Next Article in Journal
Studies on Upgradation of Waste Fish Oil to Lipid-Rich Yeast Biomass in Yarrowia lipolytica Batch Cultures
Next Article in Special Issue
Effect of Drying Methods on Volatile Compounds of Burdock (Arctium lappa L.) Root Tea as Revealed by Gas Chromatography Mass Spectrometry-Based Metabolomics
Previous Article in Journal
Carotenoid and Carotenoid Ester Profile and Their Deposition in Plastids in Fruits of New Papaya (Carica papaya L.) Varieties from the Canary Islands
Previous Article in Special Issue
NMR Characterization of Ten Apple Cultivars from the Piedmont Region
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Foods
Volume 10
Issue 2
10.3390/foods10020435
foods-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Discrimination of the Geographical Origin of Soybeans Using NMR-Based Metabolomics

by
Yaoyao Zhou
1,†,
Seok-Young Kim
1,†,
Jae-Soung Lee
1,
Byeung-Kon Shin
2,
Jeong-Ah Seo
3,
Young-Suk Kim
4,
Do-Yup Lee
5,* and
Hyung-Kyoon Choi
1,*
1
College of Pharmacy, Chung-Ang University, Seoul 06974, Korea
2
National Agricultural Products Quality Management Service, Gimcheon 39660, Korea
3
School of Systems Biomedical Science, Soongsil University, Seoul 06978, Korea
4
Department of Food Science and Engineering, Ewha Womans University, Seoul 03760, Korea
5
Center for Food and Bioconvergence, Department of Agricultural Biotechnology, Research Institute for Agricultural and Life Sciences, CALS, Seoul National University, Seoul 08826, Korea
*
Authors to whom correspondence should be addressed.
†
These authors contributed equally to this work.
Foods 2021, 10(2), 435; https://doi.org/10.3390/foods10020435
Submission received: 25 January 2021 / Revised: 10 February 2021 / Accepted: 12 February 2021 / Published: 17 February 2021
(This article belongs to the Special Issue Metabolomics Applications in Food Analysis: Quality, Safety, and Health)
Browse Figures
Versions Notes

Abstract

:
With the increase in soybean trade between countries, the intentional mislabeling of the origin of soybeans has become a serious problem worldwide. In this study, metabolic profiling of soybeans from the Republic of Korea and China was performed by nuclear magnetic resonance (NMR) spectroscopy coupled with multivariate statistical analysis to predict the geographical origin of soybeans. The optimal orthogonal partial least squares-discriminant analysis (OPLS-DA) model was obtained using total area normalization and unit variance (UV) scaling, without applying the variable influences on projection (VIP) cut-off value, resulting in 96.9% sensitivity, 94.4% specificity, and 95.6% accuracy in the leave-one-out cross validation (LOO-CV) test for discriminating between Korean and Chinese soybeans. Soybeans from the northeastern, middle, and southern regions of China were successfully differentiated by standardized area normalization and UV scaling with a VIP cut-off value of 1.0, resulting in 100% sensitivity, 91.7%–100% specificity, and 94.4%–100% accuracy in a LOO-CV test. The methods employed in this study can be used to obtain essential information for the authentication of soybean samples from diverse geographical locations in future studies.

Graphical Abstract

1. Introduction

Soybean (Glycine max L. Merr.) is an important legume for food, animal feed, and biofuels [1]. Soybean contains 40% protein; therefore, it is a common, rich and easily accessible protein source. It is also rich in soluble sugars/dietary fiber (35%) and oil (20%), 85% of which is cholesterol-free, and contains both monounsaturated and polyunsaturated fatty acids [2]. Soybean metabolites exhibit a wide range of pharmacologically beneficial effects, including antioxidant, hypo-cholesterolemic, anticarcinogenic, immunostimulatory, antidiabetic activities, and reduction of osteoporosis risk [2,3].
With the increase in international agricultural trade and consumer demand for safe and high-quality food, identifying the geographical origin of agricultural products has increased in importance. Soybeans are one of the most traded agricultural products worldwide. As soybean trade between countries increases, some soybean distributors and processing companies mix relatively inexpensive imported soybeans with domestic soybeans and sell them without revealing the exact source of origin information. Identifying the geographical origin of soybeans is important from several perspectives. First, the safety issues of genetically modified (GM) soybeans and low-quality imported soybeans should be considered. More than half of the soybeans produced worldwide are GM soybeans; this has raised consumer concerns regarding allergy risks [4,5]. In addition, there are quality degradation issues, such as microbial contamination and pesticide detection, during the storage and distribution of imported soybeans. Second, regarding the nutritional properties of soybeans, previous studies have shown that the levels of primary and secondary metabolites such as soybean sugars, proteins, fatty acids, and phytochemicals differ depending on the geographical origin [6,7]. Third, different soybean processing technologies are required depending on the soybean origin because the content of useful nutrients or anti-nutritional factors (trypsin inhibitors) varies. Processing technologies that can minimize the destruction of beneficial nutrients while removing unnecessary nutrients from the soybeans must be developed and applied appropriately depending on the origin [8]. Therefore, it is necessary to develop a method for determining the geographical origin of soybeans that can solve the problem of intentionally mislabeling the origin of soybeans, which has disrupted the soybean distribution system and created confusion among consumers.
Metabolomics has been widely used in previous studies to observe the changes in metabolites of agricultural products from different geographical origins using high-performance liquid chromatography (HPLC), gas chromatography (GC), capillary electrophoresis-time-of-flight (CE-TOF) combined with mass spectrometry (MS), nuclear magnetic resonance (NMR) spectroscopy, and Fourier-transform infrared (FT-IR) spectroscopy [9,10,11,12,13]. Among various analytical platforms, NMR spectroscopy has been widely used to characterize food resources because of its simultaneous detection of diverse compounds within a complex mixture, high reproducibility, and noninvasive nature [14,15]. Several studies have analyzed different soybean varieties using NMR, GC-MS, and ultra-performance liquid chromatography (UPLC)-tandem mass spectrometry (MS/MS) analyses [16,17]. The profiling of isoflavones and anthocyanins in black soybeans from different geographical locations in southwest China has been performed using HPLC-MS [7]. Metabolite fingerprinting has been widely used as an important tool for authentication of food and agricultural products [18]. This allows the determination of exact geographical origin and the detection of any unusual quality of raw materials in both unprocessed and processed products [18,19]. However, comparative and comprehensive metabolic profiling using NMR spectroscopy has rarely been applied to soybean samples from different geographical regions.
Accordingly, in the present study, soybean samples obtained from major soybean-producing areas and their local markets were used to establish a representative prediction model for discriminating the geographical origin of soybeans from Korea and China. NMR-based metabolic profiling coupled with multivariate statistical analysis was employed to determine the geographical origin of soybeans from Korea and China.

2. Materials and Methods

2.1. Soybean Samples

Authentic Korean soybeans from eight regions harvested in 2016 were provided from the National Agricultural Products Quality Management Service of the Republic of Korea. These were obtained from Gyeonggi-do Anseong, Gangwon-do Yeongwol, Chungcheongbuk-do Eumseong, Chungcheongnam-do Cheonan, Jeollabuk-do Imsil, Jeollanam-do Yeonggwang, Gyeongsangbuk-do Uiseong, and Gyeongsangnam-do Geochang, 500 g to 2 kg from each region. Chinese soybeans were purchased from online suppliers in October 2016. Nine Chinese samples were obtained from the three major and largest soybean producing regions including northeastern (Heilongjiang, Jilin, Liaoning), middle (Hebei, Shandong, and Hubei), and southern (Zhejiang, Guangdong, and Guangxi) regions, 500 g from each region, and the products and suppliers’ information of these soybean products are listed in Table S1. Geographical information on Korean and Chinese soybean samples is presented in Figure S1. Four replicates from each pool were collected for each region, ground in liquid nitrogen using a blender, freeze-dried, and then stored at −80 °C until NMR analysis. For measurement of size, 10 grains of soybean samples from each region were randomly selected and measured. The length along the hypocotyls was measured as soybean size using a digimatic caliper (0–150 mm, S. Tools, Wuxi, China).

