Simple QSPR Modeling for Prediction of the GC Retention Indices of Essential Oil Compounds

Abstract

A Quantitative structure-property relationship (QSPR) was developed to predict the retention index (RI) of 151 essential oil compounds. The predictive quality of the QSPR model was tested for an external set of 30 compounds, randomly chosen out of 151 compounds. A suitable set of molecular descriptors was calculated and the best-fitting descriptors were selected by using stepwise multiple linear regressions (SW-MLR). A simple SW-MLR model with three selected descriptors was obtained. The stability and the predictive performance of the proposed model was verified using both internal (cross-validation by leave one out, leave group out, bootstrap, Y-scrambling) and external validations. The model applicability domain (AD) was checked by the leverage approach to verify prediction reliability. This model, with high statistical significance (R2train = 0.933, R2test =0.945, R2adj =0.932, Q2LOO =0.928, Q2LGO =0.910, Q2Boost =0.928), could be used to predict the retention indices of the studied molecules.