Prediction of seed coat proportion in narrow-leafed and yellow lupins by near-infrared reflectance spectroscopy (NIRS)

Abstract

Breeding programs oriented to decrease hull percentage in lupins and consequently to improve their nutritional value could be greatly assisted by a reliable technique to predict seed coat proportion (SCP) in whole, intact seeds. The objective of this work was to evaluate the potential of near-infrared reflectance spectroscopy (NIRS) to predict the percentage of seed coat in two lupin species. Samples (n = 627) of seeds of different lines and crosses of narrow-leafed lupin (Lupinus angustifolius, n = 447) and yellow lupin (L. luteus, n = 180) were scanned in the VIS-NIR range and SCP subsequently determined by dissection and weighing of dry seed fractions (reference data). Calibrations were developed by multivariate regression (modified partial least squares (PLS)), testing different mathematical treatments. The best equation was obtained when both species were pooled in one set of samples and the best treatment of the spectral data consisted in a second subtraction order, over an interval of 8 data points (16 nm) and a smoothing segment of 8 data points. Cross validation for the selected equation showed a coefficient of determination of 88% and a standard error of 0.54%, for an average SCP of 22.63% and a standard deviation of the reference data of 1.55%. It was concluded that NIRS can be effectively applied in breeding programs as a fast, non-destructive screening tool to determine SCP in narrow-leafed and yellow lupins.