Abstract: As the third-generation antibacterial macrolides，ketolides not only have good activity against drug-resistant strains，but also overcome the induction of resistance to drugs，and have become the focus in the investigation of the macrolides.The 2D-QSAR(2-dimensional quantitative structure-activity relationships) and HQSAR(hologram QSAR) were used to develop QSAR models with a training set of 49 6-O-aryl ketolide derivatives.Both 2D-QSAR and HQSAR models yielded good results，including high conventional correlation(r²) coefficients(0.849，0.975) and across-validated(q²) coefficients(0.803，0.926).Both of the models were validated by an external test set of 9 compounds，which gave a satisfactory predictive r² of 0.782 and 0.878，respectively.This study might provide insights into further structural optimization of macrolides.