Near Infrared Spectroscopy for Determination of the Protein Composition of Rice

Transcription

1 Food Sci. Technol. Res., 14 (2), , 28 Near Infrared Spectroscopy for Determination of the Protein Composition of Rice Flour Jie Yu CHen 1*, Yelian miao 2, Satoshi Sato 1 and Han ZHang 1 1 Faculty of Bioresource Sciences, Akita Prefectural University, Akita 1-195, Japan 2 College of Life Science and Pharmaceutical Engineering, Nanjing University of Technology, Jiangsu 219, China Received June 19, 27; Accepted November 14, 27 Protein content and protein composition are considered very important factors in influencing the cooking and processing characteristics of rice. In the present study, the possibility of using near infrared reflectance spectroscopy (NIRS) to measure the protein composition (prolamin, globulin and glutelin) of rice flour was examined. The NIR spectra (11-25 nm) of a total of 119 rice flour samples with different protein compositions and particle sizes were acquired with a NIR spectrometer. Prediction accuracy of protein content was subsequently examined, revealing a similar accuracy (standard error of prediction (SEP) of.22%) to previous studies. NIR calibration models for determining protein composition were also examined using the partial least square (PLS) regression method. The best models were generated using multiplicative scatter correction-pretreated spectra, giving SEP of.18%, 6% and.25% for prolamin, globulin and glutelin, respectively. The findings show that NIR spectroscopy has the potential to serve as a rapid method for predicting the protein composition of rice flour. Keywords: Near infrared reflectance spectroscopy, Rice flour, Protein composition, Prolamin; Glutelin, Globulin, Determination Introduction Rice is a staple food material in Asian countries, being rich in protein as well as starch. Protein content and protein composition are important factors influencing the quality of cooked rice (Masushige et al. 1994; Furukawa et al. 26) as well as the aroma and flavor of sake (Furukawa et al. 24, 26). In addition, rice flour has recently attracted keen interest as food material, and various new foods using rice flour have been developed, including rice flour bread, rice flour noodles and various rice confectioneries. The protein content and protein composition of rice flour are important factors influencing its processing characteristics as a food material and the quality of these processed foods. Protein analysis of rice is therefore very important. Chemical analysis (the Kieldahl method) is traditionally used for determining rice protein content, and sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) is used for the determination of rice protein composition. However, these methods are *To whom correspondence should be addressed. jiey_chen@akita-pu.ac.jp environmentally unfriendly and time-consuming. Near infrared (NIR) spectroscopic techniques have been used as a rapid and cost-effective analytical way to assess rice quality. A number of studies (e.g., Osborne 26) have reported calibration models for determining the protein content of rice. Delwiche et al. (1995) developed NIR calibration models for protein analysis of ground-milled rice. Delwiche et al. (1996) subsequently extended the application of NIR reflectance spectra to whole-grain milled rice samples, revealing acceptable results of protein content. Barton et al. (1998) reported the five optimal spectral geometries (including transmission and reflectance measurements and shortwavelength NIR) required for development of a NIR model, and in 2 sequentially examined three spectroscopic techniques (NIR, Raman and NMR) for analyzing rice properties, including protein content. Sohn et al. (24a) updated the NIR calibration model for determining protein content using derivative NIR spectra and investigated the optimal derivative condition. Moreover, Sohn et al. (24b) sequentially examined two types of spectroscopic techniques (NIR and FT-Raman) for analysis of rice protein content and amylose.

