Design and application of an expert system for mobile phase optimisation in reversed-phase liquid chromatography.
The selection of the optimum composition for the mobile phase in reversed-phase high-performance liquid chromatography (HPLC) is a complex task; conventional approaches require the expenditure of significant amounts of time by the analyst, particularly for complex mixtures of solutes of biological origin. Some of the existing strategies for the automated optimisation of mobile phase composition (e.g. Simplex), may fail if the elution order of the components changes; or they may require that standards be chromatographed in order to establish the retention behaviour of each component in a mixture (e.g. resolution mapping). These problems may be overcome if the retention behaviour of each individual solute can be established from the chromatogram of the mixture. In this regard, components can be tracked by exploiting the spectral information generated by a rapid scanning photodiode array detector. Unfortunately this information is often insufficiently detailed to allow an unambiguous model of retention behaviour to be constructed. The system developed by the Authors uses these spectral data as a basis for constructing one or more hypothetical retention models, each of which is refined or rejected as further information is obtained during the progress of the experiment. To improve the reliability of the retention models proposed by the system, the spectral data are utilised in a number of tests designed to assess the purity of each chromatographic peak. The information so generated may be used in conjunction with any previously acquired spectral data both to select an appropriate method for extracting spectra for each component from the matrix of (A, lambda, t) data and to establish reliability parameters for the resultant spectra. The development and philosophy of the expert system developed for eluent optimisation in reversed-phase HPLC is discussed.