Multiple linear regression calibrations for barley and malt protein based on the spectra of hordein

Abstract

The feasibility of using NIR spectral information from barley and malt hordein was assessed as to the suitability of developing improved NIR calibrations to predict protein in barley and malt. Using extracted hordein it was possible to gain more information on wavelengths relevant to predict protein with reduced errors. Strong correlations for grain protein and NIR wavelengths were found at 1,116, 1,268, 2,040, 2,068, 2,188 and 2,300 nm. Multiple linear regression equations provided improved predicting power for barley and malt protein with a standard error of prediction of 0.15 and 0.17%, respectively, whereas partial least squares regression gave a standard error of prediction of 0.22 and 0.27% for barley and malt, respectively. The use of NIR becomes more pronounced in breeding programs as NIR is a rapid and non-destructive technique allowing the screening of early generation lines with limited grain quantities. Also, the spectral analysis of native components from resting grain components will assist in building calibrations that provide qualitative values rather than just ranking breeding lines.