ISSN : 2287-5174(Online)
Estimating the Amino Acid Composition in Three Different Sample Types of Soybean by Near Infrared Spectroscopy (NIRS)
The needs of consumers for agricultural products are now widely diversified with attention focused not only on the major constituents, but also on its other physiological functions. Soybean (Glycine max) is one of the world’s most important sources of nutritional and functional components such as protein, amino acids and isoflavone.
Amino acids such as lysine and tryptophan are essential for human and monogastric nutrition (Wu, et al. 2002). The lack of methionine, lysine, threonine, and other essential amino acids can limit the nutritional efficiency of the feed (Fontaine, et al. 2002). Thus the modification of the amino acid profile in soybean seed for high nutritional values and successful feed compounding is one of the most important objectives in quality breeding programs in soybean. Chromatographic amino acid analysis requires oxidation and hydrolysis of the protein followed by ion exchange chromatography. This wet chemical procedure, however, is quite complicated, expensive and time-consuming.
Near infrared spectroscopy (NIRS) is a rapid and nondestructive technology that does not require chemicals or reagents, It is a multi-element technique, and sample measurement using this technique is being successfully implemented throughout the agricultural industry (Schmilovitch, et al. 2000). Since the late 1980s, NIRS has been used for measuring the internal composition of biological materials (Fontaine, et al. 2001). Most published studies show that NIRS can accurately estimate the content of several internal components of plants such as dry matter and soluble solids (Lee, et al. 2009; Lee, et al. 2010), fatty acids (Velasco, et al. 1999; Kavalenko, et al. 2006), carotenoids (Berardo, et al. 2009) as well as several inorganic components (Petisco, et al. 2005). Recently, the scope of NIRS has been expanded to include the determination of the physiological indices of crops (Jeong, et al. 2008) or the classification of normal and artificial aged corn (Min and Kang. 2008), and the evaluation of product quality of herbal medicinal extract (Mohri, et al. 2009). For soybean, protein (Choung, et al. 2001), amino acids (Fontaine, et al. 2001), fatty acids (Igne, et al. 2008; Kovalenko, et al. 2006), anthocyanin (Kim, et al. 2008) and isoflavone (Sato, et al. 2008) were also predicted using NIRS.
In grain-related NIR research for the analysis of amino acid, Williams et al. (1984) reported satisfactory results in determination coefficient (R2) ranging from 0.66 to 0.96 in correlating NIR spectral data of ground wheat and barley to their amino acid concentrations. Wu et al. (2002) showed the applicability of NIR spectroscopy for amino acid analysis of milled rice powder. In their study, calibration models had high determination coefficients (R2 = 0.85-0.98), except for those of cysteine, histidine, and methione.
In soybean, NIRS is currently used for the analysis of whole seeds and ground seeds powder for amino acid (Fontaine et al. 2001; Igne, et al. 2008). The experiments conducted by them demonstrated that the predictive ability of amino acid calibration models was dependent on sample type (whole grain or ground). In addition, as suggested by Fontaine et al. (2001), the accuracy of NIR screening for amino acid concentrations in soybeans may be improved by grinding grain samples.
However, available NIRS instruments using whole seeds require a 300-500 g sample size to operate (Downey, 1994). NIRS analyses using ground seeds powder is destructive of the seed. Additionally, significant time and technique are required to grind seeds. Near-infrared analyses of intact seeds are commonly made on bulk samples of variable size, depending on the instrument and device used. Furthermore, instruments including single-seed sample holders are available (Downey, 1994). A further improvement in the application of NIRS technique would imply the minimization of sample size and the reduction of destructive use of samples, time and labor in measuring of samples. The analysis of single seeds for the important components using the non-destructive method is the more appropriate way when dealing with soybean breeding.
The objective of this work is to find the best sample types for the more accurate prediction and non-destructive way of using the NIRS technique in estimating amino acids in soybean by comparing three different sample types, single seed, whole seeds, and milled seeds powder.
MATERIALS AND METHODS
This work was conducted with 143 seed samples (individual accessions) of soybean. The seed materials, conserved at the National Agrobiodiversity Center (genebank), Rural Development Administration (RDA), Korea were selected from different geographical origins in Korea, China, Japan, and other subcontinents. The moisture content of seed materials was controlled to 5-7% following the regulation for seed conservation of medium-term storage in the National Agrobiodiversity Center.
The three different sample types, single seed, whole seeds (10-12 seeds) and milled seeds powder (0.5~0.6 g) were analyzed.
