This paper develops a rapid method using near infrared (NIR) spectroscopy for analyzing the antioxidant activity of brown rice as total phenol content (TPC) and radical scavenging activity by DPPH (2,2-diphenyl-2-picrylhydrazyl) expressed as gallic acid equivalent (GAE). Brown rice (
As one of the oldest domesticated grains, rice serves as the staple food for half of the global population. China is the first leading producer of rice in the entire world [
The contents of antioxidative compounds in brown rice mainly depend on the varieties, cultivation conditions, and processing. It is well known that long-term storage can degrade the quality of brown rice because of oxidation. In China, for economic reasons, it is profitable to sell degraded or fraud brown rice by extracting pigments and other active components before putting them on the supermarket shelf. Therefore, it is necessary to develop a rapid and reliable method for determining the quality of brown rice, especially its antioxidant activity.
Traditional methods for evaluating antioxidant activity generally can fall into two classes: direct determination of antioxidant capacity and determination of the levels of the main antioxidant components [
Compared with traditional methods, NIR spectroscopy has many advantages including less sample preparation, reduced analysis time and cost. Therefore, NIR has been widely used for rapid analysis of antioxidant activity in various food products [
Brown rice samples were collected from domestic markets and the producing areas are Yunnan (25), Hunan (21), Sichuan (27), Guizhou (25), and Shanxi (23). All the 121 samples are harvested and analyzed in 2013. All the samples were cleaned and stored at 25°C before analysis.
The method in [
A modified version of the Folin-Ciocalteu assay [
The DPPH (2,2-diphenyl-1-picrylhydrazyl) radical scavenging activity of the extracts was measured using the method in [
The NIR diffuse reflectance spectra of all the samples were collected in 4000–12000 cm−1 on a Bruker-TENSOR37 FTIR system (Bruker Optics, Ettlingen, Germany) using OPUS software. All the spectra were measured with a PbS detector and an internal gold background as the reference. The resolution was 4 cm−1 and the scanning interval was 1.929 cm−1. Therefore, each spectrum had 4148 wavelengths. The scanning number was 64, because more scans did not reduce the signal-to-noise ratio significantly.
To obtain a set of representative objects for training/validating the calibration models, the DUPLEX algorithm [
The quantitative modeling was performed using PLS and iPLS [
The content ranges reference values of antioxidant activity in calibration samples were 1.0933–6.0464 for TPC and 0.7833–5.8125 (mg GAE/g) for DPPH, respectively. The content ranges of TPC and DPPH in validation samples were 1.2135–5.9406 and 0.8121–5.4903 (mg GAE/g), respectively. This indicates that the DUPLEX method can split the data properly for calibration and validation. To reduce the unwanted spectral variations, three preprocessing methods, including smoothing, taking second-order derivative (D2) [
Results of full-spectrum PLS (FS-PLS) models for predictions of antioxidant activity of brown rice.
TPC (mg GAE/g) | DPPH (mg GAE/g) | |||||
---|---|---|---|---|---|---|
RMSECVa | RMSEPb | LVsc | RMSECV | RMSEP | LVs | |
Raw data | 0.242 | 0.303 | 7 | 0.269 | 0.283 | 8 |
Smoothing | 0.196 | 0.232 | 6 | 0.243 | 0.244 | 9 |
D2 | 0.191 | 0.209 | 6 | 0.233 | 0.219 | 8 |
SNV | 0.162 | 0.167 | 6 | 0.199 | 0.211 | 7 |
bRMSEP: root mean squared error of prediction.
cNumber of PLS components.
The raw NIR spectra and the smoothed, second-order derivative (D2), and standard normal variate (SNV) spectra of brown rice objects.
For iPLS models, the total spectral range was sequentially (from 4000 cm−1 to 12000 cm−1) split into 20 spectral intervals with an equal width of 400 cm−1. At each interval, a PLS model is built to predict TPC and DPPH values. The numbers of PLS components were determined to obtain the lowest RMSECV. With each data preprocessing method, three intervals with the lowest RMSECV values were selected and combined to build the final iPLS model. The spectral intervals selected for predictions of TPC and DPPH are listed in Table
Results of interval partial least squares (iPLS) models for predictions of antioxidant activity of brown rice.
TPC (mg GAE/g) | DPPH (mg GAE/g) | |||||
---|---|---|---|---|---|---|
Selected intervals | RMSEPa | LVsb | Selected intervals | RMSEP | LVs | |
Raw data | 3, 6, 8 | 0.179 | 4 | 2, 5, 6 | 0.175 | 6 |
Smoothing | 3, 4, 8 | 0.131 | 5 | 2, 4, 6 | 0.188 | 4 |
D2 | 2, 4, 5 | 0.117 | 4 | 1, 3, 4 | 0.181 | 4 |
SNV | 3, 4, 6 | 0.062 | 4 | 2, 3, 6 | 0.141 | 5 |
bNumber of components of the final iPLS models including all the selected spectral intervals.
Root mean squared errors of cross-validation (RMSECV) obtained for each spectral interval by interval partial least squares (iPLS) with standard normal variate (SNV). The number at each bar indicates the number of PLS components at each wavelength interval.
A rapid method for determination of antioxidant activity was developed by near infrared (NIR) spectroscopy and chemometrics. By comparison of the results by FS-PLS and iPLS, wavelength selection can significantly improve the calibration accuracy of TPC and DPPH. The most suitable data preprocessing method was SNV. With SNV transformation and the selected wavelength ranges, the RMSEP is 0.062 mg GAE g−1 for TPC and 0.141 mg GAE g−1 for DPPH radical, respectively. The proposed method will provide a useful alternative tool to the physical and chemical analysis methods for brown rice.
The authors declare that there is no conflict of interests regarding the publication of this paper.
The authors are grateful for the financial support from the Public Welfare Social Development Project of Zhejiang Province (no. 2013C33032), the National Public Welfare Industry Project of China (no. 201210092, 2012104019), and Zhejiang Province Department of Education Fund Item (no. Y201122027).