2.2. Chemicals and Reagents

Methanol-d4 (MeOD, 99.8% atom D) and sodium deuteroxide (NaOD, 99.5% atom D; 40% in D2O) were obtained from Cambridge Isotope Laboratories, Inc. (Andover, MA, USA). Deuterium oxide (D2O, 99.9% atom D) including 0.05% 3-(trimethylsilyl) propionic-2, 2, 3, 3-d4 acid sodium salt (TSP) and potassium phosphate (KH2PO4) was purchased from Sigma-Aldrich (St. Louis, MO, USA).

2.3. Climate Data for Soybean Cultivation Regions in Korea and China

The monthly average temperature (°C) and total precipitation (mm) data for soybean cultivation regions in Korea and China in 2016 were obtained from the Korea Meteorological Administration (https://data.kma.go.kr) (accessed on 12 November 2020) and China Meteorological Administration (https://data.cma.cn/en) (accessed on 12 November 2020), respectively. The average annual temperature and precipitation were calculated by averaging monthly climate data for one year (January to December 2016).

2.4. NMR Measurement and Peak Assignment

To extract various groups of metabolites in soybeans, aqueous methanol (50% D2O-MeOD mixture) was used as an extraction solvent in this study [20,21]. Fifty milligrams of soybean powder and 1.5 mL of 50% D2O-methanol mixture (D2O was titrated to pH 6 using NaOD) were transferred into a 2-mL centrifuge tube, and then vortexed and sonicated for 1 min and 15 min, respectively. Thereafter, the material was centrifuged at 17,000× g, 4 °C for 10 min. Buffer solution of 0.1 M KH2PO4 was prepared from D2O and NaOD was used to adjust pH to 6. The clear supernatant was filtered with a 0.45-µm Whatman filter (PTFE, Sigma-Aldrich), and 600 µL of the sample was transferred into 5-mm NMR tube (Norell, Landisville, NJ, USA).
A 600-MHz Bruker Avance spectrometer (Bruker, Germany) was employed to analyze soybean samples at a temperature of ~25 °C to record all NMR spectra. For 1H-NMR spectra, 64 K data points were obtained with a relaxation delay of 2.0 s and a spectral width of 12626.3 Hz. A scan number of 128 and an acquisition time of 2.6 s were used. Water suppression was conducted to exclude the region between δ = 4.7 to 5.0. For two-dimensional NMR spectra, 1H-1H correlation spectroscopy (COSY) spectra were acquired under the following conditions: 32 scans, relaxation delay of 2.0 s, and 7812.5 Hz spectral width. 1H-13C heteronuclear single quantum correlation (HSQC) spectra were obtained with 32 scans, 2.0 s relaxation delay, and 6631.3 Hz spectral width. The following pulse sequence described by Suh et al. [22] was used to collect 1H-NMR spectra: relaxation delay –90°–t1–180°–t1–acquire for time t2. The relaxation delay of t1 was incremented together with the increasing delay, and the relaxation delay was 2.0 s for time t2. Baseline correction and assignments of all 1H-NMR spectra were performed by using Chenomx NMR suite software (version 8.2, Chenomx, Edmonton, AB, Canada) and further identification of metabolites was performed based on the HMDB database (HMDB, http://www.hmdb.ca/) (accessed on 18 January 2020). Non-overlapping peaks were used for the peak assignment. MestReNova (version 6.0.4, Mestrelab Research, Santiago de Compostela, Spain) was employed to measure the J value of the peaks, and to identify the peaks of 1H-1H COSY and 1H-13C HSQC spectra.

2.5. Data Processing and Statistical Analyses

Binning and normalization of 1H-NMR spectral data were performed using Chenomx NMR suit software. Baseline corrected NMR spectral data ranging from 0.08 to 10.00 ppm were segmented into a series of small bins (total 245) with widths of 0.04 ppm while excluding the water suppression region (4.70–4.86 ppm). Then the spectra were normalized into total area normalization and standardized area normalization. Intensities in each binned spectral data acquired from total area normalization and standardized area normalization were calculated by relative intensities to the total area of all bins and area of reference peak, respectively. Results of the binned datasets were converted to Microsoft Office Excel (version 2007, Microsoft, Redmond, WA, USA) compatible format to measure each compound by its loading value. Binning values of compounds having multiple non-overlapping peaks were summed. Then, the data was imported into SIMCA (version 15.0, Umetrics, Umeå, Sweden) for (orthogonal) partial least squares-discriminant analysis ((O)PLS-DA) of soybean samples (n = 68). Optimal (O)PLS-DA models were determined by good-fit parameter; R2Y and predictability parameter; Q2Y, as well as R2Y-intercept values and Q2Y-intercept values, which were obtained by permutation tests. When establishing the OPLS-DA model, two scaling methods, unit variance (UV) and Pareto, were applied and the results compared to identify the most optimal scaling method. UV scaling and Pareto scaling are methods of dividing each variable by a scaling factor as the standard deviation and square root of standard deviation of the intensity of each metabolite in all samples, respectively. Leave-one-out cross-validation (LOO-CV) was performed to detect and prevent overfitting of the models. LOO-CV leaves out one of the data and the model is built on the remaining data. Then the left out data are predicted repeatedly from the new model until the entire data have been predicted at least once [23].
Sensitivity, specificity, and accuracy were calculated to evaluate the classification performance of the model based on the class prediction value of the sample obtained from the LOO-CV (Y-predcv) by using SIMCA software. Sensitivity is the parameter that measure the classification ability of the model for correct class of cases, whereas specificity measures the prediction ability of the model for correct class of controls [24]. Accuracy means the total proportion of correct class in both cases and controls, which measures the veracity of the model [25]. Receiver operating characteristics (ROC) curve analysis was performed using SIMCA software.

3. Results & Discussion

3.1. Size Measurement of Soybean Samples

The geographical distribution of the collected Korean and Chinese soybean samples is shown in Figure S1. Soybean samples from each region were photographed (Figure S2), and their sizes were measured. To obtain a representative sample and verify its practical significance, soybeans obtained from local markets were used in the experiments. The average size of Korean soybeans was 8.38 mm, which was significantly larger than that of Chinese soybeans (7.60 mm), as shown in Figure S3A. The size of Chinese soybeans differed widely from region to region, and even from province to province within the same region, whereas the size of Korean soybeans showed no significant regional differences as shown in Figures S2 and S3B. The yield of soybeans was influenced by various factors such as climate, adopted cultural practice, cultivars, and soil conditions [26]. For example, under soil moisture or phosphorous deficiency conditions, the size and number of soybean seeds were markedly decreased, which was a major cause of low soybean yields [27,28]. Therefore, the difference in the size of soybeans between Korea and China or among Chinese regions (and between province and province in the same region) might be affected by differences in the cultivars or environmental factors, such as temperature, precipitation, and soil conditions. As shown in the climate data (Figure S4), there was no significant difference in the average annual temperature and precipitation between Korea and China. However, in China, the average annual temperature and precipitation were significantly different among the three regions. When comparing the climate data of each province in China, the average annual precipitation was significantly different by province only for the middle region; however, the average annual temperature was not different among provinces in all respective regions.
The large-seeded species might have an advantage during seedling establishment, whereas the small-seeded species might have an advantage during seed production under various environmental conditions [29]. Although cultivar information in both countries could not be obtained for the present study, the selection of larger soybean varieties might have occurred since ancient times in Korea [30]. Hence, it is assumed that this tendency has been ongoing and has resulted in a larger size of representative soybean cultivars in Korea than in China.

