Selecting optimal scouting runs for retention modelling in reversed phase liquid chromatography

Many applications in modern liquid chromatography, such as the optimization of resolution in method development, the validation of the robustness of a method, and the unambiguous identification of an analyte, rely on accurate retention time predictions. For this purpose, empirical models relating the retention factor of an analyte to an independent experimental variable, such as the fraction of strong eluting solvent in the mobile phase, are often used. Frequently used models in liquid chromatography are the linear solvent strength (LSS) model [1] and the Neue-Kuss (NK) model [2]. To obtain an adequate retention model for a compound of interest, a number of experimental input or scouting runs, that can be executed in isocratic mode and/or gradient elution mode, are required. The minimum number of scouting runs is typically equal to the number of fitting parameters in the model (i.e., two for the LSS model, three for the NK model). The selection of scouting runs can have a large impact on the resulting prediction accuracy, whereby the selection of scouting runs leading to the highest prediction accuracy moreover seems to be analyte dependent. When complex mixtures containing multiple analytes are evaluated, the selection of scouting runs for an adequate retention time prediction of all compounds becomes even more challenging. In this study, the effect of different sets of scouting runs on the obtained retention time prediction is evaluated. More specifically, the retention time prediction accuracy in isocratic and gradient elution mode is examined when experimental isocratic and/or gradient retention data are modelled to the NK model. It is investigated how well the obtained retention models obtained from isocratic and gradient data transfer to retention predictions in isocratic and gradient elution modes, in all possible combinations. It is demonstrated that the selection of the initial scouting runs is very important to obtain a satisfactory prediction accuracy, and that this selection depends on the retention properties of the compound under consideration. This implies that there is no ‘universal’ set of initial scouting runs that can be applied to any compound under any set of separation conditions. [1] L.R. Snyder, J.W. Dolan, High-Performance Gradient Elution: The Practical Application of the Linear-Solvent-Strength Model, J. Wiley & Sons, Hoboken, 2007. [2] Neue UD, Kuss HJ. Improved reversed-phase gradient retention modeling. J Chromatogr A. 2010;1217(24):3794–803