Aspects of Optimisation of Separation of Drugs by Chemometrics

Abstract: Statistical experimental designs have been used for method development and optimisation of separation. Two reversed phase HPLC methods were optimised. Parameters such as the pH, the amount of tetrabutylammonium (TBA; co-ion) and the gradient slope (acetonitrile) were investigated and optimised for separation of erythromycin A and eight related compounds. In the second method, a statistical experimental design was used, where the amounts of acetonitrile and octane sulphonate (OSA; counter ion) and the buffer concentration were studied, and generation of an ?-plot with chromatogram simulations optimised the separation of six analytes.The partial filling technique was used in capillary electrophoresis to introduce the chiral selector Cel7A. The effect of the pH, the ionic strength and the amount of acetonitrile on the separation and the peak shape of R- and S-propranolol were investigated.Microemulsion electrokinetic chromatography (MEEKC) is a technique similar to micellar electrokinetic chromatography (MEKC), except that the microemulsion has a core of tiny droplets of oil inside the micelles. A large number of factors can be varied when using this technique. A screening design using the amounts of sodium dodecyl sulphate (SDS), Brij 35, 1-butanol and 2-propanol, the buffer concentration and the temperature as factors revealed that the amounts of SDS and 2-propanol were the most important factors for migration time and selectivity manipulation of eight different compounds varying in charge and hydrophobicity. SDS and 2-propanol in the MEEKC method were further investigated in a three-level full factorial design analysing 29 different compounds sorted into five different groups. Different optimisation strategies were evaluated such as generating response surface plots of the selectivity/resolution of the most critical pair of peaks, employing chromatographic functions, simplex optimisation in MODDE and 3D resolution maps in DryLab™.Molecular descriptors were fitted in a PLS model to retention data from the three-level full factorial design of the MEEKC system. Two different test sets were used to study the predictive ability of the training set. It was concluded that 86 – 89% of the retention data could be predicted correctly for new molecules (80 – 120% of the experimental values) with different settings of SDS and 2-propanol.Statistical experimental designs and chemometrics are valuable tools for the development and optimisation of analytical methods. The same chemometric strategies can be employed for all types of separation techniques.