Quantitative structure-retention relationship (QSRR) study was performed forpredicting the gas chromatography retention indices of 96 polycyclic aromatichydrocarbons (PAHs) on SE-52 capillary columns.First,descriptors werecalculated using Materials Studio (MS) software.Then the best-fittingdescriptors were selected by using stepwise multiple linear regression(SW-MLR).Then QSRR models were built by adopting multiple linearregression (MLR) and back propagation-artificial neural network(BP-ANN).The results showed prediction correlation coefficient R of 0.9971and 0.9976 for MLR and BP-ANN respectively.