2 Determination of the Protein Composition of Rice Flour In these studies, the protein content of milled rice and ground rice flour could be measured with very good accuracy using the NIR method. Recently, a spectrum-measuring method for direct application on grain and brown rice was also developed. The grain-measuring method is used for quality control of brown rice, and the flour-measuring method is used for quality control of rice flour as food material. However, neither NIR technique has been applied for the determination of rice protein composition. The present study was designed to investigate the possibility of using the NIR technique as a means for determining the protein composition of rice flour. Moreover, another objective was to examine the measurement accuracy of NIR in determining the protein content of rice flour samples with different protein compositions and particle sizes. Materials and Methods Rice Sample Preparation In order to remove the effects of degree of particle size of rice flour as much as possible, rice flour samples with various particle sizes should be collected. Thus, we collected the rice flour samples with various particle sizes with the following descriptions. Eight kinds of brown rice with different protein compositions (Shyunyou, 133 LGCsoft, Akitakomati, Menkoina, Hitomebore, Akita39, Tatukomoti, and Kinunohada) were purchased from the Ogata Village Country Elevator Public Corporation, Akita Prefecture, Japan. All rice samples were grown at Ogatamura, Akita Prefecture, Japan, in 23. As shown in Figure 1, whole-grain milled rice milled to 9 and 7% was obtained using a laboratory miller (Chioda, HS-4). The whole-grain milled rice, as well as the brown rice samples, were then ground by the Awaji Flour-Milling Company (Makino, DD-2-37). Finally, the rice flour obtained was separated by particle size as shown in Figure 1. A total of 119 rice flour samples with different particle sizes were obtained as study samples. Spectra Collection Spectroscopic collection was performed using a NIR scanning monochromator (system model 65; Foss NIRSystems Inc., Silver Spring, MD) in the reflectance mode. Rice flour samples (about 2 g) were placed in a standard ring cup and then scanned. All spectral data were recorded as the logarithm of the reciprocal of reflectance (log 1/R) in a wavelength range of 11 and 25 nm at 2-nm intervals to give a total of 7 data points per sample. Each spectrum was an average of 32 scans and one spectrum was obtained per sample. All operations were per- Milling Brown rice samples Milling Whole-grain milled rice, degree of milling 9% (based on the brown rice weight) Whole-grain brown rice Whole-grain milled rice, degree of milling 7% (based on the brown rice weight) Grinding Grinding Grinding Rice flour (Degree of milling 9%) Rice flour (Brown rice) Rice flour (Degree of milling 7%) Sorting Sorting Sorting Flour (15-25 µm) Flour (15-25 µm) Flour (15-25 µm) Flour (16-15 µm) Flour (16-15 µm) Flour (16-15 µm) Flour (75-16 µm) Flour (75-16 µm) Flour (75-16 µm) Flour (63-75 µm) Flour (63-75 µm) Flour (45-63 µm) Fig. 1. Preparation procedure for rice flour samples with different particle sizes.

3 134 J. Y. Chen et al. Table 1. Mean, range and standard deviation (SD) of the total protein content and protein composition of the rice flour samples used for calibration and prediction. Calibration (n=8) Prediction (n=39) Mean Range SD Mean Range SD Total protein (%) Prolamin (%) Globulin (%) Glutelin (%) formed at room temperature (25 C). Reference Analysis Total protein content (N 5.95) was determined using the Kjeldahl method with an analyzer unit (23, ACTAC, Japan) in duplicate assays. Replicate results were averaged for data analysis. The standard deviation of differences of duplicate protein content measurements with the Kjeldahl method is.11%. Protein composition was analyzed by SDS-PAGE as described by Laemmli (197) with slight modifications. The SDS-PAGE was run with 15% acrylamide resolving gel and 4.75% acrylamide stacking gel using a powered mini- PAGE system (AE-6531, ATTO, Japan). Rice flour samples were dissolved in 1 ml of sample buffer (4% SDS, 48% urea, 2% glycerol, 5% β-mercaptoethanol, and.5 M Tris- HCl, ph 6.8), heated at 1 C for 5 min, and centrifuged at 1, g to remove particulates. Supernatant aliquots of 1 μl were loaded in each well. Gel electrophoresis was carried out at a constant of 1 V for 1.5 h. A SDS-PAGE molecular weight (MW) standard, composed of a cocktail of proteins (6.5-2 kda) (Bio-Rad, Broad Range, Japan), was also run. The gel was stained with Coomassie Brilliant Blue G25 and destained with pure water. To quantify the protein bands, the polyacrylamide gels were scanned with a calibrated densitometer (Bio-Rad, GS-8, Japan). The MWs of individual protein bands were estimated from the regression line of the protein migration distance versus the logarithm of the MW, and the areas of corresponding peaks were integrated with the 1-D analysis software (Bio-Rad, Quantity One Ver.4.5.2) built into the densitometer. The bands were assigned to proteins according to Iwano et al. (21) and Tanaka et al. (198) as follows: bands of kda were assigned to α-glutelin, 26 kda to globulin, kda to β-glutelin, and 13 kda to prolamin. The glutelin composition was obtained from the combined α- and β-glutelin content. The protein composition of the rice