A consideration for structure-taste correlations of perillartines using pattern-recognition techniques.

The relationships between molecular structure and taste quality (sweet or bitter) or several perillartine derivatives were investigated using pattern-recognition techniques. For the classification of these compounds into two classes (sweet or bitter), a significant discriminant function was developed by the use of linear learning machine. All the compounds were assigned correctly to their observed taste classes by the function involving three parameters (one hydrophobic and two steric). In addition, the K-L transformation technique was used for examination of classification results.