Free energy perturbation methods using molecular dynamics have been used to calculate the absolute free energy of association of two ligand-protein complexes. The calculations reproduce the snificantly more negative free energy of ation of biotin to streptavidin, compared to N-L-acetyltryptophanamlide/a-chymotrypsln. This difference in free energy ofassocition is due to van der Waals/dlspersion effects in the nearly ideally preformed cavity that streptavidin presents to biotin, which involves four ryptophan residues.One of the exciting developments in computer modeling of complex molecules in solution has been the capability to calculate relative free energies of association of these molecules and to relate these values to experiment (1, 2). This development has been catalyzed by methodological advances (3, 4) and increased computer capabilities. In favorable cases, relative free energies of association within 1 kcal/mol (1 cal = 4.184 J) of experiment have been achieved (2). In such cases, the calculations could be of use in experimental ligand design. However, inaccuracies in molecular mechanical force fields and representation of the system and, even more importantly, limitations in one's ability to completely sample the relevant regions of conformational space, have restricted the number of systems to which such free energy calculations could be applied to give chemical accuracy (5,6).Nonetheless, such free energy calculations can be very valuable and interesting even when such accuracy is not achieved, because mechanistic insight into noncovalent association in general and protein-ligand design in particular can be extracted from them (7,8 (14) have had success with this approach for nucleic acid bases in organic solvents, and Lee et al. (15) have calculated the absolute free energy of association of phosphorylcholine analogs to an immunoglobulin, but no one has carried out such a large and dramatic change as in the biotin-streptavidin association studied by Miyamoto and Koilman (9). In that paper, either all of biotin or all of biotin but the terminal COj group were mutated to dummy atoms both in water and in the protein; in either case, a AG for association in the range of -20 kcal/mol could be calculated, in good agreement with experiment. Given the approximations in that study (neglect of any changes in intramolecular energies of biotin free and bound and underestimate of translational/rotational entropy losses due to the use of hydrogen bond restraints) and the difficulty in precisely estimating the magnitude of the errors, we turned to another ligand-protein complex with much lower affinity, with a ligand of size comparable to biotin to use as a control. The results of that calculation, presented here along with the results of the biotin-streptavidin calculation, enable fundamental insights into the nature ofligand-protein associations.
METHODSWe used the free energy perturbation method to calculate the binding free energy of complexation. This method uses an easily derived equation from class...