3.2. Identification of Soybean Metabolites Using NMR Spectroscopy

Representative 1H-NMR spectra of Korean and Chinese soybeans are shown in Figure S5. We obtained 68 1H-NMR spectra using four experimental replications from eight and nine soybean samples from Korea and China, respectively. A total of 25 metabolites were identified, as listed in Table 1. These included ten amino acids (alanine, asparagine, aspartate, choline, galactarate, glutamate, isoleucine, leucine, tryptophan, and valine), three carbohydrates (glucose, raffinose/stachyose, and sucrose), nine organic acids (2-oxoglutarate, acetate, acetoacetate, citrate, formate, fumarate, malonate, succinate, and tartarate), one fatty acid (2-hydroxyisobutyrate), and others (hypoxanthine and oxy-purinol). Using 1H-1H COSY and 1H-13C HSQC spectral data, 16 and 23 metabolites were confirmed, respectively (Figure S6).

3.3. Discrimination and Prediction of Korean and Chinese Soybean Samples

The selection of the most appropriate normalization and scaling methods is an important step for improving the biological information of metabolomics data as it decreases any unwanted biases originating from biological and technical variance and adjusts different ranges between samples or variables for comparison [31,32]. Normalization is a row operation that reduces the significant intensity variation of metabolites between samples (sample-to-sample variation); therefore, all samples can be compared with each other. Scaling is a column operation that adjusts the intensity variance between metabolites (metabolite-to-metabolite), and thus, all metabolites can be compared with each other [33].
OPLS-DA and permutation tests were performed to determine the optimal normalization (standardized area normalization versus total area normalization) and scaling methods (UV versus Pareto) to discriminate between Korean and Chinese soybeans. The number of components was selected using the autofit function in SIMCA software, which extracts a significant number of principal components from the models. The optimal model was determined by a good-fit parameter, R2Y, and a predictability parameter, Q2Y, with values close to 1, which indicate good fitness and prediction of the experiment, respectively.
The R2Y- and Q2Y-intercept values from permutation tests should be below 0.40 and 0.05, respectively, in a statistically valid OPLS-DA model [34]. As listed in Table 2, the highest R2Y value of 0.882 and Q2Y value of 0.783 were obtained by applying the total area normalization, UV scaling, and five OPLS components (1+4) with satisfactory R2Y- and Q2Y- intercept values of 0.254 and –0.487, respectively. The R2Y-intercept value of 0.254 and Q2Y-intercept value of –0.487 from the permutation test proved the statistical validity of the model without overfitting of the data. The total area normalization used in the model for discriminating between Korean and Chinese soybeans divided each metabolite peak area by the total peak area; therefore, each sample had the same total peak area unit of 1. This is one of the most commonly used normalization methods in NMR metabolomics. Thus, each peak intensity can be expressed as a fraction of the total peak intensity (as a percentage), making it possible to compare each metabolite level between samples in the same unit [33]. The UV scaling method is one of the easiest ways to adjust the metabolic variations by giving each metabolite the same importance, making the standard deviation equal to 1 for all metabolites [35,36].
Based on the selected total area normalization and UV scaling methods, various variable influences on projection (VIP) values were applied to select the optimal OPLS-DA model. The VIP values from the 25 metabolites are listed in Table S2. As listed in Table S3, the best OPLS-DA model (R2Y 0.882 and Q2Y 0.783) for discriminating between Korean and Chinese soybean samples was constructed without applying the VIP cut-off value.
OPLS-DA-derived score plots (Figure 1A) showed clear discrimination between Korean and Chinese soybean samples. Y prediction plots using the LOO-CV are shown in Figure 1B. These were used to evaluate the ability of the OPLS-DA model to determine the origin of soybean samples (i.e., Korea or China). A threshold value of 0.5 was adopted to classify the origin of the soybean samples. Except for one Korean sample (Kangwon-Yeongwol) and two Chinese samples (Heilongjiang and Guangdong provinces), most of the samples were classified correctly, showing 96.9% sensitivity, 94.4% specificity, and 95.6% accuracy (Table 3). ROC curve analysis showed the area under the curve (AUC) value of 1.0 for predicting the geographical origin of soybean samples from Korea and China (Figure S7A). This identified 25 metabolites which could be used as potential biomarkers differentiating the geographical origin of soybean sample from Korea and China. Thus, the OPLS-DA model obtained in the present study is useful for discriminating and predicting Korean and Chinese soybean samples.
The relative levels of metabolites in the soybean samples from Korea and China are listed in Table S4. Among the 25 total metabolites, nine showed significantly different relative levels when comparing Korean and Chinese soybean samples. Relative levels of alanine, citrate, isoleucine, tartarate, and valine were significantly higher (p < 0.05) in Korean soybeans than Chinese soybeans, whereas those of asparagine, choline, galactarate, and tryptophan were significantly higher in Chinese soybeans than Korean soybeans. These nine metabolites (VIP value over 0.87) were suggested as influential contributors to the differentiation between Korean and Chinese soybean samples (Table S2).
Furthermore, this study investigated the possibility of a discriminant model for soybean samples from Korea and northeastern China. In geographic coordinates, these two sites are located relatively close along the latitudinal and longitudinal axis (38° N, 127° E in Korea, 41–48° N, 122–129° E in northeastern China). As shown in Figure S8A,B, soybean samples from Korea and northeastern China were well discriminated. The LOO-CV test showed an accuracy of 100% for predicting origin (Figure S8C). Accordingly, soybean samples from Korea and northeastern China could also be differentiated by NMR-based metabolic profiling, even though the geographical locations of Korea and northeastern China are relatively close.
Differences in the metabolic profile levels of seed plants including soybeans are influenced by genetic factors, such as cultivar; however, they are also influenced by environmental factors [37,38]. Based on thousands of years of the rich planting experience of breeders and on specific environmental conditions, many scientists and breeders in each region have developed their own environmentally compatible soybean varieties, such as drought-, cold-, disease-, and insect-resistant varieties, to accommodate environmental conditions in soybean breeding programs or using historical landraces maintained by farmers for their seed lots. Therefore, the influence of the variety and environmental background factors on differences in metabolites between Korean and Chinese soybean samples could be considered. Korea and China are the main countries that have a long cultivation history and broad genetic diversity of soybean, and soybean genotypes collected from these countries are known to be different [39,40,41,42]. The history of soybean cultivation in Korea began 2500 years ago, and different cultivars with various forms and agronomic characteristics were formed over the years, which were utilized as genetic resources for the improvement of modern varieties. However, these native soybeans were quickly replaced by newly developed varieties, which lowered the genetic diversity of Korean soybean varieties [43]. The history of soybean cultivation in China began 3000 years ago, and the northern, Huang Huai (area between Yellow River–Huai Rivers) and southern regions are the major cultivation areas [44]. China has a diverse climate with varying soil characteristics and rainfall patterns. Over 1400 soybean cultivars have been developed in China between 1923 and 2013 [45] through a scientific breeding program; therefore, the parents of the cultivars were landraces, cultivars, breeding lines, and exotic introductions. During the past 20 to 40 years, the yield level of soybean cultivars has increased from 750 kg/hm2 to 2000 kg/hm2 in China [46].
Although we could not obtain cultivar information for soybeans, the significantly different relative levels of various metabolites between Korean and Chinese soybeans might be due to different environmental conditions, such as temperature, precipitation, and soil conditions (type, particle size, and nutrient content), which are closely related to metabolic expression. The growth and metabolic characteristics of the same cultivar varied depending on the environmental conditions of the growing region, which caused differences in its cellular metabolism [47].
There are several studies reporting the effect of various soil conditions on the yield and quality characteristics of soybean seeds. Higher soybean seed yield was observed in soil with low phosphorous and potassium content, and the yield was also improved under the physical properties of soil with a higher clay content, which was due to the high plant-available water in that type of soil during the growing season [48]. Soybean seed composition varied depending on the soil nutrient content. In the correlation analysis of soil nutrients and seed composition, low contents of soil organic matter and elements such as N, C, K, B, and Zn were related to the lower content of soybean protein and oleic acid [49]. Similarly, amino acid content in soybean was positively correlated with the contents of soil elements such as B, Mo, Se, K and N [50]. Soil moisture stress is also an important abiotic factor for modulating the soybean yield and quality. Under a deficient soil moisture condition, soybean seed yield was reduced and the contents of total protein, palmitic acid, linoleic acid, sucrose, raffinose, stachyose, N, P, K and Ca were markedly decreased, whereas those of total oil, stearic acid, oleic acid, linolenic acid, Fe, Mg, Zn, Cu and B were increased [51]. Therefore, it is suggested that the contents of soil nutrients or moisture during the soybean reproductive periods could be critical factors for influencing the yield and nutrient value (various metabolites) in soybean seeds.
In our study, there was no significant difference in the average annual temperature and precipitation of Korea and China (Figure S4), so it is assumed that the soil moisture or nutrient contents might be different depending on the soil conditions in each country. Even with the same precipitation environment, soil moisture content could be different according to various types, textures, and particle size of soils in each region or country, because the moisture-holding capacity could differ depending on the soil condition. Therefore, other environmental factors, such as soil conditions, rather than temperature and precipitation, and diverse soybean cultivars improved over centuries to adapt to the particular environmental conditions of each region, might be important factors influencing the different levels of metabolites in Korean and Chinese soybeans.

