Structure-olfactive threshold relationships for pyrazine derivatives.

Structure-olfactive threshold relationships for 40 pyrazine derivatives have been studied by multivariate statistical analysis. Variable descriptors used to describe the molecules studied were obtained using molecular-mechanics software. A correlation coefficient of 0.82 was obtained when all the molecules were included in the model. When the model was established for closely related subsets of molecules, the correlation coefficients obtained were higher and the established models were bilinear. Such models allow the identification of optimal structures corresponding to low olfactive thresholds for the subsets studied. Surprisingly, we find that the optimum structures are included in the set of 40 molecules. The efficiency of the models was supported by the cross-validation technique, where the correlation coefficients were found to be good with respect to the precision of the values of the olfactive thresholds.