Retention prediction of adrenoreceptor agonists and antagonists on a diol column in hydrophilic interaction chromatography

Abstract

Retention prediction models using multiple linear regression (MLR) and artificial neural networks (ANN) were developed for adrenoreceptor agonists and antagonists chromatographed on a diol column under hydrophilic interaction chromatographic (HILIC) mode at three pH conditions (3.0, 4.0 and 5.0). Using stepwise MLR, the retention behavior of the analytes was satisfactorily described by a five-predictor model; the predictors being the percentage of acetonitrile in the mobile phase (% ACN), the logarithm of partition coefficient (logD), the number of hydrogen bond donor (HBD), the number of hydrogen bond acceptor (HBA), and the total absolute atomic charge of the molecule (TAAC). Among the five descriptors, % ACN had the strongest effect on the retention as indicated by its relatively higher standardized coefficient compared to the other four predictors. The inclusion of the four predictors which are related to the properties of the compounds (logD, HBD, HBA and TAAC), suggested hydrophilic interaction, hydrogen bonding and ionic interaction as possible mechanisms of retention of the analytes on the studied system. The models derived from MLR also showed adequate fit as proven by the high correlation (R2 as high as 0.9667) between observed and predicted logk values for the training set and good predictive power on the test set (R2 greater than 0.97). ANN analyses of the data were also conducted using the five predictors derived from MLR as inputs and logk as output. The trained ANNs showed better predictive abilities as compared to MLR models as indicated by relative higher R2 and lower root mean square error of predictions (RMSEP) for both training and test sets. The derived models can be used as references for method development and optimization of chromatographic conditions for the separation of adrenoreceptor agonists and antagonists.