Canonical Discriminant Analysis Applied to the Headspace GC Profiles of Coffee Cultivars.

Chemometric methods were applied for analyzing the relationship between the classification of coffee cultivars and their volatile components. Six typical cultivars were selected from Coffea arabica L., and their headspace profiles were analyzed by GC and GC/MS. A canonical discriminant analysis, with GC peaks as the variables, suggested the existence of a relationship between the sensory characteristics and canonical variables. The six coffee cultivars were divided into three groups, respectively having a roasted note, sweet aroma and fermented odor. It is suggested that Brazil and Mandheling varieties had a roasted note derived from methylpyrazine, while Mocha coffee had a fermented odor derived from 2,3-butanedione.