The analysis conditions of soybean by near infrared spectroscopy system are summarized in Table 1. Single seed samples were analyzed following the method of Font, et al. (2004) using the NIRSystems model 6500 spectrophotometer (Foss-NIRSystems, Inc., Silver Spring, MD) equipped with a DCFA (Direct Contact Food Analyzer) module in the reflectance mode. Intact single seed was placed on sample holder with a diameter of 3 mm. Twenty single seeds per one accession were scanned and their average spectra were recorded as individual files at 2 nm intervals in the 400~2,500 nm wavelength range. Whole seeds samples were analyzed using the round cup (outer diameter 5 cm, inner diameter 3.5 cm) with a quartz window in the same NIRSystem with a spinning module. Milled seeds powder sample type was grounded with a ball mill and sieved with a 1.0 mm. Same seed materials are scanned by NIRS as three different sample types.
The parameters of NIRs for equation statistics and calibration are summarized in Table 2. using the program WINISI II v.1.50 (Infrasoft International, LLC, Port Matilda, PA), different calibration equations for three sample types and 17 different amino acids were developed on the calibration set (n=143). Calibration equations were computed using the first or second derivative of raw optical data (log1/R, where R is reflectance) with several combinations of segment (smoothing) and derivative (gap) sizes [i.e., (1,4,4,1 which means 1 = number of derivative of spectra, 4 = extent of gap over which the derivative was to be calculated, 4 = the smoothing of point, 1 = second smoothing), (1,8,8,1), (2,4,4,1), (2,10,10,1)]. These parameters in the mathematical processing for different sample types and raw data for 17 amino acids components were sought through trial and error in order to maximize the coefficient determination in cross validation and to minimize the standard error of cross validation.
To correlate the spectrum data and the contents of the parameters, modified partial least squares (MPLS) was used as a regression method in the wavelength range 400~2,500 nm. In addition, standard normal variate and detrend transformations (SNV-DT) were used to correct baseline offset due to scattering effects from differences in particle size among samples. The calibration equation were optimized by removing outliers for samples using the following criteria: samples with large residuals values (the difference between the predicted and the actual values) and T-outliers (T>2.5) or H-outliers (H>3).
The different calibration equations obtained in the calibration process were then cross-validated using an internal validation set (25% of total samples randomly taken by software routine) that included outliers removed from the calibration set. The calibration statistics used were standard error of calibration (SEC), standard error of cross-validation (SECV), and coefficient determination of cross-validation (1-VR). The SEC and SECV were calculated as the square root of the mean square for the residuals on the calibration set spectra and cross-validation set, respectively. The SECV is used as the statistic for determining the best number of independent variables for the calibration equation. The prediction ability of each of the calibration equations was determined based on the coefficient of determination on internal validation (R2) and standard error of prediction (SEP).
Table 1. The analysis conditions of soybean by near infrared spectroscopy system (NIR 6500, USA).
Table 2. Parameter of NIRs for equation statistics and calibration.
Assay of amino acids
Each milled powder sample that was scanned by NIRS was used for chemical analysis.
About 0.3 g of each sample was weighed and 40 ml of 6 N-HCl was added. The hydrolysis was maintained for 24 h at 110℃ in test tubes with screw caps under a nitrogen atmosphere. Afterwards, the acid was evaporated from all samples using a rotary evaporator. Samples were washed three times with distilled water, and the residue was redissolved in sodium citrate buffer (pH 2.2). After dilution to a known volume, the hydrolysate was filtered with filter paper (Whatmann No. 5). About 1 ml of each filtered hydrolysate was put into an auto-sampler bottle and injected into L-8500A high-speed amino acid analyzer (Hitachi, Japan). The amount of each amino acid in the samples was calculated with reference to the standard solution. The standard 17 amino acid solutions were obtained from Wako (Wako-shi, Japan).
RESULTS AND DISCUSSION
Variation amino acid in soybean
Seventeen different amino acids were detected by HCl hydrolysis-HPLC method. Mean values, ranges and standard deviations (SD) for amino acids in the samples used in the calibration set are shown in Table 3. One hundred forty three samples were detected 17 amino acids and showed low contents of cysteine and methionine, respectively. This result was similar to other studies on soybean (Fontaine, et al. 2001; Kovalenko, et al. 2006). In the sample set, there was a comparatively wide variation in the fifteen amino acids in a comparison to the variation of other cereals such as wheat, barley and corn, making the population suitable for NIRS calibration (Fontanine, et al. 2002). But the samples varied less widely than the variability reported in the literature of soy (Fontanine, et al. 2001). The means for different amino acids with the exceptions of cystein and methionine in the calibration sample set were between 1.21 and 7.45%.
