Relationship between the matrix effect and the physicochemical properties of analytes in gas chromatography.

The phenomenon "matrix-induced chromatographic response enhancement" (matrix effect) causes quantitative errors in gas chromatography (GC) analyses. This effect varies according to the analyte nature, matrix type and concentration, and GC-system parameters. By focusing on the physicochemical properties of analytes, a predictive model was developed for the matrix effect using quantitative structure-property relationships. Experimental values of the matrix effect were determined for 58 compounds in a serum extract obtained from solid-phase extraction as the matrix. Eight molecular descriptors were selected, and the matrix-effect model was developed by multiple linear regression. The developed model predicted values for the matrix effect without any further experimental measurements. It also indicated that the molecular polarity (particularly H-bond donors) and volume of the analyte increase the matrix effect, while hydrophobicity and increasing number of nonpolar carbon atoms in the analyte decrease the matrix effect. The model was applied to the analysis of barbiturates. The predicted values indicated that N-methylation decreases the matrix effect, and the relative predicted values were effective for the selection of an internal standard. The obtained insight into the matrix effect and the prediction data will be helpful for developing quantitative analysis strategies.