Development of Quantitative Structure-Property Relationship (QSPR) Models of Aspartyl-Derivatives Based on Eigenvalues (EVA) of Calculated Vibrational Spectra

Abstract

In present study, validated linear quantitative structure-property (sweetness) relationship (QSPR) models were developed, based on an experimental sweetness index and the calculated vibrational spectra of 158 aspartyl-derivatives. The quantum chemical AM1 method was applied to calculate the vibrational spectra of the compounds. The EVA descriptors were derived from the vibrational spectra for a combination of various values of σ (2.5, 5, 10,20 and 40 cm−1) and L (5, 10, 20 and 40 cm−1), which were correlated with an experimental sweetness index of the compounds using the multiple linear regression method. Several QSPR equations were constructed and analyzed. Among the various combinations, the statistically best QSPR equations were achieved when σ =10 and L = 10 were chosen. Four equations were selected as QSPR models, namely, model_1, model_2, model_3 and model_4. Satisfactory statistical values were found for the fit of the training set, R2=0.618, 0.691, 0.605 and 0.604, internal validation Q2=0.540, 0.647, 0.550 and 0.548, and the external predictive measures for the test set, $$ {R}_{pre.}^2 $$=0.614, 0.616, 0.633 and 0.633 were obtained for dubbed 1 to 4 respectively. The application domain of the models was determined. The results of the Y-randomization test of the models support the finding that the statistical values obtained for the QSPR models were not due to just chance. Graphic visualizations of the EVA profiles of some of the selected compounds showed that the presence of hydrophobic moiety in the structure such as those of saturated hydrocarbons, e.g. heptane in ASP_215, tert-butyl in ASP_63, cyclohexane in ASP_201 and pentane in ASP_199, which is related to a high sweetness index, might be monitored in the EVA profiles of aspartyl-derivatives.