Statistics NIRS calibration
Table 4 to 6 summarizes the performance parameters obtained for the calibration equations by using single seed, whole seeds and ground seeds. Over-all results showed that the R2 of NIR calibration and 1-VR of cross-validation developed from samples scanned using milled powder were highest, followed by single seed, and then whole seeds were the lowest. The NIRS calibration equations for 11 other amino acids with exception of six amino acids (cysteine, methionine, tryptophan, phenylalanine, lysine, and histidine) in milled powder also showed high coefficient of determination (0.82-0.95) (Table 6). The R2 of calibration and 1-VR of cross-validation for other ten different amino acids with exception of cysteine, methionine, valine, tryptophan, phenylalanine, lysine, and histidine developed from samples scanned using single seed were not poor coefficient of determination with R2 0.80-0.89 and 0.62-0.70, respectively (Table 4). On the other hand, calibration equation developed with NIRS data scanned with whole seeds showed the lowest accuracy and reliability compared with the other sample groups (Table 5). The only calibration equation development for glutamic acid analyzed by NIRS data scanned with whole seeds showed higher than 0.8 R2 coefficient.
In general, most of the amino acid calibration models of this experiment developed from samples scanned using whole seeds and milled seeds powder were higher than those previously reported by Pazdermik et al. (1997). In their study used a total of 90 NIRS calibration samples, R2 values of 0.06-0.83 and 0.38-0.85 were obtained for whole seeds and ground seeds soybean samples, respectively. This could likely be attributed to a much larger calibration set used in this study (143 vs 90 samples)
It is a normal finding that the standard errors obtained by cross-validation are slightly higher and the 1-VR values slightly lower than those parameters of the calibration (Wu, et al. 2002). In our results, the standard errors obtained by cross-validation were slightly higher than those parameters of calibration. However, the 1-VR values were much lower than the parameters of calibration. The calibration models for cysteine, methionine, phenylalanine and histidine obtained using single bean and milled powder showed very low values. Because of the very low values, these calibrations were not suitable to estimate those amino acids.
Raw NIRS spectrum for 143 samples scanned using three different sample types are shown in Figure 1. In the infrared wavelength range (700-2,500 nm), the width of data spectrum for milled powder sample was the widest, followed by the single bean, and that of whole beans was the thinnest. These results were similar to those reported by other authors who experimented on the whole seeds and ground powder of cereals. Tkachuk, et al (1987) have compared whole pea to ground powder to predict protein more accurately and stated ground pea performed much better than the whole pea. They also stated that the reason why the higher SEC and SEP values and the lower RSQ values for whole pea prediction may be attributed in large part to sampling error arising from the large particle size of whole peas. Agricultural products selectively absorb NIR radiation and then yield information about the molecular bonds within the material being measured. Because NIR radiation are scattered between the large particle size of whole soybeans so that the accurate information about material could not be produced from whole soybeans, no satisfactory results from whole seeds samples in this research may be obtained. The R2 coefficients of single soybean perimeter were higher than results of whole soybeans and showed higher RSQ coefficient than 0.8 with the exception of seven amino acids, which suggests that satisfactory accuracy of NIR predictions may be achieved without grinding the seed samples.
The present findings suggest that eleven different amino acids, such as aspartic acid, threonine, serine, glutamine, glycine, alanine, valine, isoleucine, leucine, arginine and proline, of soybean seeds in the powder could be estimated for the most accurate prediction method. The study also showed that the use of R2 coefficient of single seed perimeter in the application of NIRS technology can make possible the use of nondestructive sampling results to a more accurate and faster method of prediction and analysis. This is a useful way as the method of analysis in determining the desired characteristics in the soybean seeds to be used for breeding purposes.
Table 3 Laboratory reference value statistics of milled soybean powder for the NIRs calibration set.
Table 4. Statistics of calibration, cross-validation and prediction validation in NIRs equations for contents of amino acids in soybean with the single soybean sample type.
Table 5 Statistics of calibration, cross-validation and prediction validation in NIRs equations for contents of amino acids in soybean with the whole soybeans sample type.
Table 6 Statistics of calibration, cross-validation and prediction validation in NIRs equations for contents of amino acids in soybean with the milled soybean powder sample type.
Fig. 1 Raw NIRS spectrum for 143 samples scanned using three different sample types.
This work was supported by a grant from the National Academy of Agricultural Science (PJ00855303), Rural Development Administration, Rep. of Korea.
2.Choung MG, Baek IY, Kang ST, Han WY, Shin DC, Moon HP, Kang KH. 2001. Determination of protein and oil contents in soybean seed by near infrared reflectance spectroscopy. Korean J. Crop Sci. 46: 106-111.
3.Downey, G., 1994. Grain analysis by NIRS: Is the harvest in?. In: G.D. Batten, P.C. Flinn, L.A. Welsh & A.B. Blakeney (Eds), Leaping Ahead with Near Infrared Spectroscopy. Royal Australian Chemical Institute, Lorne, Australia. pp.136-147.
