2022 Volume 70 Issue 12 Pages 863-867
Apricot and Peach Kernels are commercial crude drugs used in many formulas of traditional Japanese medicine, Kampo. Although their applications are quite different, it is difficult to distinguish them using conventional methods such as HPLC. The study aimed at near-infrared (NIR) metabolic profiling to discriminate Apricot and Peach Kernels (Armeniacae Semen and Persicae Semen) collected from Japanese markets. A fast, simple, non-destructive, and robust NIR measurement of kernel surface with no sample pre-treatment was achieved in situ. Principal component analysis and orthogonal partial least squares discriminant analysis (OPLS-DA) models showed discrimination between the two crude drugs with good fitting and prediction values. These results indicate that NIR metabolic profiling is useful for discriminating Apricot and Peach Kernels based on their chemical constituents using a simple and non-destructive procedure.
Apricot Kernel (Armeniacae Semen, Kyounin in Japanese) is defined as the seed of P. armeniaca Linneá, P. armeniaca Linneá var. ansu Maximowicz, or P. sibirica Linneá.1) Peach Kernel (Persicae Semen, Tounin in Japanese) is defined as the seed of Prunus persica Batsch or P. persica Batsch var. davidiana Maximowicz (Rosaceae). Apricot Kernel is used to treat asthma, coughing, dyspnea, and body edema and is found in many traditional Kampo formulas, such as Mao-to and Makyoukannseki-to. On the other hand, Peach Kernel is used to treat “Oketsu,” a blood stasis pattern, and it is often included in many traditional Kampo formulas, such as keishibukuryogan and tokakujokito. Thus, the clinical applications of the two crude drugs are quite different.2–4) In traditional Japanese Kampo medicine, Yoshimasu, in his textbook “Yakucho,”5) describes Apricot Kernel mainly curing cough, phlegm, and asthma with abdominal distention and Peach Kernel mainly curing blood stagnation, lower abdominal distention, and pain. The two crude drugs are slightly different in morphology. In Apricot Kernel, numerous vascular bundles run from chalaza throughout the seed coat, appearing as thin vertical furrows, and at its terminus, it branches into scattered shapes. Whereas in Peach Kernel, vascular bundles rarely branch from chalaza through the seed coat and appear as dented longitudinal wrinkles and terminate at its terminal end with few branches as well.1,6) Therefore, it is not easy to distinguish between the two crude drugs, so chemical analytical identification methods are required.
Both crude drugs’ chemical constituents, mainly cyanogenic glycosides, amygdalin, and fatty acids, are similar. Hence, HPLC provides chromatograms with similar patterns that make discrimination between the two crude drugs difficult.7)
However, there is a known difference in amygdalin and scopoletin content between the two crude drugs. In other words, the description of Apricot Kernel properties in the Japanese Pharmacopoeia states that it “tastes bitter,” while Peach Kernel is said to have a “slightly bitter” taste. Japanese Pharmacopoeia also states that Peach Kernel contains at least 1.2% amygdalin, while Apricot Kernel contains at least 2.0% amygdalin. Therefore, Apricot Kernel is known to have a bitterer taste.1)
Chemical analyses play an important role in these practices. Analytical techniques, such as LC, gas chromatography (GC), Fourier-transform infrared spectroscopy (FT-IR), near-infrared (NIR) spectroscopy, TLC, MS, NMR, and their serial or parallel combinations, have been successfully employed to improve and ensure the quality of botanical health products.8–11)
Vibrational spectroscopy techniques, such as NIR spectroscopy,12–14) provide complementary information on the molecular structure and enable in situ measurement with no sample pre-treatment. NIR has the advantage of being extremely fast, easy-to-measure, robust, and needing no sample alteration. In this way, NIR has unique properties as a non-destructive technique providing simpler and faster analytical procedures. The main drawback of NIR is the wide nonselective bands of the spectral profile, so there are very few studies on crude drugs with NIR.12,15) With the evolution of chemometrics (multivariate data analysis and modeling), NIR is gaining strong acceptance in food science and technology. Therefore, a detailed study showing the relationship between compounds and bitterness has been reported in NIR quality assessment of almonds of the same genus as the above two crude drugs by measuring the seed coat with no sample pre-treatment.16) It was reported that peach and apricot kernels were identified by NIR using one lot of each these crude drugs for comparison.7)
NIR has been used as a comprehensive and straightforward method for analyzing herbal medicines and foods, and the differences in amygdalin content and fatty acids are expected to be easily discriminated due to the quantitative nature of NIR. Analyses for quality control of crude drugs are necessary to collect a wide variety of samples.
