Prediction of protein and amino acid contents in whole and ground lentils using near-infrared reflectance spectroscopy

Abstract

Lentil is an important source of plant-based protein, and the protein and amino acid contents have a significant influence on its nutritional quality and value. In this study, near-infrared reflectance spectroscopy (NIRS) models were developed by partial least squares (PLS) regression to predict the crude protein and 18 amino acid contents of lentil seeds. The effects of sample status (whole and ground), type of spectrometer (PerkinElmer DA 7250 and FT 9700), and amino acid/protein correlation on model performance were analyzed and evaluated. The DA 7250 models and FT 9700 models of protein and 14 amino acids, except histidine, tyrosine, methionine and cysteine, showed good statistical results with coefficients of determination for calibration (R2C) higher than 0.652 and residual predictive deviation (RPD) values higher than 1.57. The DA 7250 models achieved similar accuracy for the determination of compositions in whole and ground samples. Two spectrometers had no significant difference (p > 0.05) for measurement in ground lentils. NIRS models could perform better for certain amino acids when they were highly correlated to protein. Overall, NIRS had a significant potential for rapid, accurate and simultaneous prediction of protein and most amino acid contents in lentils.