Experimental designs for modeling retention patterns and separation efficiency in analysis of fatty acid methyl esters by gas chromatography-mass spectrometry.

Journal of chromatography. A

PubMedID: 21851946

Skartland LK, Mjøs SA, Grung B. Experimental designs for modeling retention patterns and separation efficiency in analysis of fatty acid methyl esters by gas chromatography-mass spectrometry. J Chromatogr A. 2011;1218(38):6823-31.
The retention behavior of components analyzed by chromatography varies with instrumental settings. Being able to predict how changes in these settings alter the elution pattern is useful, both with regards to component identification, as well as with regards to optimization of the chromatographic system. In this work, it is shown how experimental designs can be used for this purpose. Different experimental designs for response surface modeling of the separation of fatty acid methyl esters (FAME) as function of chromatographic conditions in GC have been evaluated. Full factorial, central composite, Doehlert and Box-Behnken designs were applied. A mixture of 38 FAMEs was separated on a polar cyanopropyl substituted polysilphenylene-siloxane phase capillary column. The temperature gradient, the start temperature of the gradient, and the carrier gas velocity were varied in the experiments. The modeled responses, as functions of chromatographic conditions, were retention time, retention indices, peak widths, separation efficiency and resolution between selected peak pairs. The designs that allowed inclusion of quadratic terms among the predictors performed significantly better than factorial design. Box-Behnken design provided the best results for prediction of retention, but the differences between the central composite, Doehlert and Box-Behnken designs were small. Retention indices could be modeled with much better accuracy than retention times. However, because the errors of predicted tR of closely eluting peaks were highly correlated, models of resolution (Rs) that were based on retention time had errors in the same range as corresponding models based on ECL.