This study reports the development of a method for discriminating Peach and Apricot Kernels as commercial crude drugs collected from Japanese markets by combining NIR spectroscopy and multivariate analysis.
Ten specimens of Apricot and Peach kernels were analyzed. Five samples from a large number of specimens were measured. Detailed information on the plant materials is provided in Table 1.
| Lot No. | NIR No. | Crude Drug | Obtained from* | Code No. | Production area | Collection year | |
|---|---|---|---|---|---|---|---|
| ● | 1 | 1–5 | Apricot Kernel | National Institute of Biomedical Innovation, Japan | NIB0272 | Shanxi, China | Unknown |
| ▲ | 2 | 6–10 | Apricot Kernel | National Institute of Biomedical Innovation, Japan | NIB0425 | Shanxi, China | 2011 |
| ■ | 3 | 11–15 | Apricot Kernel | National Institute of Biomedical Innovation, Japan | NIB0426 | Shaanxi, China | 2009 |
| ◆ | 4 | 16–20 | Apricot Kernel | National Institute of Biomedical Innovation, Japan | NIB0452 | Guizhou, China | 2012 |
| ▼ | 5 | 21–25 | Apricot Kernel | National Institute of Biomedical Innovation, Japan | NIB0744 | Ganzhou, China | 2018 |
| ● | 6 | 26–30 | Peach Kernel | National Institute of Biomedical Innovation, Japan | NIB0271 | Hebei, China | Unknown |
| ▲ | 7 | 31–35 | Peach Kernel | National Institute of Biomedical Innovation, Japan | NIB0428 | Shaanxi, China | 2009 |
| ■ | 8 | 36–40 | Peach Kernel | National Institute of Biomedical Innovation, Japan | NIB0439 | Hebei, China | 2009 |
| ◆ | 9 | 41–45 | Peach Kernel | National Institute of Biomedical Innovation, Japan | NIB0731 | Hebei, China | 2010 |
| ▼ | 10 | 46–50 | Peach Kernel | Meishi Inc. | C17033601 | Sichuan, China | 2012 |
* Sample information obtained from National Institute of Biomedical Innovation, Japan (Nibio) was sited in the database website; Comprehensive Medicinal Database (http://mpdb.nibiohn.go.jp/).
A set of 50 Apricot and Peach kernel samples was measured on the seed coat. The corresponding raw spectra are shown in Fig. 1a. The spectra were pre-processed using standard normal variate (SNV) before principal component analysis (PCA) (Fig. 1b).
Detailed peaks were assigned to the metabolites by comparing with the pure compounds, as shown in Table 2. These assignments included cyanogenic glycosides, amygdalin, and a fatty acid, stearic acid. The NIR spectra of amygdalin and palmitic acid, as well as NIR No. 1 as Apricot Kernel and NIR No. 26 as Peach Kernel. are shown in Fig. 2. In the NIR spectrum, the main peaks were assigned to a cyanogenic glycoside, in which the CN bond of amygdalin was identified at 4379 cm−1 (Fig. 2, I). In contrast, 5665 and 5775 cm−1 fatty acid peaks might be attributed to the second overtone of the C–H stretch (Fig. 2, II).
| Compound | NIR (cm−1) |
|---|---|
| Amygdalin | 3902, 3991, 4231, 4379, 4670, 4760, 5206, 5978, |
| Stearic acid | 4250, 4325, 5665, 5775, 7175, 8225 |
The peaks follow those listed in Table 2 for metabolite identification using NIR.
