QSRR modelling for the investigation of gas chromatography retention indices of flavour and fragrance compounds on Carbowax 20 ​M glass capillary column with the index of ideality of correlation and the consensus modelling

Abstract

The goal of the present study was to use a quantitative structure-retention relationship (QSRR) for the retention indices of 1179 flavour and fragrance organic compounds using the Monte Carlo algorithm of CORAL software. All the organic compounds were represented by SMILES notation for computation of descriptor of correlation weight (DCW). The dataset of 1179 flavour and fragrance organic compounds was used to make nine splits, each of which was further segmented into four sets: training, invisible training, calibration, and validation. The task of the index of ideality correlation (IIC) was analysed in-depth and it was found that the QSRR models generated by the use of IIC were more robust and significant. Two target functions i.e. TFA (IICweight=0.0), TFB (IICweight=0.2) were applied to build 18 QSRR models. The established QSRR model with TFB having Rvalidation2=0.9015 for split 6 was considered as the prime model. The reliability and robustness of the prime model was also confirmed by the numerical value of Qvalidation2=0.9000 and Qcalibration2=0.8919. The common promoters of increase and decrease of endpoint were also extracted from three splits 5, 6 and 9. Moreover, consensus modelling employing the split 6 layout of dataset distribution improves the prediction performance by enhancing the numerical value of Rvalidation2 from 0.9015 to 0.9241.