Species differentiation by multivariate analysis of headspace volatile patterns from canned pacific salmon

Multivariate statistical analyses were applied to forty four volatile compounds measured by static headspace gas chromatography with the aim of differentiating between four species of canned Pacific salmon, namely chum (Oncorhynchus keta), coho (O.kisutch), pink (O.gorbuscha), and sockeye (O.nerka). Using principal component analysis as a data reduction technique, the first ten principal components (PC) were retained and explained 87% of the total variation in the data. PC1, PC3, and PC4 regrouped several volatiles important toward separation between salmon species, while PC2 comprised variation accounted for by the loss of chromatographic separation from year to year. In order to generate functions which segregate species into separate groupings, three types of discriminant analyses were carried out based on the scores of the ten principal components. Nonparametric discriminant analysis produced a higher percentage of correct classification in the four respective species groupings (98.4%) than linear (88...