Comparative Molecular Field Analysis (CoMFA) methods were used to produce a 3D-QSAR model that correlated the catalytic efficiency of rat hepatic aryl sulfotransferase (AST) IV, expressed as log(k(cat)/K(m)), with the molecular structures of its substrates. A total of 35 substrate molecules were used to construct a CoMFA model that was evaluated on the basis of its leave-one-out cross-validated partial least-squares value (q(2)) and its ability to predict the activity of six additional substrates not used in the training set. The model was constructed using substrate conformations that favored (1) proton abstraction by the catalytic histidine residue, (2) an in-line sulfuryl-group transfer mechanism, and (3) constraints imposed by the residues lining the substrate binding pocket of a homology model of AST IV. This CoMFA model had a q(2) value of 0.691, and it successfully predicted the activities of the six molecules not used in the training set. A final CoMFA model was constructed using the same methodology but with molecules from both the training set and the test set. Its q(2) value was 0.701, and it had a non-cross-validated r(2) value of 0.922. The contour coefficient map generated by this CoMFA was overlaid on the amino acids in the substrate-binding pocket of the homology model of AST IV and found to show a good fit. Additionally external validation was obtained by using the CoMFA model to design substrates that show high activities. These results establish a methodology for prediction of the substrate specificity of this sulfotransferase based on CoMFA methods that are guided by both the homology model and the catalytic mechanism of the enzyme.