Abstract :
Multivariate regression modeling techniques (PLS-1 regression modeling) were applied to ordinary UV spectral absorption data obtained on solutions containing inclusion complexes formed between homochiral modified cyclodextrins (methyl(beta)-cyclodextrin, (alpha)-, (beta)-, and (gamma)-carboxymethyl cyclodextrins and (alpha)-, (beta)-, and (gamma)hydroxypropyl cyclodextrins) and four guest molecules of pharmaceutical interest (ephedrine, norephedrine, norepinephrineL-bitartrate, and tryptophan methyl ester). The PLS-1 regression models were developed by correlating the known enantiomeric composition of laboratory prepared samples with ordinary UV absorption spectral data. The regression models were subsequently validated with laboratory-prepared test sets. The rms percent relative error in the predicted mol fraction of (1S, 2R)-(+)-ephedrine, (1S, 2R)-(+)-norephedrine, (R)-(-)-norepinephrine-L-bitartrate, and D-tryptophan methyl ester obtained with the independently prepared test sets was heavily dependent on the host molecule used.
