese mulberry fruits possesses high nutritional value, whereas the chemical components and mineral ingredients significantly differ among types.
The content of rutins in dark Hongguo1 mulberries reached up to 0.32 mg/g, which is higher than the value previously reported (Ercisli & Orhan, 2007). The quantity of anthocyanin contained in Hongguo 2 mulberries was 0.34 g/100 g, whereas it F I G U R E 5 Canonical-correlation analysis of chemical components and mineral elements was undetectable in Baishen mulberries (light mulberries). The quantity of the rutins and anthocyanin contained in these mulberries presents with identical changing pattern, which is consistent with previous findings (Donno et al., 2015;Ercisli & Orhan, 2007).
Appropriate understanding of the chemical composition of mulberry ruits can explore and identify novel resources of natural antioxidants. (Kara & Erçelebi, 2013;Wang, Xiang, Wang, Tang, & He, 2013). PCA is primarily used as a tool in exploratory data analysis and for establishing predictive models, and for visualizing genetic distance and relatedness between populations. PCA can be performed by eigenvalue decomposition of a data covariance matrix or singular value decomposition of a data matrix. In this study, PCA demon- Canonical-correlation analysis demonstrated that multiple ingredients, such as reducing sugar, crude protein, crude fat, and pectin are intimately correlated with mineral elements, and play a coordinated role. However, relatively weak correlation is observed between mineral elements and the accumulation of active ingredients. Se and Cu exert a significant effect upon the function of active ingredients.
Besides the genetic and physiological influence, chemical components and nutrition of the mulberry fruits are probably affected by the environmental factors, such as the soil chemical properties and climatic conditions, agronomic conditions including the harvesting techniques during different stages of maturity and technical factors, such as disposal after harvesting and conditions for processing and storage (Donno et al., 2012(Donno et al., , 2012Sadia et al., 2014). Taken together, our findings add evidence to the modification of farming methods, aiming to improve the quality and nutritional components of mulberry fruits. Moreover, the findings obtained from this investigation offers reference for the selection of the mulberry variety.

ACK N OWLED G M ENTS
The authors thank the staff from the planting and cultivation base of Xia Jin County for their assistance and support.

CO N FLI C T O F I NTE R E S T
None declared.