Multiple quantitative structure-pungency correlations of capsaicinoids

The comprehensive mechanistic understanding of pungency and the binding interactions between pungent capsaicinoids from foods and their receptors have attracted increasing attention in food sensory and pharmaceutical fields. In this study, linear and quadratic statistically significant quantitative structure-pungency correlations have firstly been established for capsaicinoids by combining genetic function approximation and brute force approach and subsequently validated by the tests of cross validation, randomization, external prediction, Roy's r_m² metrics and Golbraikh-Tropsha's criteria. The resultant optimal predictive correlation models have strong internal and external predictive capacities (r² = 0.949-0.989, r²_CV = 0.860-0.955, r²_pred = 0.859-0.904), which elucidate the elementary electrostatic, hydrogen bonding, hydrophobic and steric structural requirements for the pungent perception of capsaicinoids. Finally, a series of new capsaicinoids was designed based on the insights from the established correlation models, and most of which showed excellent predicted pungency potency and acceptable ADMET properties.