USE OF MULTIVARIATE CHARACTERIZATION, DESIGN AND ANALYSIS IN ASSAY OPTIMIZATION

A procedure is proposed for utilizing information from previous work o n development of LC assays in order to facilitate the analysis of nove l compounds related to those previously analysed. The procedure employ s a multivariate method from the field of chemometrics, partial least squares analysis (PLS) to combine quantitative information on the chem ical properties of a compound with a quantitative description of the c olumn and the mobile phase and then to use this information to form a regression model for the retention time. A test of the procedure was m ade by using data on nucleoside analogues studied in our laboratory. D ata obtained from chromatographic studies of seven compounds tested in a total of 28 combinations of columns and mobile phases (3-5 per comp ound) were used to calculate a PLS model. The model was then used to p redict retention times of nine other substances and the results were c ompared with experimental data. The predictions were (115 +/- 82%) (95 % confidence interval) of the experimentally observed retention times. The results are encouraging and the method will be subject to further and extended investigations.