flour samples (prolamin, globulin and glutelin) was obtained by multiplying the fraction of each protein by the total protein content. Chemometric Analysis Samples were divided into calibration and validation sets as follows. Initially, samples within the parent set were sorted by the presence of each chemically-determined component. Starting with the lowest component sample, the first and third samples were assigned to the calibration set, and the second sample to the validation set. The next group of three samples was similarly assigned, and so on, until the last group. Thus, 8 samples were included in the calibration set and 39 in the validation set. The protein content and composition (prolamin, glutelin and globulin) of the rice flour samples are shown in Table 1. The calibration and validation sample sets showed similar means and standard deviations. The calibration set was then used to establish calibration models (multivariate equations) with the spectral data and laboratory reference values, while the validation set was used to evaluate the resulting models. The partial least square (PLS) regression method (Martens and Naes 1989) was used to develop calibration models for determining the protein content and protein composition of the rice flour samples based on calibration sample set. The PLS calibrations were performed with Unscrambler 7.6 (CAMO, Oslo, Norway). The optimum number of factors required to minimize overfitting was based on the standard error of cross validation (SECV). The validation sample set was then used to test the performed calibration. The standard error of prediction (SEP) and the correlation coefficient of the reference values versus the NIR values (Rpred) were calculated. Pretreatment of the spectra included second derivative (Savitzky and Golay 1964) and multiplicative scatter correction (MSC) (Martens and Naes 1989) using the Unscrambler software. Results and Discussion Development of PLS Calibrations The correlation between the protein content and each protein component was investigated using all rice flour samples, because rice flour samples with different protein composition are necessary in order to comprehensively analyze the possibility of NIR measurement of the protein composition. Table 2 shows the correlation coefficient between the protein content and

4 Determination of the Protein Composition of Rice Flour Table 2. Correlation coefficients between total protein content and protein component of rice flour samples. Protein Prolamin Glutelin Globulin Protein Prolamin Glutelin Globulin each protein component. Each correlation coefficient was relatively low, except for the correlation coefficient between prolamin and globulin. In other words, the rice flour samples used in this study had different protein compositions and were appropriate for investigating the possibility of NIR measurement of the protein components, such as prolamin and glutelin, of rice flour. The spectra of rice flour samples with different particle sizes are represented in Figure 2, which shows the spectrum signals before and after second derivative and MSC pretreatment. The main absorption bands were observed at wavelengths of 1466, 1934, 214, 2288 and 2322 nm. The 1466-nm band was related to the O-H (starch) and N-H (protein) stretch first overtone (Osborne et al. 1993), the 1934-nm band to the O-H bend second overtone related to water, the 214-nm band to the O-H and C-O bend second overtone related to starch, and the and 2322-nm bands to the N-H stretch first overtone related to protein and C-H bend second overtone related to oil, respectively. The raw spectra also presented a large vertical shift from baseline (Figure 2a), affecting the entire spectral region, due to differences in the scattering behavior of the samples with different particle sizes. As can be seen, this baseline shift could not be effectively prevented by second derivative pretreatment (Figure 2b), but was successfully eliminated by MSC pretreatment (Figure 2c). Table 3 shows the results of the calibration models obtained with the PLS regression method calculated using the raw spectra, second derivative spectra, and MSC-pretreated spectra. Regarding the total protein model, the results obtained from the second derivative spectra were very similar and showed no improvement compared to those obtained from the raw spectra. However, the MSC-pretreated spectra produced clearly better results than both the raw spectra and second derivative spectra. The correlation coefficient of calibration (R cal ), standard error of calibration (SEC), standard error of cross-validation (SECV), correlation coefficient of prediction (R pred ) and standard error of prediction (SEP) were.99,.18%,.21%,.99 and.22% in the total protein model, respectively. This result is graphically depicted in Log (1/R) d 2 log (1/R) µm µm µm µm µm Log (1/R) µm µm µm µm µm µm µm µm µm µm (a) (b) (c) Fig. 2. Raw NIR spectra (a), spectra after second derivative (b) and MSC (c) correction of rice flour samples with different particle sizes. 135 Figure 3a in a plot of actual versus predicted values. The predicting error for total protein, a SEP of.22%, was similar to the results obtained by Sohn et al. (24a and 24b), demonstrating that the total protein content of rice flour samples with different particle sizes could be satisfactorily estimated using the NIR method. Regarding the compositions of prolamin, globulin and glutelin, satisfactory PLS calibration models were obtained as shown in Table 3. Similar to the total protein content, derivative treatment resulted in fewer factors, but did not effectively improve the prediction accuracy of the calibration models. Regarding glutelin, MSC processing produced calibration and prediction statistics superior to those obtained with the raw spectra and second derivative spectra. However,

