Regresar

A PRELIMINARY QUANTITATIVE STRUCTURE-RELATIVE SWEETNESS RELATIONSHIP

Abstract:

A predictive quantitative structure-relative sweetness relationship (QSRSR) was developed using a data set of 202 sugars and sweeteners. The Dragon software was used to calculate 1932 non-conformational molecular descriptors, which were then analyzed through a multivariable linear regression analysis together with the wellknown replacement method (RM) variable subset selection technique. A seven-parameter model is achieved. For the model validation, the data set was split into a traning set (142 sweeteners) and a test set (60 sweeteners). The results reflect that the spectral positive sum from the weighted-by-mass Burden matrix [SpPos_B (m)], which belong to the 2D matrix-based descriptors, has a high relevance as a synergic effect for this purpose, ie, it has the higher value of the standardized coefficients. Thus, this model can be used in order to predict the sweetness of both unevaluated and un-synthesized sweeteners, as well as complements other studies reported in the literature.