We report a 3D QSAR study of almost 300 structurally diverse small molecule antagonists of the integrin α4β1 whose biological activity spans six orders of magnitude. The alignment of the molecules was based on the conformation of a structurally related ligand bound to the αIIBβ3 and αvβ3 integrins in X-ray crystallographic studies. The molecular field method, CoMSIA, was used to generate the 3D QSAR models. The resulting models showed that the lipophilic properties were the most important, with hydrogen bond donor and steric properties less relevant. The models were highly significant (r(2)=0.89, q2(LOO)=0.67, r(2) (test set)=0.76), and could make robust predictions of the data (SEE=0.46, SEP=0.78, SEP (test set)=0.66). We predicted the antagonist activities of a further ten compounds with useful accuracy. The model appears capable of predicting α4β1 integrin antagonist activity to within a factor of five for compounds within its domain of applicability. The implications for design of improved integrin antagonists will be discussed.