Abstract: In this work, in order to establish a predictive model using TQ analyst software to determine amylose content by attenuated total reflectance-Fourier transform mid-infrared (ATR FT-MIR) spectroscopy, the correlation between amylose content and the spectral variables selected by principal component analysis (PCA) and orthogonal partial least squares (OPLS) was explored and the difference in the interpretation of the selected variables for the models was compared. The results showed that the characteristic waveband selected by OPLS was 969–1 158 cm-1, mainly corresponding to the crystalline and amorphous regions of amylose, and was also the characteristic band of the C–O–C stretching vibration of α-1,4-glycosidic bonds. The prediction performance of the model developed in this region was improved compared with those based on the full-band spectra and in the region of 800–1 200 cm-1, with a correlation coefficient of 0.999 8, a root mean square error for calibration (RMSEC) of 0.587%, a root mean square error for prediction (RMSEP) of 6.26%, and a relative percent deviation (RPD) of 5.177 8. The correlation coefficient between the predicted value and the real value was 0.962 7. Therefore, the variables selected by OPLS could interpret most of the chemical characteristics in the mid-infrared region of amylose, and enhance the analytical capability of the prediction model.

Key words: amylose; attenuated total reflectance-Fourier transform mid-infrared spectroscopy; variable selection; orthogonal partial least squares