Mathematics Faculty Articles

Characterization of a comprehensive flavor database

Document Type

Article

Publication Date

8-26-2011

Publication Title

Journal of Chemometrics

Keywords

flavor fingerprint, principal component analysis, clustering analysis

ISSN

0886-9383

Volume

25

First Page

550

Last Page

560

Abstract

Flavor perception involves, among a number of physiological and psychological processes, the recognition of chemicals by olfactory and taste receptors. The highly complex and multidimensional nature of flavor perception challenges our ability to both predict and design new flavor entities. Toward this endeavor, classifications of flavor descriptors have been proposed. Here, we developed a fingerprint‐based representation of a large data set comprising 4181 molecules taken from the commercially available Leffingwell & Associates Canton, Georgia, USA database marketed as Flavor‐Base Pro©2010. Flavor descriptions of the materials in this database were composite descriptions, collected from numerous sources over the course of more than 40 years. The flavor descriptors were referenced against a detailed and authoritative sensory lexicon (ASTM, American Society for Testing and Materials publication DS 66) comprising 662 flavor attributes. Comparison of clustering analysis, principal component analysis, and descriptor associations provided similar conclusions for various mutually correlated descriptors. Regarding analysis of the flavor similarity of the molecules, the clustering performed provided a means for the quick selection of molecules with either high or low flavor similarity description. Preliminary comparison of the chemical structures to the flavor description demonstrated the feasibility but also the complexity of this task. Additional studies including different structural representations, careful selection of subsets from this data set, as well as the use of a number of classification methods will demonstrate the utility of structure–flavor associations. This work shows that the flavor information contained in databases, such as that used in the present study, can be analyzed following standard chemoinformatics methods.

Comments

Copyright © 2011 John Wiley & Sons, Ltd.

DOI

10.1002/cem.1399

This document is currently not available here.

Peer Reviewed

Find in your library

Share

COinS