A retention index-based QSPR model for the quality control of rice

Abstract

The purpose of work presented here was to calibrate and validate a mathematical model based on a quantitative structure-property relationship for modeling the retention indices (I) of 137 volatile organic compounds (VOCs) measured in the headspace of rice using a Divinylbenzene-Carboxen-Polydimethylsiloxane (DVB-CAR-PDMS) fiber in the solid-phase microextraction-gas chromatography-mass spectrometry (SPME-GC-MS) analysis. The dataset was split into training, validation and test sets according to the Balanced Subsets Method (BSM). The study was divided into three different steps. In the first step, 1753 conformation-independent descriptors were considered for modeling. In the second step, 1145 conformation-dependent descriptors were taken into account to obtain a model. Finally, in the last step both conformation-independent and conformation-dependent descriptors were used to build the model. A three-descriptor model was retained as the optimal one in all cases. Conformation-dependent descriptors led to models with no appreciable improvement over those obtained with conformation-independent descriptors. The final conformation-independent QSPR model was used as a tool for the quality control of volatile contaminants of rice by predicting the retention indices in a set of 46 rice contaminants.