Use of multivariate analysis for optimization of separation parameters and prediction of migration time, resolution,...

The optimization of separation parameters in chromatography for better separation and resolution of analytes continues to be a labor intensive procedure usually performed by a trial and error method. A multivariate analysis in the form of multilinear regression (MLR) is used to optimize separation parameters and predict the migration behavior, resolution, and resolution per unit time of achiral (4-chlorophenol, pentachlorophenol, clonazepam, and diazepam) and chiral (1,1'-binaphthyl 2,2'-dihydrogen phosphate (BNP), and 1,1'-bi-2-naphthol (BOH)) compounds in MEKC. Separations of achiral and chiral analytes were performed using an achiral (poly(sodium N-undecylenic sulfate)) molecular micelle and chiral (poly(sodium N-undecanoyl-L-leucylvalinate) or poly(sodium N-undecanoyl-L-isoleucylvalinate)) molecular micelle, respectively, at various operating temperatures, applied voltages, pH values, and molecular micelle concentrations in the BGE. The separation parameters were subsequently used as input variables for MLR models. The models were validated with independent samples. The root-mean-square percent relative error (RMS%RE) is used as a figure of merit for characterizing the performance of the migration time, resolution, and resolution per unit time models. The RMS%RE obtained for predicted migrated times, resolutions, and resolution per unit time of 4-chlorophenol, pentachlorophenol, clonazepam, diazepam, BNP, and BOH ranged between 8 and 19%. The same experimental procedure was used to optimize the separation parameters of six other chiral analytes of different compound class. The predicted migration times, resolutions, and resolution per unit time of the chiral as well as the achiral analytes compare favorably with the experimental migration times and resolutions, indicating versatility and wide applicability of the technique in MEKC.