Quantitative structure-activity relationships (QSARs) techniques are routinely used in modern computer-aided drug design. In this review, the common generalization and computational procedures of QSAR methods (CoMFA, CoMSIA, and HQSA) are described in detail. The predictive ability of CoMFA and CoMSIA models depends directly on the quality of molecular alignment, the selection of probe, the difference of steps and type of charge. Moreover, it is worth noting that the active conformation plays a key role in molecular alignment, and it is a very difficult task to select the active conformations for drug molecules. The approaches to determine the active conformation are also reviewed. Furthermore, strategies including the selection of different fields and the chemometric methods for QSAR model to improve predictive capabilities between the structures of drugs and the biological activities are suggested in this review. The predictive ability of HQSAR models is directly dependent on the hologram length, fragment size, and distinction parameters. By using these techniques, our recent case studies of QSAR on two categories of drugs are presented. One is a series of the inhibitors of estrogen receptor (ER), 3-arylquinazolinethione derivatives, which is a key drug target for the treatment of osteoporosis and breast cancer. The other is a series of inhibitors of human eosinophil phosphodiesterase, 5,6- dihydro-(9H)-pyrazolo[3,4-c]-1,2,4- triazolo [4,3-a]pyridines, which is a drug target for the treatment of inflammation. Satisfactory models of QSAR (with high predictive ability) on two categories of drugs were obtained with optimized parameters. According to the models of QSAR obtained in our study, new drug molecules with higher activity were proposed.