The essential oils from leaves of 20 commercial citrus accessions maintained by the University of California, Riverside Givaudan Citrus Variety Collection and selected on the basis of their odor profile were analyzed by GCMS/FID. The main components were quantified while the semi-quantitative percentage composition data was compiled with data from other publications for sample visualization, classification and comparison with leaf oils from other citrus accessions. Some compositional clusters aligned closely with the taxonomic clades of sweet orange, bitter orange, and C. hystrix while other clades like the mandarins and lemons showed distinct chemical sub-groups. Characteristic compounds for the clusters included linalyl acetate and linalool (bitter orange leaf), sabinene (sweet orange leaf), methyl N-methyl anthranilate (mandarin leaf), γ-terpinene (yuzu leaf), citronellal (C. hystrix), limonene, citronellal and citral (lemons and citrons). A chemometric approach combined with t-SNE cluster plots can be more informative than taxonomic assignments when considering flavor and fragrance characteristics.
Keywords: Citrus leaf; essential oil; machine learning; phytochemistry; terpenoids.
© 2022 Wiley-VHCA AG, Zurich, Switzerland.