3.4. Discrimination and Prediction of Domestic Chinese Soybean Samples

Korean soybeans could not be categorized (data not shown) according to various provinces, possibly because most of the provinces in Korea have similar climatic, soil, and rainfall conditions as well as soybean varieties grown on a small land scale. Therefore, we focused on the differentiation in domestic Chinese soybean samples.
Various normalization and scaling methods were tested to determine the optimal PLS-DA model to differentiate domestic Chinese soybean samples. The highest R2Y value of 0.898 and Q2Y value of 0.651 were obtained by applying standardized area normalization, UV scaling, and six PLS components satisfying R2Y and Q2Y intercept values in the PLS-DA model (Table 4). Standardized area normalization divides the intensity of each metabolite by a constant concentration of the internal standard compound, allowing the measurement of the contribution of each metabolite to the spectrum [52]. Because internal standards added prior to extraction can monitor and correct the intensity drift that might occur during extraction and instrument analysis, standardized area normalization can reduce the difference in extraction efficiency between samples [53,54]. Meanwhile, the Pareto scaling method emphasizes weak peaks with high biological relevance and reduces the effect of intense peaks, thereby reducing the effect of noise variables more than the UV scaling method [55].
Based on the selected standardized area normalization and UV scaling methods, various VIP cut-off values were applied to select the optimal OPLS-DA model. The VIP values of the 25 metabolites are listed in Table S5. When a VIP cut-off value of 1.0 was applied, the best PLS-DA model was obtained with 11 metabolites showing the highest R2Y value of 0.887, Q2Y value of 0.789, R2Y intercept value of 0.151, and Q2Y intercept value of –0.480 (Table S6).
Chinese soybean samples from the northeastern region (NR), middle region (MR), and southern region (SR) were successfully differentiated by NMR-based metabolic profiling (Figure 2A). In the development of the PLS-DA model for discriminating between more than two groups, cross-validation could be conducted in three groups that were binary coded with the three dummy Y variables, (0 1 1) for samples in class 1, (1 0 1) for samples in class 2, and (1 1 0) for samples in class 3, where 0 and 1 indicate the control and case, respectively [56]. Therefore, three independent class approaches were employed in our study. Y prediction plots after performing the LOO-CV are shown in Figure 2B–D. When designating control groups as both NR and SR independently, Ypredcv values (predicted Y values of each left-out sample calculated after performing cross-validation) of each control and case groups were correctly classified with a threshold value of 0.5 (Figure 2B,D). Only two misclassified samples were detected in the model designating the MR to the control group (Figure 2C). In summary, 100% sensitivity, specificity, and accuracy were obtained in the model designating the NR and SR to the control group, and 100% sensitivity, 91.7% specificity, and 94.4% accuracy were obtained in the model designating the MR to the control group (Table 3).
In ROC curve analysis, all AUC values for predicting the origin of soybean samples from three Chinese regions were 1.0 (Figure S7B–D). It was confirmed that the selected 11 metabolites could be used as potential biomarkers for the differentiation of geographical origin of soybean samples from three Chinese regions.
Seventeen metabolites showed significantly different levels among soybeans from the three regions of China analyzed in our study (Table S7). Among these 17 metabolites, four metabolites (glucose, hypoxanthine, leucine, and tartarate) showed significantly higher levels in the NR, and SR was characterized by significantly higher levels of malonate alone. However, in the MR, no particular metabolite showed a relatively higher level than the other regions.
The present study assumed that the differences in soybean metabolites among Chinese regions might be affected by different environmental conditions including temperature, precipitation, and soil as well as cultivar differences. Based on China’s regional climate statistics, there were significant differences in the average annual temperature and total precipitation among the three Chinese regions in 2016 (Figure S4). The average annual temperature in the NR of China was the lowest compared to the other regions, and there was no significant difference in temperature between the MR and SR. The average annual precipitation was the highest in the SR of China compared to the other regions, whereas no significant difference was found in the average annual precipitation between the NR and MR. Sugar accumulation has been observed in soybeans obtained from the cold northeastern region of China, whereas amino acid and total protein content was abundant in soybeans sourced from the warm southern region in China [38]. North-to-south latitudinal gradients in China have a greater influence on metabolite profile changes in soybeans than east-to-west longitudinal gradients [38]. This is consistent with our study results showing the lowest temperature and highest soybean glucose level, which are characteristic features of soybeans from northeastern China. Plants exposed to low temperatures accumulate starch-derived sugars to adapt to cold stress and use sugars as an energy source or cryoprotectant [57].
Previous studies have reported that the accumulation of alanine and gamma-aminobutyric acid (GABA) in soybean roots is the most characteristic response to flood stress [58]. This is consistent with our findings that soybean alanine levels were the highest in the SR of China, which had relatively higher precipitation than the other regions. The accumulation of amino acids such as alanine and GABA is an important adaptation mechanism for storing insufficient carbon and nitrogen in an oxygen-deficient environment [58]. Therefore, flood-stressed plant cells maintain intracellular osmotic pressure by accumulating amino acids to compensate for the reduced soluble carbohydrate level [58].
In the present study, Chinese soybeans could be differentiated based on their cultivation locations (i.e., northeastern, middle, and southern regions), suggesting that regionally adapted varieties were characterized in each region of China under their own inherent growing environments. We previously reported the classification of soybeans from Korea and China using FT-IR spectroscopy [59]. The NMR-based differentiation and prediction methods for identifying the geographical origin of soybeans in the present study could be used as a complementary method to the FT-IR-based technology to provide orthogonal criteria for more precisely discriminating soybean samples.

4. Conclusions

In the present study, NMR spectroscopy coupled with multivariate statistical analysis was employed to differentiate the geographical origin of soybeans. We succeeded in discriminating and predicting the origin of Korean and Chinese soybeans using the OPLS-DA model with LOO-CV. In addition, soybeans from the northeastern, middle, and southern regions of China were differentiated using PLS-DA models. Based on the requirements of growers and consumers, various soybean varieties have been planted and grown worldwide. These soybean varieties have been utilized in soybean products such as soy oil, soy sauce, tofu, and soymilk, and the market life of soybean varieties is short. Hence, prediction models and databases for investigating the geographical origin of soybeans should be regularly updated (at least every 2–3 years). As an investigative work, we established a simple, efficient, and convenient method for discriminating the precise geographical origin of soybeans in the present study. This method could be widely used for the detection of soybean samples with intentionally mislabeled geographical origin. In future studies, soybean samples from diverse geographical locations worldwide that have been harvested in different years will be investigated using the methods developed in the present study.

