Comparison of Spectral and Molecular Analyses for Classification of Long Term Stored Wheat Samples.

Abstract

This study aimed to determine whether NIR spectroscopy and protein band analysis can differentiate the grain samples of 15 wheat genotypes stored for different periods: Group Ⅰ (91 weeks), Group Ⅱ (143 weeks), Group Ⅲ (194 weeks), and Group Ⅳ (246 weeks). Samples were harvested from previously-conducted field trials, and stored at +4 ℃. A-PAGE and SDS-PAGE methods were utilized to separate gliadin and glutenin fractions, respectively. A qualitative calibration model based on the Support Vector Machine (SVM) method was generated and validated using NIR spectra taken from samples. Results indicated storage length did not have an effect on molecular band fractions. Use of this method would not be considered an effective tool for discrimination of samples stored for different lengths of time. Spectral techniques may have potential in sorting samples based on their storage time. The SVM calibration model generated here had an acceptable true classification rate (over 80%) for separating all groups, while only Groups Ⅱ and Ⅳ were precisely separated (100% true classification rate) in the validation step.