Optimisation of chromatographic separations by use of a chromatographic response function, empirical modelling and multivariate analysis

The chiral separation of the drug substance R,S-oxybutynin chloride on a reversed phase HPLC system has been optimised by use of empirical modelling and multivariate analysis. The separation was characterised by a new chromatographic response function developed to modulate both quality of separation and retention time. The study includes a comparison between three different multivariate techniques (multi-layer feed-forward neural networks, multiple linear regression and partial least squares regression) of their capabilities to model the new chromatographic response function and predict its value for new experiments. It was indicated that the most accurate models were achieved with neural networks, although partial least squares regression could also be used to solve the problem since it gives the major directions for the optimal settings of the variables.