Predictive quantitative structure retention relationship models for ion-exchange chromatography

Abstract

A database of probe molecules and their reported ion-exchange chromatographic data was collected from the literature, after which an extensive set of both traditional and novel molecular property descriptors were computed for each probe molecule. A genetic algorithm/partial least squares (GA/PLS) approach was then used on the data to create a predictive Quantitative Structure-Retention Relationship (QSRR) model of retention where a subset of the original data was used for training and the remainder of the data as a test set. The utility of this model was demonstrated by using it to predict the chromatographic behavior of compounds not included in the training set. The results presented in this paper demonstrate the utility of modern QSRR modeling to predict chromatographic behavior in ion-exchange systems.