Simultaneous Prediction of the Logarithmic Capacity Factor of Some Aliphatic and Aromatic Compounds on Five Different Stationary Phases in RP-LC Using Artificial Neural Network

Abstract

The present work is a quantitative structure-retention relationship (QSRR) study based on multiple linear regressions (MLR) and artificial neural network (ANN) to predict the retention behavior of solutes in reverse-phase liquid chromatography based on their structures. To attain this goal, the capacity factors (log k′) of a collection of 65 aliphatic and aromatic solutes on five stationary phases (Zorbax SB-C18, Zorbax Rx-C18, Hypersil C18, Hypersil C8 and Zorbax C8) were selected as a data set. By using Dragon software, various descriptors were calculated for all molecules in the data set. The descriptors that appeared in this model were selected by stepwise multiple linear regression (MLR). These descriptors were used as inputs of the constructed ANN and its output was logarithm capacity factors of aliphatic and aromatic compounds on five different columns. Comparison between statistical results calculated for MLR and ANN model reveals that all statistics have improved considerably in case of the ANN model. The improved statistics by the ANN would suggest the existence of nonlinear relation between selected molecular descriptors and their retention in reverse phase liquid chromatography.