Predicting retention indices of PAHs in reversed-phase liquid chromatography: A quantitative structure retention relationship approach

Abstract

In this work, the liquid chromatography retention time in monomeric and polymeric stationary phases of PAHs was investigated. Quantitative structure retention relationship approach has been successfully performed. At first, 3224 molecular descriptors were calculated for the optimized PAHs structure using Dragon software. Afterwards, the modelled dataset was divided using the CADEX algorithm into two subsets for internal and external validation. The genetic algorithm-based on a multiple linear regression was used for feature selection of the most significant descriptors and the model development. The selected models with five descriptors: nCIR, GGI3, GGI4, JGT and DP14 were used for the monomeric column and nR10, EEig01x, L1m, H5v and HATS6v were introduced for the polymeric column. Robustness and predictive performance of the suggested models were verified by both internal and external statistical validation. The good quality of the statistical parameters indicates the stability and predictive power of the suggested models. This study demonstrated the suitability of the established models in the prediction of liquid chromatographic retention indices of PAHs.

Keywords

• molecular descriptors
• genetic algorithm
• multiple linear regression
• prediction