5 136 J. Y. Chen et al. Table 3. NIR calibration and prediction results for the total protein content and protein composition of rice flour samples. Pretreatment Factors R cal SEC(%) SECV(%) R pred SEP(%) Bias(%) RPD raw Total protein D MSC raw Prolamin D MSC raw Globulin D MSC raw Glutelin D MSC raw, none; D2, second derivative spactra; MSC,multiplicative scatter correction; R cal, correlation coefficient of calibration; SEC, standard error of calibration;secv, standard error of cross-validation; R pred, correlation coefficient of prediction; SEP, standard error of prediction; RPD, ratio of the standard deviation of the reference data in the prediction sample set to SEP R=.99 SEP=.22 % Bias= 1 % (a) Total protein (%) 75-16µm 16-15µm 15-25µm R=.94 SEP=.18 % Bias= % 75-16µm µm (b) 15-25µm Prolamin (%).8.6 R=.93 SEP= 6 % Bias= % µm 16-15µm (c) 15-25µm Globulin (%) R=.97 SEP=.25 % Bias= 3 % µm 16-15µm (d) 15-25µm Glutelin (%) Fig. 3. Scatter plots of actual and predicted protein content and protein composition of rice flour samples. regarding the compositions of prolamin and globulin, MSC processing did not effectively improve the accuracy. The R pred and SEP were.94 and.18% for prolamin,.93 and 6% for globulin, and.97 and.25% for glutelin; these results are graphically depicted in Figures 3b, c and d, respectively. To determine the usefulness of the calibration models, the ratio of the standard deviation of the reference data of the prediction sample set to the SEP (RPD) is usually utilized (Williams 21; Chen et al. 25, 26). A RPD value of more than 2.5 is regarded as adequate for screening. In this study, RPD values of more than 2.5 were obtained, suggesting that the calibration models are suitably accurate for measurement of the protein content and protein composition of rice flour. Discussion of NIR Calibration Models can be used to discuss the contributions of individual

6 Determination of the Protein Composition of Rice Flour wavelengths to a PLS calibration model, since a regression coefficient spectrum shows characteristic peaks and troughs that can indicate which wavelength range is important for the calibration model (Martens et al. 1989; Chen et al. 23, 24). Figure 4 shows the regression coefficients of the PLS calibration models of total protein (a), prolamin component (b), globulin component (c) and glutelin component (d), respectively. The regression coefficient spectrum for total protein (Figure 4a) showed many remarkable peaks related to protein at wavelengths of 143, 1982, 252 and 2292 nm (Osborne 1993). The peak at wavelength of 143 nm could be correlated to the absorption band of N-H stretching first overtone. The peaks at wavelengths of 1982 and 252 nm could be correlated to the combined absorption band of N-H asymmetrical stretching and amide II mode corresponding to the wavelengths of 198 and 25 nm. The peak at wavelength of 2292 nm could be correlated to the combined absorption band of N-H stretching and C=O stretching mode corresponding to the wavelength of 2294 nm related to amino acid. The other regression coefficient spectra for protein composition (Figures 4b, c and d) showed some remarkable peaks related to protein at wavelengths of 1456, 146, 15, 1522, 1958, 196, 2, 248, 2128, 217, 2292 and 23 nm. The peaks at wavelengths of 1456, 146, 15 and 1522 nm could be correlated to the absorption band of N-H 137 stretching first overtone corresponding to the wavelengths of 146, 15 and 152 nm. The peak at wavelengths of 1958, 196, 2 and 248 nm could be correlated to the combined absorption band of N-H asymmetrical stretching and amide II mode corresponding to the wavelengths of 196, 2 and 25 nm. The peak at wavelength of 217 nm could be correlated to the combined absorption band of two amide I and amide III modes. The peak at wavelengths of 2128, 2292 and 23 nm could be correlated to the combined absorption band of N-H stretching and C=O stretching mode corresponding to the wavelengths of 2132 and 2294 nm related to amino acid. Taken together, these results suggest that the PLS calibration models for protein composition of rice flour were established based on the absorptions of protein. The above results showed the possibility of using the NIR technique as a means for determining the protein composition of rice flour. However, further research is needed to examine the best spectrum measuring method and spectral range for practical use, as they were not examined in this paper. Conclusions NIR reflectance spectroscopy of rice flour with different particle sizes could be used to measure total protein content with similar accuracy (SEP.22%) to previous studies. Moreover, measurement of the protein composition (pro (a): total protein (b): prolamin (c): globulin (d): glutelin Fig. 4. in the PLS calibration models based on MSC pretreated spectra (11-25 nm) for the protein content and protein composition of rice flour.

