A Comparative QSRR Study on Enantioseparation of Ethanol Ester Enantiomers in HPLC Using Multivariate Image Analysis, Quantum Mechanical and Structural Descriptors

Abstract

Quantitative structure–retention relationship study was carried out for predicting the retention times of twenty‐six 1‐(2‐naphtyl)‐1 ethanol ester enantiomers in high‐pressure liquid chromatography by using original molecular, quantum mechanical, and multivariate image analysis (MIA) descriptors. Multiple linear regressions, partial least squares (PLS), and partial component regression (PCR) models combined with genetic algorithm (GA) were constructed as variable selection methods by using molecular descriptors to investigate the potential relationship between the selected descriptors and the retention times of the enantiomers. The molecular descriptors were generated from the molecular structure of enantiomers and calculated by the DRAGON software. Besides 504 DRAGON descriptors, additional 38 quantum mechanical descriptors were obtained using density functional theory/B3LYP/6‐31G method. Then, the results of the obtained models were compared with MIA descriptors that are pixels of two‐dimensional (2D) chemical structures. MIA descriptors were analyzed by correlation ranking‐PCR and PLS methods. The predictive ability of the models was evaluated using cross‐validation and an external test set. The results showed that GA is a good method for variable selection using structural descriptors combined with the PLS model. Since MIA descriptors are 2D pixels of the chemical structures, they can differentiate between the (R) and (S) isomers of enantiomers better than the quantum mechanical and structural descriptors. Comparison between the different models indicated that the GA‐PLS and MIA‐PLS models were the best methods.