Plant Soil Environ., 2012, 58(4):196-203 | DOI: 10.17221/684/2011-PSE

Near infrared spectroscopy for deoxynivalenol content estimation in intact wheat grain

V. Dvořáček, A. Prohasková, J. Chrpová, L. Štočková

Crop Research Institute, Prague-Ruzyně, Czech Republic

Non-invasive determination of deoxynivalenol (DON) still presents a challenging problem. Therefore, the present study was aimed at a rapid determination of DON in whole wheat grain by means of FT-NIR spectroscopy, with a wide range of concentrations for potential applications in breeding programs and common systems of quality management using partial least square calibration (PLS) and discriminant analysis technique (DA). Using a set of artificially infected wheat samples with a known content of DON, four PLS models with different concentration range were created. The broadest model predicting DON in the concentration range of 0-90 mg/kg possessed the highest correlation coefficients of calibration and cross validation (0.94 and 0.88); but also possessed the highest prediction errors (SEP = 6.23 mg/kg). Thus the subsequent combination of DA as the wide range predictive model and the low-range PLS model was used. This technique gave more precise results in the samples with lower DON concentrations - below 30 mg/kg (SEP = 2.35 mg/kg), when compared to the most wide-range PLS model (SEP = 5.95 mg/kg).Such technique enables to estimate DON content in collections of artificially infected wheat plants in Fusarium resistance breeding experiments.

Keywords: FT-NIR; Triticum aestivum; Fusarium; DON; discriminant analysis; PLS regression

Published: April 30, 2012  Show citation