Chemometric investigation of mustard seed

Abstract

Levels of the 11 major glucosinolates and two other characteristics (thousand seed weight and colour) were determined in samples of the mustard seed (of Sinapis alba, Brassica nigra and Brassica juncea) and in samples of the seed of similar Brassica plants ( Sinapis arvensis and Brassica napus) available from the local testing stations, the local producers and suppliers. The aim of the work was to rationalize the mustard seed quality control using multivariate statistical methods discriminating the individual oilseed types. The data obtained from 19 oilseed samples were subjected to principal components analysis (PCA). PCA conducted with only six original variables (sinigrin, gluconapin, progoitrin and glucoibervirin level, thousand seed weight and colour) was able to separate clusters of Sinapis alba and Brassica napus seeds. A new PCA conducted with a reduced set of four variables (gluconapin, progoitrin and glucoibervirin level and thousand seed weight) readily distinguished between the seeds of Brassica nigra, Brassica juncea and Sinapis arvensis. An acceptable grouping of samples (percent of grouped cases correctly classified was 100) was achieved by stepwise linear discriminant analysis. Only seven variables (thousand seed weight, levels of progoitrin, gluconapin, glucoibervirin, glucoerucin, gluconasturtiin and 4-hydroxyglucobrassicin) contributed to adequate differentiation of B. juncea, B. nigra. and S. arvensis seeds