Supplementary Materials

The following are available online at https://www.mdpi.com/2304-8158/10/2/435/s1, Figure S1: Geographical distribution of Korean and Chinese soybean samples. Map of Korea with 8 provinces of soybean samples (A), Map of China with 3 divided regions of soybean samples consisting 9 provinces (B). Figure S2: Representative pictures of soybean samples from Korea and China. Figure S3: Average size of soybean samples from Korea and China. Comparison of average soybean size in Korea and China (A), Comparison of individual soybean size of each region and province in Korea and China (B). Figure S4: Climate data for soybean cultivation regions in Korea and China in 2016. The average value of monthly mean temperature (A) and total precipitation (B) from January to December. Figure S5: Representative 600 MHz 1H-NMR spectra of Korean (A) and Chinese (B) soybean samples. Figure S6: 1H-1H correlation spectroscopy (COSY, (A)) and 1H–13C heteronuclear single quantum correlation (HSQC, (B)) spectra of soybean sample. Figure S7: Receiver operating characteristics (ROC) curves and area under the curve (AUC) values for distinguishing geographical origin of soybeans. AUC value of 25 metabolites discriminating soybean samples from Korea and China (A). AUC value 11 metabolites discriminating Chinses soybean samples from NR and MR/SR (B), MR and NR/SR (C), and SR and NR/MR (D). Figure S8: Discrimination model for soybean samples from Korea and northeastern China. OPLS-DA score plots (A) derived from the 1H-NMR spectra of soybean samples for discriminating the geographical origin of Korea and northeastern China. Permutation test plots (B) with 100 permutations of OPLS-DA model. Leave-one-out cross-validated score plots (C). Table S1: Products and suppliers’ information of Chinese soybean samples. Table S2: Variable importance projection (VIP) values of metabolites for discriminating Korean and Chinese soybean samples. Table S3: Parameters of OPLS-DA models based on various VIP cut-off values for discriminating Korean and Chinese soybean samples based on total area normalization and UV scaling methods. Table S4: Relative levels of metabolites in Korean and Chinese soybean samples. Table S5: Variable importance projection (VIP) values of metabolites for discriminating different origins of the Chinese soybean samples. Table S6: Parameters of PLS-DA models based on various VIP cut-off values for discriminating different origins of the Chinese soybean samples based on standardized area normalization and UV scaling methods. Table S7: Relative levels of metabolites in Chinese soybean samples.

