Premium
Predictive QSPR modelling for the olfactory threshold of a series of pyrazine derivatives
Author(s) -
Pal Pallabi,
Mitra Indrani,
Roy Kunal
Publication year - 2013
Publication title -
flavour and fragrance journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.393
H-Index - 70
eISSN - 1099-1026
pISSN - 0882-5734
DOI - 10.1002/ffj.3135
Subject(s) - quantitative structure–activity relationship , chemistry , partial least squares regression , biological system , pyrazine , molecular descriptor , in silico , molecule , chemometrics , computational chemistry , stereochemistry , statistics , mathematics , organic chemistry , chromatography , biochemistry , gene , biology
Nowadays, much interest is shown in the prediction of different organoleptic properties of molecules for their prominent use in a variety of formulations and food products. Against this background, cheminformatics modelling was done in the present work for the development of predictive models for the olfactory threshold (log T ) of a series of 74 pyrazine derivatives employing in silico techniques such as genetic function algorithm (GFA) and genetic partial least squares algorithm (G/PLS). Different categories of descriptors were calculated for the work. After validating the models both internally and externally, a parabolic GFA‐spline model was selected as the best model which showed significant predictive quality [ R pred 2= 0.822, = 0.641, = 0.019]. Interpretation of the descriptors concluded that increased hydrophobic surface area, along with an increase in the positive charge weighted solvent accessible surface area (as interpreted from the Jurs_DPSA _2 descriptor) in the molecules, provides a low value of log T . Again, the presence of an isopropyl group and electron richness of the molecules significantly influences the property profile (log T ) of the molecules. Finally, 27 pyrazine derivatives were designed based on the present analysis and good in silico prediction values for odour threshold, i.e. low log T values were obtained from the developed model. Thus, use of this optimized quantitative structure–property relationship (QSPR) model may be used to screen molecule databases as well as modify the structures for selection of potent entities. Copyright © 2013 John Wiley & Sons, Ltd.