To aims to reduce the dimensionality of a multivariate dataset, principal component analysis (PCA) is a widely applied unsupervised method in multivariate data analysis.13) Data points derived from the NIR spectra (3900–10000 cm−1) of 50 samples were pre-processed by Standard Normal Variate (SNV) and centering scaling method. PCA extracted three significant principal components cumulatively accounting for 91.5% of the total variance (PC1 = 74.8%, PC2 = 12.1%, PC3 = 4.5%), which is shown in a 3D score plot, which proposed a slight differentiation between the two crude drugs (Fig. 3a). A 2D-PCA (principal components PC1 and PC2) score plot is shown in Fig. 3b to understand it in detail. The first component (PC1) was mainly responsible for the separation. The combination of PC1 and PC2, showed good separation of the two crude drugs. The PCA loading line plot (Fig. 3c) was used to identify the differentiation of these chemical constituents between the two crude drugs. The loading line plot (Fig. 3c) shows the main regions differentiating between the two crude drugs. PC1 shows peaks at 4260, 4745, and 5186 cm−1 in the positive region, which could be related to C–H bonds and closely identical to peaks of cyanogenic glycosides shown in Table 2. On the other hand, PC1 negative regions exhibit peaks at 5680 and 5785 cm−1, which could correspond to the C–H links of fatty acids shown in Table 2.7,16)
(a) 3D score plot of principal components PC1, PC2, and PC3 scores, (b) score plot of PC1 and PC2 scores, (c) loading line plot for PC1 and PC2.
To validate the PCA model, OPLS-DA was performed to demonstrate outstanding discrimination between Apricot and Peach Kernels to show the score plot as Fig. 4a, where the x-axes (t[1]) corresponds to between the two crude drugs and the y-axes (to[1]) to within variability of each crude drug. OPLS-DA was performed on the PCA models to add two classes of Apricot and Peach Kernels as objective variables, checked the multiple correlation coefficient (R2Y) and predictive ability parameter (Q2) of the resulting model, of which model presented a good fit (R2Y = 0.89) and prediction (Q2 = 0.73), indicating the model’s linearity and predictive ability, respectively, to the response using two components: Apricot and Peach Kernel. With R2 > 0.65 and Q2 > 0.5, the model is regarded as adequate.17)
The x-axes (t[1]) corresponds to between the two crude drugs and the y-axes (to[1]) to within variability of each crude drug. t; the score vector for discrimination between apricot and peach kernel. (b) S-line plot for the class separation to visualize the p1(ctr) lording colored according to the absolute value of p(corr).
A permutation test was performed17) (Fig. 5) to validate the incidence of overfitting in the prediction model. In this test, the provisional prediction models are constructed based on various data matrices in which objective and explanatory variables are combined randomly many times to calculate the R2 and Q2 of each provisional model. The correlation coefficients between the original and permuted data matrices versus both R2 and Q2 are plotted on the x- and y-axes, respectively. The y-intercept of the regression line in the plot was used to estimate the index of overfitting. In general, R2 < 0.3 and Q2 < 0.05 are the standard values.18,19) The test calculated R2 = 0.205 and Q2=−0.438 for Apricot Kernel and R2 = 0.228 and Q2=−0.534 for Peach Kernel, which are within the standard values.
The number of permutations to 50 times was performed in one component. The y-intercept of the regression line in the plot was used to estimate the index of overfitting.
