A series of 2-benzoxazolinone, diazocoumarin and quinazoline derivatives have been shown to inhibit HIV replication in cell culture. To understand the pharmacophore properties of selected molecules and design new anti-HIV agents, quantitative structure–activity relationship (QSAR) study was developed using a descriptor selection approach based on the stepwise method. Multiple linear regression method was applied to relate the anti-HIV activities of dataset molecules to the selected descriptors. Obtained QSAR model was statistically significant with correlation coefficient R 2 of 0.84 and leave one out coefficient Q 2 of 0.73. The model was validated by test set molecules giving satisfactory prediction value (R 2 test) of 0.79. Molecules also were docked on HIV integrase enzyme and showed important interactions with the key residues in enzyme active site. These data might be helpful for design and discovery of novel anti-HIV compounds.