Author Contributions

Conceptualization, D.-Y.L. and H.-K.C.; methodology, D.-Y.L. and H.-K.C.; validation, J.-A.S., Y.-S.K. and D.-Y.L.; formal analysis, Y.Z. and S.-Y.K., and J.-S.L.; investigation, Y.Z., S.-Y.K., and J.-S.L.; resources, B.-K.S.; data curation, Y.Z. and S.-Y.K.; writing—original draft preparation, Y.Z., S.-Y.K., and H.-K.C.; writing—review & editing, S.-Y.K. and H.-K.C.; supervision, D.-Y.L. and H.-K.C.; project administration, H.-K.C.; funding acquisition, H.-K.C. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIP) (NRF-2015R1A5A1008958), Korea Institute of Planning and Evaluation for Technology in Food, Agriculture, Forestry and Fisheries (IPET) through Advanced Production Technology Development Program, funded by Ministry of Agriculture, Food and Rural Affairs (MAFRA) (316081-04).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data presented in this study are available on request from the corresponding author.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Ferguson, B.J.; Indrasumunar, A.; Hayashi, S.; Lin, M.-H.; Lin, Y.-H.; Reid, D.E.; Gresshoff, P.M. Molecular analysis of legume nodule development and autoregulation. J. Integr. Plant Biol. 2010, 52, 61–76. [Google Scholar] [CrossRef] [PubMed]
  2. Jooyandeh, H. Soy products as healthy and functional foods. Middle East J. Sci. Res. 2011, 7, 71–80. [Google Scholar]
  3. Fournier, D.B.; Erdman, J.W., Jr.; Gordon, G.B. Soy, its components, and cancer prevention: A review of the in vitro, animal, and human data. Cancer Epidemiol. Biomark. Prev. 1998, 7, 1055–1065. [Google Scholar]
  4. Berdal, K.G.; Holst-Jensen, A. Roundup Ready® soybean event-specific real-time quantitative PCR assay and estimation of the practical detection and quantification limits in GMO analyses. Eur. Food Res. Technol. 2001, 213, 432–438. [Google Scholar] [CrossRef]
  5. Stevenson, S.E.; Woods, C.A.; Hong, B.; Kong, X.; Thelen, J.J.; Ladics, G.S. Environmental effects on allergen levels in commercially grown non-genetically modified soybeans: Assessing variation across North America. Front. Plant Sci. 2012, 3, 196. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  6. Lee, J.H.; Choung, M.G. Comparison of nutritional components in soybean varieties with different geographical origins. J. Korean Soc. Appl. Biol. Chem. 2011, 54, 254–263. [Google Scholar] [CrossRef]
  7. Wu, H.J.; Deng, J.C.; Yang, C.Q.; Zhang, J.; Zhang, Q.; Wang, X.C.; Yang, F.; Yang, W.Y.; Liu, J. Metabolite profiling of isoflavones and anthocyanins in black soybean [Glycine max (L.) Merr.] seeds by HPLC-MS and geographical differentiation analysis in Southwest China. Anal. Methods 2017, 9, 792–802. [Google Scholar] [CrossRef]
  8. Qin, G.N.; Verstegen, M.W.A.; Van der Poel, A.F.B. Effect of temperature and time during steam treatment on the protein quality of full-fat soybeans from different origins. J. Sci. Food Agric. 1998, 77, 393–398. [Google Scholar] [CrossRef]
  9. Longobardi, F.; Ventrella, A.; Napoli, C.; Humpfer, E.; Schütz, B.; Schäfer, H.; Kontominas, M.G.; Sacco, A. Classification of olive oils according to geographical origin by using 1H NMR fingerprinting combined with multivariate analysis. Food Chem. 2012, 130, 177–183. [Google Scholar] [CrossRef]
  10. Wu, Z.; Zhao, Y.; Zhang, J.; Wang, Y. Quality assessment of Gentiana rigescens from different geographical origins using FT-IR spectroscopy combined with HPLC. Molecules 2017, 22, 1238. [Google Scholar] [CrossRef] [Green Version]
  11. Yudthavorasit, S.; Wongravee, K.; Leepipatpiboon, N. Characteristic fingerprint based on gingerol derivative analysis for discrimination of ginger (Zingiber officinale) according to geographical origin using HPLC-DAD combined with chemometrics. Food Chem. 2014, 158, 101–111. [Google Scholar] [CrossRef] [PubMed]
  12. Zhang, L.; Wang, X.; Guo, J.; Xia, Q.; Zhao, G.; Zhou, H.; Xie, F. Metabolic profiling of Chinese tobacco leaf of different geographical origins by GC-MS. J. Agric. Food Chem. 2013, 61, 2597–2605. [Google Scholar] [CrossRef]
  13. Zhao, J.; Zhao, Y.; Hu, C.; Zhao, C.; Zhang, J.; Li, L.; Zeng, J.; Peng, X.; Lu, X.; Xu, G. Metabolic profiling with gas chromatography-mass spectrometry and capillary electrophoresis-mass spectrometry reveals the carbon-nitrogen status of tobacco leaves across different planting areas. J. Proteome Res. 2016, 15, 468–476. [Google Scholar] [CrossRef] [PubMed]
  14. Consonni, R.; Cagliani, L.R. Geographical characterization of polyfloral and acacia honeys by nuclear magnetic resonance and chemometrics. J. Agric. Food Chem. 2008, 56, 6873–6880. [Google Scholar] [CrossRef] [PubMed]
  15. Pereira, G.E.; Gaudillere, J.P.; Van Leeuwen, C.; Hilbert, G.; Maucourt, M.; Deborde, C.; Moing, A.; Rolin, D. 1H NMR metabolite fingerprints of grape berry: Comparison of vintage and soil effects in Bordeaux grapevine growing areas. Anal. Chim. Acta 2006, 563, 346–352. [Google Scholar] [CrossRef]
  16. Lin, H.; Rao, J.; Shi, J.; Hu, C.; Cheng, F.; Wilson, Z.A.; Zhang, D.; Quan, S. Seed metabolomic study reveals significant metabolite variations and correlations among different soybean cultivars. J. Integr. Plant Biol. 2014, 56, 826–836. [Google Scholar] [CrossRef]
  17. Harrigan, G.G.; Skogerson, K.; Macisaac, S.; Bickel, A.; Perez, T.; Li, X. Application of 1H NMR profiling to assess seed metabolomic diversity. A case study on a soybean era population. J. Agric. Food Chem. 2015, 63, 4690–4697. [Google Scholar] [CrossRef] [PubMed]
  18. Longobardi, F.; Innamorato, V.; Di Gioia, A.; Ventrella, A.; Lippolis, V.; Logrieco, A.F.; Catucci, L.; Agostiano, A. Geographical origin discrimination of lentils (Lens culinaris Medik.) using 1H NMR fingerprinting and multivariate statistical analyses. Food Chem. 2017, 237, 743–748. [Google Scholar] [CrossRef]
  19. Mannino, G.; Di Stefano, V.; Lauria, A.; Pitonzo, R.; Gentile, C. Vaccinium macrocarpon (Cranberry)-based dietary supplements: Variation in mass uniformity, proanthocyanidin dosage and anthocyanin profile demonstrates quality control standard needed. Nutrients 2020, 12, 992. [Google Scholar] [CrossRef] [Green Version]
  20. Kim, H.K.; Verpoorte, R. Sample preparation for plant metabolomics. Phytochem. Anal. 2009, 21, 4–13. [Google Scholar] [CrossRef] [PubMed]
  21. Kim, H.K.; Choi, Y.H.; Verpoorte, R. NMR-based metabolomic analysis of plants. Nat. Protoc. 2010, 5, 536–549. [Google Scholar] [CrossRef] [PubMed]
  22. Suh, H.W.; Kim, S.H.; Park, S.J.; Hyun, S.H.; Lee, S.Y.; Auh, J.H.; Lee, H.J.; Cho, S.M.; Kim, J.H.; Choi, H.K. Effect of Korean black raspberry (Rubus coreanus Miquel) fruit administration on DNA damage levels in smokers and screening biomarker investigation using 1H-NMR-based metabolic profiling. Food Res. Int. 2013, 54, 1255–1262. [Google Scholar] [CrossRef]
  23. Rubingh, C.M.; Bijlsma, S.; Derks, E.P.; Bobeldijk, I.; Verheij, E.R.; Kochhar, S.; Smilde, A.K. Assessing the performance of statistical validation tools for megavariate metabolomics data. Metabolomics 2006, 2, 53–61. [Google Scholar] [CrossRef] [Green Version]
  24. Szymańska, E.; Saccenti, E.; Smilde, A.K.; Westerhuis, J.A. Double-check: Validation of diagnostic statistics for PLS-DA models in metabolomics studies. Metabolomics 2012, 8, 3–16. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  25. Zhu, W.; Zeng, N.; Wang, N. Sensitivity, Specificity, Accuracy, Associated Confidence Interval and Roc Analysis with Practical Sas Implementations; NESUG Proceeding of Health Care and Life Sciences: Baltimore, MD, USA, 2010. [Google Scholar]
  26. Mattioni, N.M.; Schuch, L.O.B.; Villela, F.A.; Mertz, L.M.; Peske, S.T. Soybean seed size and quality as a function of soil compaction. Seed Sci. Technol. 2012, 40, 333–343. [Google Scholar] [CrossRef]
