Use of cluster and principal component analysis in quantitative structure-retention relationship study

The retention time of 42 barbituric acid derivatives was determined on a porous graphitized carbon (PGC) column using water-acetonitrile mixtures as eluents. Linear correlations were calculated between the logarithm of the capacity factor and the acetonitrile concentration in the eluent. To find the relationship between the retention parameters and physicochemical parameters of barbiturates, principal component analysis followed by two-dimensional nonlinear mapping and cluster analysis was used. Calculations were carried out on both the correlation and the correlation matrices. The retention of barbituric acid derivatives on PGC column does not depend significantly on the hydrophobicity parameters of solutes suggesting that the selectivity of PGC column is different from that of traditional reversed-phase columns, although the eluents are typical reversed-phase eluents. The comparison of the various methods of calculation (using correlation or covariance matrix) indicated that they exert considerable effect on the results and therefore their simultaneous application is highly recommended.