Binding affinities of a series of benzamide inhibitors for trypsin were evaluated by molecular dynamics (MD) simulations using a linear response approach. The calculated results show that for models with one, two or three parameters, the coefficients of linear regression are similar, while the coefficients of "leave-one-out" cross- validation of the two-parameter and three-parameter models are obviously superior to that of the one-parameter model. Moreover, the binding free energies of these compounds in the test set can be well predicted by the two-parameter and three-parameter models. The predictions by the two-parameter model are a little better than those by the three-parameter model. To the compounds in the test set, the average error is only 1.15 kJ/mol between the actual and the predicted binding free energies by the two-parameter model. The relationships between the structures and activities of the benzamides can be derived from the binding free energies and the trajectories from MD simulations, which provide us with some important structural information about development of new drugs.

Key words: TRYPSIN, INHIBITOR, INHIBITOR, DRUGS, KINETIC STUDY