  27. Brown, E.A.; Caviness, C.E.; Brown, D.A. Response of selected soybean cultivars to soil moisture deficit. Agron. J. 1985, 77, 274–278. [Google Scholar] [CrossRef]
  28. Gutierrez-Boem, F.; Thomas, G.W. Phosphorus nutrition and water deficits in field-grown soybeans. Plant Soil 1999, 207, 87–96. [Google Scholar] [CrossRef]
  29. Moles, A.T.; Westoby, M. Seedling survival and seed size: A synthesis of the literature. J. Ecol. 2004, 92, 372–383. [Google Scholar] [CrossRef]
  30. Lee, G.A.; Crawford, G.W.; Liu, L.; Sasaki, Y.; Chen, X. Archaeological soybean (Glycine max) in east Asia: Does size matter? PLoS ONE. 2011, 6, e26720. [Google Scholar] [CrossRef] [PubMed]
  31. Kohl, S.M.; Klein, M.S.; Hochrein, J.; Oefner, P.J.; Spang, R.; Gronwald, W. State-of-the art data normalization methods improve NMR-based metabolomic analysis. Metabolomics 2012, 8, 146–160. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  32. van den Berg, R.A.; Hoefsloot, H.C.; Westerhuis, J.A.; Smilde, A.K.; van der Werf, M.J. Centering, scaling, and transformations: Improving the biological information content of metabolomics data. BMC Genom. 2006, 7, 142. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  33. Craig, A.; Cloareo, O.; Holmes, E.; Nicholson, J.K.; Lindon, J.C. Scaling and normalization effects in NMR spectroscopic metabonomic data sets. Anal. Chem. 2006, 78, 2262–2267. [Google Scholar] [CrossRef] [PubMed]
  34. Eriksson, L.; Johansson, E.; Kettaneh-Wold, N.; Trygg, J.; Wikström, C.; Wold, S. Part I: Basic Principles and Applications. In Multi-and Megavariate Data Analysis, 2nd ed.; Umetrics Academy: Stockholm, Sweden, 2006; pp. 39–101. [Google Scholar]
  35. Li, B.; Tang, J.; Yang, Q.; Cui, X.; Li, S.; Chen, S. Performance evaluation and online realization of data-driven normalization methods used in LC/MS based untargeted metabolomics analysis. Sci. Rep. 2016, 6, 38881. [Google Scholar] [CrossRef] [Green Version]
  36. Weljie, A.M.; Newton, J.; Mercier, P.; Carlson, E.; Slupsky, C.M. Targeted profiling: Quantitative analysis of 1H NMR metabolomics data. Anal. Chem. 2006, 78, 4430–4442. [Google Scholar] [CrossRef]
  37. Beleggia, R.; Platani, C.; Nigro, F.; de Vita, P.; Cattivelli, L.; Papa, R. Effect of genotype, environment and genotype-by-environment interaction on metabolite profiling in durum wheat (Triticum durum Desf.) grain. J. Cereal Sci. 2013, 57, 183–192. [Google Scholar] [CrossRef]
  38. Cohen, H.; Shir, O.M.; Yu, Y.; Hou, W.; Sun, S.; Han, T.; Amir, R. Genetic background and environmental conditions drive metabolic variation in wild type and transgenic soybean (Glycine max) seeds. Plant Cell Environ. 2016, 39, 1805–1817. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  39. Dong, Y.S.; Zhao, L.M.; Liu, B.; Wang, Z.W.; Jin, Z.O.; Sun, H. The genetic diversity of cultivated soybean grown in China. Theor. Appl. Genet. 2004, 108, 931–936. [Google Scholar] [CrossRef]
  40. Lee, J.D.; Yu, J.K.; Hwang, Y.H.; Blake, S.; So, Y.S.; Lee, G.J.; Nguyen, H.T.; Shannon, J.G. Genetic diversity of wild soybean (Glycine soja Sieb. and Zucc.) accessions from South Korea and other countries. Crop Sci. 2008, 48, 606–616. [Google Scholar] [CrossRef]
  41. Lee, J.D.; Vuong, T.D.; Moon, H.; Yu, J.K.; Nelson, R.; Nguyen, H.T. Genetic diversity and population structure of Korean and Chinese soybean [Glycine max (L.) Merr.] accessions. Crop Sci. 2011, 51, 1080–1088. [Google Scholar] [CrossRef]
  42. Li, Z.; Nelson, R.L. Genetic diversity among soybean accessions from three countries measured by RAPDs. Crop Sci. 2001, 41, 1337–1347. [Google Scholar] [CrossRef] [Green Version]
  43. Han, O.K.; Abe, J.; Shimamoto, Y. Genetic diversity of soybean landraces in Korea. Korean J. Crop Sci. 1999, 44, 256–262. [Google Scholar]
  44. Li, Y.; Guan, R.; Liu, Z.; Ma, Y.; Wang, L.; Li, L.; Lin, F.; Luan, W.; Chen, P.; Yan, Z.; et al. Genetic structure and diversity of cultivated soybean (Glycine max (L.) Merr.) landraces in China. Theor. Appl. Genet. 2008, 117, 857–871. [Google Scholar] [CrossRef]
  45. Qin, X.; Feng, F.; Li, D.; Herbert, S.J.; Liao, Y.; Siddique, K.H.M. Changes in yield and agronomic traits of soybean cultivars released in China in the last 60 years. Crop Pasture Sci. 2017, 68, 973–984. [Google Scholar] [CrossRef]
  46. Xiong, D.; Zhao, T.; Gai, J. Genetic bases of improved soybean cultivars released from 1923 to 2005 in China—A historical review. Front. Agric. China 2010, 4, 383–393. [Google Scholar] [CrossRef]
  47. Das, A.; Rushton, P.J.; Rohila, J.S. Metabolomic profiling of soybeans (Glycine max L.) reveals the importance of sugar and nitrogen metabolism under drought and heat stress. Plants 2017, 6, 21. [Google Scholar] [CrossRef] [Green Version]
  48. Cox, M.S.; Gerard, P.D.; Wardlaw, M.C.; Abshire, M.J. Variability of selected soil properties and their relationships with soybean yield. Soil Sci. Soc. Am. J. 2003, 67, 1296–1302. [Google Scholar] [CrossRef]
  49. Bellaloui, N.; Hanks, J.E.; Fisher, D.K.; Mengistu, A. Soybean seed composition is influenced by within-field variability in soil nutrients. Crop Manag. 2009, 8, 1–12. [Google Scholar] [CrossRef]
  50. Gao, M.; Li, S. Relationship between soybean yield/quality and soil quality in a major soybean producing area based on a 2D-QSAR model. AIP Conf. Proc. 2017, 1839, 020071. [Google Scholar] [CrossRef] [Green Version]
  51. Wijewardana, C.; Reddy, K.R.; Bellaloui, N. Soybean seed physiology, quality, and chemical composition under soil moisture stress. Food Chem. 2019, 278, 92–100. [Google Scholar] [CrossRef] [PubMed]
  52. Zulak, K.G.; Weljie, A.M.; Vogel, H.J.; Facchini, P.J. Quantitative 1H NMR metabolomics reveals extensive metabolic reprogramming of primary and secondary metabolism in elicitor-treated opium poppy cell cultures. BMC Plant Biol. 2008, 22, 8. [Google Scholar] [CrossRef] [Green Version]
  53. Thonusin, C.; IglayReger, H.B.; Soni, T.; Rothberg, A.E.; Burant, C.F.; Evans, C.R. Evaluation of intensity drift correction strategies using MetaboDrift, a normalization tool for multi-batch metabolomics data. J. Chromatogr. A 2017, 1523, 265–274. [Google Scholar] [CrossRef]
  54. van der Kloet, F.M.; Bobeldijk, I.; Verheij, E.R.; Jellema, R.H. Analytical error reduction using single point calibration for accurate and precise metabolomic phenotyping. J. Proteome Res. 2009, 8, 5132–5141. [Google Scholar] [CrossRef]
  55. Worley, B.; Powers, R. Multivariate analysis in metabolomics. Curr. Metab. 2013, 1, 92–107. [Google Scholar]
  56. Gromski, P.S.; Muhamadali, H.; Ellis, D.I.; Xu, Y.; Correa, E.; Turner, M.L.; Goodacre, R. A tutorial review: Metabolomics and partial least squares-discriminant analysis—A marriage of convenience or a shotgun wedding. Anal. Chim. Acta 2015, 879, 10–23. [Google Scholar] [CrossRef] [PubMed]
  57. Dauwe, R.; Holliday, J.A.; Aitken, S.N.; Mansfield, S.D. Metabolic dynamics during autumn cold acclimation within and among populations of Sitka spruce (Picea sitchensis). New Phytol. 2012, 194, 192–205. [Google Scholar] [CrossRef] [PubMed]
  58. Coutinho, I.D.; Henning, L.M.M.; Döpp, S.A.; Nepomuceno, A.; Moraes, L.A.C.; Marcolino-Gomes, J.; Richter, C.; Schwalbe, H.; Colnago, L.A. Flooded soybean metabolomic analysis reveals important primary and secondary metabolites involved in the hypoxia stress response and tolerance. Environ. Exp. Bot. 2018, 153, 176–187. [Google Scholar] [CrossRef] [Green Version]
  59. Lee, B.-J.; Zhou, Y.; Lee, J.S.; Shin, B.K.; Seo, J.-A.; Lee, D.; Kim, Y.S.; Choi, H.K. Discrimination and prediction of the origin of Chinese and Korean soybeans using Fourier transform infrared spectrometry (FT-IR) with multivariate statistical analysis. PLoS ONE 2018, 13, e0196315. [Google Scholar] [CrossRef] [PubMed]
Foods 10 00435 g001 550
