A validation model for prediction of kovats retention indices of compounds isolated from origanum spp. and thymus spp. essential oils

Abstract

This work aimed to obtain a validated model for the prediction of retention times of compounds isolated from Origanum heracleoticum, Origanum vulgare, Thymus vulgaris, and Thymus serpyllum essential oils. In total 68 experimentally obtained retention times of compounds, which were separated and detected by GC-MS were further used to build the prediction models. The quantitative structure–retention relationship was employed to foresee the Kovats retention indices of compounds acquired by GC-MS analysis, using eight molecular descriptors selected by a genetic algorithm. The chosen descriptors were used as inputs for the four artificial neural networks, to construct a Kovats retention indices predictive quantitative structure–retention relationship model. The coefficients of determination in the training cycle were 0.830; 0.852; 0.922 and 0.815 (for compounds found in O. heracleoticum, O. vulgare, T. vulgaris and T. serpyllum essential oils, respectively), demonstrating that these models could be used for prediction of Kovats retention indices, due to low prediction error and high r2.