Rapid Determination of Fat, Protein and Amino Acid Content in Coix Seed Using Near-Infrared Spectroscopy Technique

The feasibility of near-infrared (NIR) spectroscopy to determine fat, protein, and amino acids (AAs) in Coix seed was investigated. Partial least squares (PLS) regression was applied to establish quantitative model. Competitive adaptive reweighted sampling (CARS) and Norris derivative smoothing (NDS) were used to improve the model accuracy. For fat, protein, Asx, threonine, serine, Glx, alanine, leucine, proline, lysine, and histidine, NDS pretreatment improved the models' performance and yielded better prediction results. Then, key variables were selected by CARS, and the PLS models could obtain the optimal results with good predictive ability and robustness. The models of protein, Asx, serine, Glx, alanine, valine, isoleucine, leucine, phenylalanine, and proline were fit for screening with the residual predictive deviation (RPD, the ratio between the standard deviation of the reference value and the root mean square error of prediction of the validation set) equal or greater than 2.50. The RPDs of fat and threonine were 1.61 and 2.00, and the other AAs' were less than 1.50. It was concluded that the NIR spectral technique was suitable for determining fat, protein, and most of AAs in Coix seed. The technique might also give a rough estimate of the contents of glycine, cysteine, methionine, tyrosine, arginine, lysine, and histidine.