reflectance spectrometry as a screening tool for prediction of lutein content in diverse wheat species (triticum spp.)

Direct analytical methods assessing lutein content (e.g., high-performance liquid chromatography [hplc]) are expensive and time-consuming. their utilization in broad mapping of genetic resources or selecting among hundreds of breeding lines is limited. therefore, the development of reliable predictive models based on reflectance spectrometry for assessing lutein content would be very useful. this study aims at finding a rapid and accurate lutein quantification method for wheat breeding purposes. reference samples were represented by hexaploid, tetraploid, and diploid wheat species (119 accessions) cultivated in 2 years. the species triticum aestivum l. predominated, with 81 accessions. two near-infrared (nir) spectroscopy devices were used to collect visible (vis) and/or nir spectra: the model foss 6500 with dispersive element and wavelength range 400–2500 nm and the nicolet antaris ii fourier transform spectrophotometer with wavelength range 1000–2500 nm. hplc was used as the reference method for assessing lutein content in wheat grain. sample grinding and proper nir and vis spectral combinations were essential for obtaining sufficient predictive parameters in calibration models. the foss 6500 calibration models were the most promising (standard error of prediction 0.46–0.78; ratio of performance to deviation 2.36–2.83 mg/kg). nir spectral instability of ground samples during 12 days in storage was detected and evoked a significant relative increase in lutein content of about 35% using the foss 6500 calibration model. after excluding five time-instable spectral regions, prediction differences of lutein content were markedly eliminated and did not exceed 7% in relative terms during 12 days of storage.