4.Font R, Rio MD, Fernandez-Martinez JM, de Haro-Bailon A. 2004. Use of near-infrared spectroscopy for screening the individual and total glucosinolate contents in Indian mustard seed (Brassica juncea L. Czern. & Coss.). J. Agric. Food Chem. 52: 3563-3569.
5.Fontaine J, Hrr J, Schirmer B. 2001. Near-infrared reflectance spectroscopy enables the fast and accurate prediction of the essential amino acid contents in soy, rapeseed meal, sunflower meal, peas, fishmeal, meat meal products, and poultry meal. J. Agric. Food Chem. 49: 57-66.
6.Fontaine J, Schirme B, Horr J. 2002. Near-infrared reflectance spectroscopy enables the fast and accurate prediction of essential amino acid contents. 2. Results for wheat, barley, corn, triticale, wheat bran/middlings, rice bran, and sorghum. J. Agric. Food Chem. 50: 3902-3911.
7.Igne B, Rippke GR, Hurburgh CR. 2008. Measurement of whole soybean fatty acids by near infrared spectroscopy. J. Am. Oil Chem. Soc. 85: 1105-1113. http://dx.doi.org/10.1007/s11746-008-1311-1
8.Jeong JC, Ok HC, Hurr OS, Kim CG. 2008. Prediction of sprouting capacity using near-infrared spectroscopy in potato tubers. Am. J. Pot. Res. 89: 309-314. http://dx.doi.org/10.1007/s12230-008-9010-x
9.Kim YH, Ahn HK, Lee ES, Kim HD. 2008. Development of prediction model by NIRS for anthocyanin contents in black colored soybean. J. Crop. Sci. 53: 15-20.
10.Kovalenko IV, Rippke GR, Hurburgh CR. 2006. Measurement of soybean fatty acids by near-infrared spectroscopy: linear and nonlinear calibration methods. J. Am. Oil Chem. Soc. 83: 421-427.
11.Lee SY, Lee YY, Lee SK, Cho YH, Ma KH, Kang HK, Gwag JG, Park KH, Lee HS. 2009. Selecting high amylose lines of rice varieties using NIR spectroscopy at the RDA gene bank conserved. Kor. J. Breed. Sci. 41: 108.
12.Lee YY, Kim JB, Lee SY, Lee HS, Gwag JG, Kim CK, Lee YB. 2010. Application of near-infrared reflectance spectroscopy to rapid determination of seed fatty acids in foxtail millet (Setaria italica (L.) P. Beauv) germplasm. Kor. J. Breed. Sci. 42: 448-454.
13.Min TG, Kang WS. 2008. Nondestructive classification between normal and artificially aged corn (Zea mays L.) seeds using near infrared spectroscopy. J. Crop Sci. 53: 314-319.
14.Mohri T, Sakata Y, Otsuka M. 2009. Quantitative evaluation of glycyrrhizic acid that affects the product quality of kakkonto extract, a traditional herbal medicine, by a chemometric near infrared spectroscopic method. J. Near Infrared Spectrosc. 17: 84-100.
15.Pazdermik DL, Killam AS, Orf JH, 1997. Analysis of amino and fatty acid composition in soybean seed using near infrared reflectance spectroscopy. Agro. J. 89: 679- 685.
16.Petisco C, Garcia-Criado B, Vazquez de Aldana BR, Zabalgogeazcoa I, Mediavilla S, Garcia-Ciudad A. 2005. Use of near-infrared reflectance spectroscopy in predicting nitrogen, phosphorus and calcium contents in heterogeneous woody plant species. Anal. Bioanal Chem. 382: 458-465.
17.Sato T, Eguchi K, Hatano T, Nishiba Y. 2008. Use of near-infrared reflectance spectroscopy for the estimation of the isoflavone contents of soybean seeds. Plant Prod. Sci. 11: 481-486.
18.Schmilovitch Z, Mizrach A, Hoffman A, Egozi H, Fuchs Y. 2000. Determination of mango physiological indices by near-infrared spectrometry. Postharv. Biol. Technol. 19: 245-252.
19.Tkachuk R, Kuzina FD, Reichert RD. 1987. Analysis of protein in ground and whole field peas by near-infrared reflectance spectroscopy. Cereal Chem. 64: 418-422.
20.Velasco L, Christian M, Heiko C. 1999. Estimation of seed weight, oil content and fatty acid composition in intact single seeds of rapeseed (Brassica napus L.) by near-infrared reflectance spectroscopy. Becker. Euphytica. 106: 79-85.
21.Williams PC, Preston KR, Norris KH, Starkey PM. 1984. Determination of amino acids in wheat and barley by near- infrared reflectance spectroscopy. J. Food Sci. 49: 17-20.
22.Wu JG, Shi C, Zhang X. 2002. Estimating the amino acid composition in milled rice by near-infrared reflectance spectroscopy. Field Crops Research. 75: 1-7.