Abstract: AIM　To build a model of two series of PPARγ agonists thiazolidinedione and aryketo-acid derivatives using 3D-QSAR method, and to reveal the structural features affecting the binding activity to PPARγ, which relates to antihyperglycemic and antihyperlipidemic activity and has a potential application to the treatment of type II diabetes. METHODS and RESULTS　48 agonists with selective activity for PPARγ were analyzed using CoMFA. Based upon the active conformation of rosiglitazone (BRL) extracted from its complex with PPARγ all agonists were aligned. The model from CoMFA showed a high ability to explain and predict the activity of PPARγ agonists with cross-validation correlation coefficient R²=0.656, that of non-cross-validataion R²=0.982, F_10,37=201.1, and SE=0.115. CONCLUSION　The CoMFA contour map indicates that the steric fields mainly contribute to the binding effect, and especially a bulky group in the arylketo-acid series favors in the increase of affinity for PPARγ, as compared to the thiazolidinedione.