Discrimination of intact almonds according to their bitterness and prediction of amygdalin concentration by Fourier transform infrared spectroscopy

Abstract

Intact almond kernels (N = 360, half sweet and half bitter) were analyzed using attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) for the prediction of amygdalin concentration and to classify them according to their bitterness. Amygdalin concentrations for sweet and bitter almonds, determined by high performance liquid chromatography, were between 0.7–350 and 15,000-50,000 mg kg−1, respectively. Concentrations were successfully predicted by applying partial least squares (PLS) to the pre-treated spectral data with R2p of 0.951 and RMSEP of 0.398. Additionally, linear discriminant analysis (LDA), quadratic discriminant analysis (QDA) and PLS-DA models were constructed to classify samples according to their bitterness. All three models provided a satisfactory discrimination of almonds into sweet and bitter categories, providing overall accuracy values of 83.3%, 86.1% and 98.6%, respectively. The results indicate the potential of ATR-FTIR spectroscopy for the reliable, easy and fast prediction of amygdalin concentration, and for almond classification according to their bitterness.