Quantitative Structure–Retention Relationship Study of a Variety of Compounds in Reversed‐Phase Liquid Chromatography: A PLS‐MLR‐STANN Approach

Abstract

AbstractA quantitative structure–retention relationships model has been developed to study the retention behavior of 87 aliphatic and aromatic compounds in Reversed‐Phase Liquid Chromatography (RPLC) on five bonded‐phase columns differing in silanol group acidity. Six numerical descriptors of Molecular Mass (M), partial charge of the most negative atom (NPCH), partial charge of the most positive hydrogen (PCHH), van der Waals volume (VOLUME), Dipole Moment (DIMO), and Highest Occupied Molecular Orbital (HOMO) have been calculated for each compound. A separate Multiple Linear Regression (MLR) model has been developed using the six descriptors for each column. Partial Least Square (PLS) combined with MLR mean effects has been used for the mechanism interpretation. The descriptors of M, HOMO, and DIMO showed the largest PLS regression coefficients. It is found in the present work that solute size, n–π donor–acceptor and dipole–dipole interactions play a major role in the RPLC retention mechanism. Also, a self‐training artificial neural network was developed using the six descriptors as its inputs. The results obtained using this model are in good agreement with the experimental results of all five columns. Superiority of this model with low standard errors and high correlation coefficients for all five stationary phases reveals nonlinear characteristics of retention in RPLC. Successful application of the six descriptors in modeling of the retention behavior of different data sets using gas chromatography (previous work) and RPLC (present work) with different stationary phases reveals the robustness of the parameters.