QSRR-ANN modelling in β-CD-modified RP-HPLC

Abstract

Introduction: Cyclodextrin (CD) added to RPHPLC mobile phase interacts with analytes, solvent components and stationary phase surface. Therefore, the retention is influenced by the analyte’s distribution between CD dissolved in the mobile phase, free CD and formed inclusion complex adsorbed onto the stationary phase, and stationary phase itself (1, 2). The research goal was to reveal the structural characteristics affecting the inclusion complexation and retention in these kinds of chromatographic systems by employing QSRR-ANN models. Materials and Methods: Mixed QSRR model included large pool of molecular descriptors, complex association constants and experimental parameters towards the retention factor of risperidone, olanzapine and structurally related impurities. The experimental space was adequately covered with central composite design, while experiments were conducted on Dionex Ultimate 3000 (U) HPLC. QSRR-ANN modelling was performed in STATISTICA Neural Networks. Results: To evaluate the individual influence of each of the descriptors, the difference in the highest and lowest retention factor value across the investigated range of the descriptor’s values was calculated. The highest ratios were associated with the following descriptors RDF075m, UE, Mor04v and CATS2D_08_PL, making them the most contributing towards the selected output. RDF075m descriptor shows the three-dimensional mass distribution calculated at a distance of 7.5 Å from the geometrical centre of the molecule and it refers to steric factors at the same distance. Groups approximately 7.5 Å distant from the geometrical centre of risperidone, olanzapine and related compounds in their optimized conformations were determined. These groups were the same ones involved in the complexation process according to previously performed NMR study. Identified groups and their steric factors are the most important for the formation of inclusion complexes, and, in this way, the value of RDF075m contributes to the retention of the selected compounds. The importance of Mor04v confirms the influence of molecular size and shape in retention in these kinds of chromatographic systems, while CATS2D_08_PL accounts for lipophilicity. Conclusions: The current study resulted in development of QSRR-ANN with remarkable performances, which enabled the elucidation of the molecular features significantly influencing the retention in β-CD-modified RP-HPLC. The pronounced effect of molecular structure on retention was best described through RDF075m, followed by UE, Mor04v and CATS2D_08_PL. Retention behaviour is also highly affected by molecular size and shape, as well as lipophilicity of the investigated compounds. Moreover, the size and polarity of the chosen CD should not be neglected, due to the consequent structural fit.