The S-line plot is tailor-made for NIR spectroscopy data showing the wavenumbers responsible for the class separation. This plot visualizes p1(ctr), the lording of the score vector t to discriminate two crude drugs, colored according to the absolute value of p1(corr). The p1(corr) is the normal loading re-expressed as the correlation (running between −1 and +1) between each X-variable and the score vector t. The S-line plot thus displays the predictive lording plot in a form resembling the original spectra, colored according to p(corr)18) (Fig. 4b). Apricot Kernel (positive values) is mainly explained by peaks at 4132 cm−1 (I, p(corr) = 0.800), which can be assigned as a C–N triple bond and is not found in the PCA loading plot. For positive values of Apricot Kernel, the peak at 4737 cm−1 (II, p(corr) = 0.775) corresponds to the O–H bond/C–O stretch combination. Peach Kernel (negative values) is mainly described by peaks at 5683 cm−1 (III, p(corr) = 0.597) and 5866 cm−1 (IV, p(corr) = 0.598) for the C–H stretch second overtone, 8353 cm−1 (V, p(corr) = 0.779), and 8393 cm−1 (VI, p(corr) = 0.785) for the 2nd overtone. Peaks in the positive values correspond to cyanogenic glycosides, while those in the negative values correspond to fatty acids.7,16) It was reported that between the two crude drugs, Apricot Kernel contains more cyanogenic glycosides, while the contents of fatty acids are similar.20) Further investigation of fatty acids is required, such as GC or HPLC analysis.
The S-line curve could be similar to the PC1 loading curve, which suggests that the PCA model is a good fit for discrimination between these crude drugs. However, for their classification and quality control with a wide variety of medicinally used samples, OPLS-DA is suitable. Further sample collection and NIR analysis are essential to improve the accuracy of this discriminant model.
In conclusion, We performed NIR metabolic profiling to discriminate Apricot and Peach Kernels as commercial crude drugs collected from Japanese markets. A fast, simple, non-destructive, and robust measurement of NIR was achieved by the in situ measurement of the kernel surface without any sample pre-treatment. PCA and OPLS-DA models presented discrimination between two crude drugs with good fitting and prediction values. These results indicated that NIR metabolic profiling is useful for discriminating Apricot and Peach kernels based on their chemical constituents with the simple and non-destructive procedure.
Ten specimens of Apricot and Peach kernels were analyzed. Five samples from a large number of specimens were measured. Detailed information on the plant materials is provided in Table 1.
General Experimental MethodsStandards for apricot and peach kernels, amygdalin, and stearic acid were obtained from FUJIFILM Wako Pure Chemical Corporation (Osaka, Japan).
NIR AnalysisSpectral reflectance NIR measurements were collected using a VIR-200, an SC-100-VIR, and an RF-SC-VIR (JASCO Co., Tokyo, Japan) equipped with an indium gallium arsenide (InGaAs) detector, a reflection accessory, and Spectra Manager Version 2.15.01 (JASCO Co.). Representative surfaces of the apricot and peach kernels with skin were directly deposited on the reflection measurement areas. Spectra were recorded in the reflectance mode from 10000 to 3900 cm−1, at a 4 cm−1 resolution and 64 scans per sample. The absorbance was computed against the background spectrum measured using a standard white plate. A mixture of standards in KBr powder was measured under the same conditions as above, and the absorbance was computed against a background spectrum measured by a standard KBr powder.
Data AnalysisA final matrix of 50 samples and 6328 variables was built from the NIR spectra. The collected spectra were imported into SIMCA software (version 14, Umetrics Co., Umeå, Sweden) and pre-processed by SNV and centering scaling for multivariate analyses, PCA, and OPLS-DA.
This work was supported by The Japan Science and Technology Agency’s Center of Innovation Program (JST COI, Grant number JPMJCE1301), and AMED (Grant No. JP20ak0101104).
All authors declare that: (I) no support, financial or otherwise, has been received from any organization that may have an interest in the submitted work; and (II) there are no other relationships or activities that could appear to have influenced the submitted work.
This article contains supplementary materials.
