Simultaneously determining amino acid contents using near-infrared reflectance spectroscopy improved by pre-processing method in rice

Abstract

Rice, the most consumed in Asian and constitute important source of amino acids. The present study evaluated the suitability of near-infrared reflectance spectroscopy to predict amino acid contents (AACs) using 1210 rice samples representing a wide range of grain qualities and genetic background. Calibration models established by combining different spectral derivatives, scatter and baseline correction methods, and regression algorithm were compared using full cross validation. A modified partial least square model (MPLS) with the spectral derivative treatment “2, 3, 3, 1” (2nd order derivative computed based on 3 data points, and 3 and 1 data points in the 1st and 2nd smoothing, respectively) and weighted multiplicative scattering correction (WMSC) method was identified as the best model for simultaneously measurement of 14 AACs in brown rice flour. The high model quality is reflected by RSQ of 0.914∼0.969, standard error of calibration (SEC) of 0.0187%∼0.0553%, 1-VR of 0.901∼0.955, SECV of 0.0214%∼0.0579%, and residual prediction deviation (RPD) of 2.58∼3.85. These results indicated that NIRS coupled with an appropriate pre-processing, multivariate regression could be a rapid and reliable method for quantification of AACs in rice breeding programs and quality control in the food industry. Factors that would influence model quality were investigated and discussed.