Figure 1. OPLS-DA score plots (A) derived from the 1H-NMR spectra of soybean samples for discriminating geographical origin, Korea and China. Leave-one-out cross-validated score plots (B) showing Korean soybeans (above dashed line) and Chinese soybeans (below dashed line) with threshold value of 0.5 (dashed line) for all samples. In case of misclassified samples, the Ypredcv values were marked with red circles.
Figure 1. OPLS-DA score plots (A) derived from the 1H-NMR spectra of soybean samples for discriminating geographical origin, Korea and China. Leave-one-out cross-validated score plots (B) showing Korean soybeans (above dashed line) and Chinese soybeans (below dashed line) with threshold value of 0.5 (dashed line) for all samples. In case of misclassified samples, the Ypredcv values were marked with red circles.
Foods 10 00435 g001
Foods 10 00435 g002 550
Figure 2. PLS-DA score plots (A) derived from the 1H-NMR spectra of soybean samples for discriminating the geographical origin of three regions of China. Leave-one-out cross-validated score plots for NR vs. MR/SR comparative group (B), MR vs. NR/SR comparative group (C), and SR vs. NR/MR comparative group (D) showing control sample (one region, above dashed line) and case sample (two regions, below dashed line) with threshold value of 0.5 (dashed line) for all samples. In case of misclassified samples, the Ypredcv values were marked with red circles. NR: northeastern region, MR: middle region (Huang-Huai-Hai and Yangtze River region), SR: southern region.
Figure 2. PLS-DA score plots (A) derived from the 1H-NMR spectra of soybean samples for discriminating the geographical origin of three regions of China. Leave-one-out cross-validated score plots for NR vs. MR/SR comparative group (B), MR vs. NR/SR comparative group (C), and SR vs. NR/MR comparative group (D) showing control sample (one region, above dashed line) and case sample (two regions, below dashed line) with threshold value of 0.5 (dashed line) for all samples. In case of misclassified samples, the Ypredcv values were marked with red circles. NR: northeastern region, MR: middle region (Huang-Huai-Hai and Yangtze River region), SR: southern region.
Foods 10 00435 g002
Table
Table 1. Assignment of 1H-NMR (nuclear magnetic resonance) spectral peaks of soybean samples.
Table 1. Assignment of 1H-NMR (nuclear magnetic resonance) spectral peaks of soybean samples.
No.CompoundsChemical ShiftMultiplicity; J ValueAssignmentAssignment Method
12-hydroxyisobutyrate1.34sH-3, H-4One-dimensional proton NMR (1D)/heteronuclear single quantum correlation (HSQC)
22-oxoglutarate2.44t; J = 6.92H-51D/correlation spectroscopy (COSY)/HSQC
2.99t; J = 6.92H-4
3Acetate1.91sH-21D/HSQC
4Acetoacetate2.28sH-41D/HSQC
5Alanine1.47d; J = 7.19H-31D/COSY/HSQC
3.78q; J = 7.19H-2
6Asparagine2.82–2.88mH-21D/COSY/HSQC
2.90–2.98mH-2
4.01q; J = 4.26H-3
7Aspartate2.62dd; J = 8.7, 14.43H-21D/COSY/HSQC
2.79dd; J = 3.78, 13.68H-2
3.91dd; J = 3.75, 4.92H-3
8Choline3.20sH-3, H-4, H-51D/COSY/HSQC
3.48–3.53mH-2
4.03–4.09mH-1
9Citrate2.54d; J = 15.362Ha, 4Ha1D/COSY/HSQC
2.68d; J = 15.362Hb, 4Hb
10Formate8.46sH-11D
11Fumarate6.52sH-2, H-31D
12Galactarate3.94sH-3, H-41D/COSY/HSQC
4.26sH-2, H-5
13Glucose3.22dd; J = 1.44, 7.95H-21D/COSY/HSQC
3.38–3.43mH-4
3.48–3.54mH-5
3.52dd; J = 3.7, 9.82H-2
3.72–3.78mH-3, H-6
3.80–3.85mH-5, H-6
4.62d, J = 7.92H-1
5.22d; J = 3.72H-1
14Glutamate2.00–2.08mH-31D/COSY/HSQC
2.10–2.18mH-3
2.28–2.40mH-4
3.75dd; J = 4.8, 2.4H-2
15Hypoxanthine8.17sH-21D/COSY/HSQC
8.20sH-8
16Isoleucine0.93t; J = 7.15H-51D/COSY/HSQC
1.00d; J = 7.15CH3
1.41–1.49mH-4
1.92–2.01mH-3
3.66d; J = 4.08H-2
17Leucine0.94t; J = 6.06H-5, CH31D/COSY/HSQC
1.64–1.78mH-3, H-4
3.69–3.76mH-2
18Malonate3.13sH-21D/HSQC
19Oxypurinol8.27sH-71D/HSQC
20Raffinose/Stachyose3.52t; J = 4.5H-4′1D/COSY/HSQC
3.69br. sH-6
3.95t; J = 6.36H-5″
4.95dd; J = 2.7, 4.1H-1″
5.41d; J = 4.5H-1
21Succinate2.42sH-2, H-31D/HSQC
22Sucrose3.55dd; J = 3.84, 6.12H-1′1D/COSY/HSQC
3.66sH-1
3.75t; J = 9.05H-3
3.76–3.84mH-6
4.04t; J = 9.05H-4′
5.39d; J = 3.84H-1
23Tartarate4.34sH-2, H-31D/HSQC
24Tryptophan7.20–7.24mH-81D/COSY/HSQC
7.18–7.28mH-9
7.32sH-2
7.71d; J = 7.92H-7
25Valine0.98d; J = 7.02CH31D/COSY/HSQC
1.05d; J = 7.14H-4
2.20–2.32mH-3
3.61d; J = 4.33H-2
s, singlet; d, doublet; dd, doublet of doublet; t, triplet; q, quartet; m, multiplet; br, broad.
Table
Table 2. Parameters of orthogonal partial least squares-discriminant analysis (OPLS-DA) models based on various normalization and scaling methods to discriminate between Korean and Chinese soybean samples.
Table 2. Parameters of orthogonal partial least squares-discriminant analysis (OPLS-DA) models based on various normalization and scaling methods to discriminate between Korean and Chinese soybean samples.
Group No.Normalization MethodScaling MethodNumber of ComponentR2YQ2YR2Y InterceptQ2Y Intercept
1StandardUV40.8440.7620.218−0.430
2Par70.8610.7790.197−0.417
3TotalUV50.8820.7830.254−0.487
4Par60.8620.7980.165−0.385
Number of components obtained from autofit function in SIMCA software; OPLS-DA, orthogonal partial least squares discriminant analysis; Standard, standardized area normalization; Total, total area normalization; UV, unit variance; Par, Pareto; The bold characters indicate the selected optimal model parameters.
Table
Table 3. Classification performance parameters (sensitivity, specificity and accuracy) of OPLS-DA model to discriminate between Korean and Chinese soybeans, and partial least squares-discriminant analysis (PLS-DA) models for discriminating between three regions for Chinese soybeans using leave-one-out cross-validation.
Table 3. Classification performance parameters (sensitivity, specificity and accuracy) of OPLS-DA model to discriminate between Korean and Chinese soybeans, and partial least squares-discriminant analysis (PLS-DA) models for discriminating between three regions for Chinese soybeans using leave-one-out cross-validation.
ClassSensitivity %Specificity %Accuracy %
Korea vs. China96.994.495.6
ChinaNR vs. MR&SR100.0100.0100.0
MR vs. NR&SR100.091.794.4
SR vs. NR&MR100.0100.0100.0
NR: northeastern region, MR: middle region (Huang-Huai-Hai and Yangtze River region), SR: southern region.
Table
Table 4. Parameters of PLS-DA models based on various normalization and scaling methods to discriminate between the different origins of Chinese soybean samples.
Table 4. Parameters of PLS-DA models based on various normalization and scaling methods to discriminate between the different origins of Chinese soybean samples.
Group No.Normalization MethodScaling MethodNumber of ComponentR2YQ2YR2Y InterceptQ2Y Intercept
1StandardUV60.8980.6510.348−0.821
2Par20.4920.3480.082−0.189
3TotalUV30.7310.6220.197−0.345
4Par40.7130.5660.142−0.468
Number of components obtained from autofit function in SIMCA software; PLS-DA, partial least squares discriminant analysis; Standard, standardized area normalization; Total, total area normalization; UV, unit variance; Par, Pareto; The bold characters indicate the selected optimal model parameters.
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Zhou, Y.; Kim, S.-Y.; Lee, J.-S.; Shin, B.-K.; Seo, J.-A.; Kim, Y.-S.; Lee, D.-Y.; Choi, H.-K. Discrimination of the Geographical Origin of Soybeans Using NMR-Based Metabolomics. Foods 2021, 10, 435. https://doi.org/10.3390/foods10020435

AMA Style

Zhou Y, Kim S-Y, Lee J-S, Shin B-K, Seo J-A, Kim Y-S, Lee D-Y, Choi H-K. Discrimination of the Geographical Origin of Soybeans Using NMR-Based Metabolomics. Foods. 2021; 10(2):435. https://doi.org/10.3390/foods10020435

Chicago/Turabian Style

Zhou, Yaoyao, Seok-Young Kim, Jae-Soung Lee, Byeung-Kon Shin, Jeong-Ah Seo, Young-Suk Kim, Do-Yup Lee, and Hyung-Kyoon Choi. 2021. "Discrimination of the Geographical Origin of Soybeans Using NMR-Based Metabolomics" Foods 10, no. 2: 435. https://doi.org/10.3390/foods10020435

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop