An accurate QSRR model for the prediction of the GC×GC-TOFMS retention time of polychlorinated biphenyl (PCB) congeners

Quantitative structure-retention relationship (QSRR) models were constructed for the GC×GC-TOFMS retention time of 209 polychlorinated biphenyl (PCB) congeners. Principal component analysis (PCA) was used to recognize groups of samples with similar behavior and assist the separation of the data into training and test sets. The best multi-linear regression (BMLR) method was used for the systematic development of multi-linear regression equations; the best regression model involved four descriptors which were related to GC×GC-TOFMS chromatographic retention of PCBs. The obtained model has good predictive ability. For the test set, it gave a predictive correlation coefficient (R) of 0.988 and an average absolute relative deviation (AARD) of 3.08%. Results of a six-fold cross-validation procedure, which were in accordance with those from validation of training and test sets, demonstrated that this model was reliable. Additionally, this paper provides a simple, practical, and effective method for analytical chemists to predict the retention times of PCBs in GC.