7 138 lamin, globulin and glutelin) could be accomplished with NIR calibration models developed using the PLS regression method with SEPs of.18%, 6% and.25% for prolamin, globulin and glutelin, respectively. NIR spectroscopy therefore has the potential to serve as a rapid method for predicting the protein composition of rice flour. References Barton-II, F. E., Windham, W. R., Champagne, E. T., and Lyon, B. G. (1998). Optimal geometries for the development of rice quality spectroscopic chemometric models. Cereal Chem., 75(3), Barton,II, F. E., Himmelsbach, D. S., Mcclung, A. M., and Champagne, E. T. (2). Rice quality by spectroscopic analysis: Precision of three spectral regions. Cereal Chem., 75(5), Chen, J. Y.; Matsunaga, R., Ishikawa, K. and Zhang, H. (23). Main inorganic component measurement of seawater using nearinfrared spectroscopy. Applied Spectroscopy., 57(11), Chen, J. Y.; Miao, Y.; Zhang, H. and Matsunaga, R. (24). Nondestructive determination of carbohydrate content in potatoes using near infrared spectroscopy. J. Near Infrared Spectroscopy., 12, Chen, J. Y., Zhang, H., Miao, Y., and Matsunaga, R. (25). NIR measurement of specific gravity of potato. Food Sci. Technol. Res., 11(1), Chen, J. Y., Zhang, H., and Matsunaga, R. (26). Rapid determination of the main organic acid composition of raw Japanese apricot fruit juices using near-infrared spectroscopy. J. Agric. Food Chem., 54(23), Delwiche, S. R., Bean, M. M., Miller, R. E., Webb, B. D., and Williams, P. C. (1995). Apparent amylase content of milled rice by near-infrared reflectance spectrophotometry. Cereal Chem., 72(2), Delwiche, S. R., Mckenzie, K. S., and Webb, B. D. (1996). Quality characteristics in rice by near-infrared reflectance analysis of whole-grain milled samples. Cereal Chem., 73(2), Furukawa, S., Mizuma, T., Kiyakawa, Y., Iida, S., Matsushita, K., Maeda, H., Sunohara, Y., and Wakai, Y. (24). Improvement in quality of sake brewed from low-glutelin rice for sake brewing. Seibutsukogaku 82: In Japanese. J. Y. Chen et al. Furukawa, S., Tanaka, K., Masumura, T., Ogihara, Y., Kiyakawa, Y., and Wakai, Y. (26). Influence of rice proteins on eating quality of cooked rice and aroma and flavor of sake. Cereal Chem., 83(4), Iwano, K., Nakazawa, N., Ito, T., Takahashi, H., Uehara, Y., and Matsunaga, R. (21). The influence of protein components in raw material rice on various enzyme activities in sake-koji. J. Brew. Soc. Japan., 96(12), In Japanese. Laemmli, U. K. (197). Cleavage of structural protein during the assembly of the head of bacteriophage T4. Nature, 227, Martens, H., and Naes, T. (1989). Multivariate Calibration. John Wiley and Sons: Chichester, UK. Masushige, H., Hirai, N., Masumura, T., and Tanaka, K. (1994). Relationship between contents, distribution of proteinbody I, II and the cooked rice taste. Breeding Sci., 44, 238. In Japanese. Osborne B. G., Fearn, T., and Hindle, P.H. (1993). Practical NIR Spectroscopy: With Applications in Food and Beverage Analysis. John Wiley & Sons, Inc., New York Osborne, B. G. (26). Application of near infrared spectroscopy in quality screening of early-generation material in cereal breeding programmes. J. Near Infrared Spectrosc., 14, Sohn, M., Barton-II, F. E., McClung, A. M., and Champagne, E. T. (24a). Near-infrared spectroscopy for determination of protein and amylase in rice flour through use of derivatives. Cereal Chem., 81(3), Sohn, M., Himmelsbach, D. S., and Barton-II, F. E. (24b). A comparative study of fourier transform raman and NIR spectroscopic methods for assessment of protein and apparent amylase in rice. Cereal Chem., 81(4), Savitzky, A., and Golay, M. J. E. (1964). Smoothing and differentiation of data by simplified least squares procedures. Anal. Chem., 36, Tanaka, K., Sugimoto, T., Ogawa, M., and Kasai, Z. (198). Isolation and characterization of two types of protein bodies in the rice endosperm. Agric. Biol. Chem., 44(7), Williams, P. C. (21). Near-infrared technology in the agricultural and food industries, In: Implementation of Near-Infrared Technology, edited by P. Williams and K. Norris, American Association of Cereal Chemists Inc., Minnesota, USA. pp