Computer simulation of protein-ligand interactions: challenges and applications.

The accurate modeling of protein-ligand interactions, like any prediction of macromolecular structure, requires an energy function of sufficient detail to account for all relevant interactions and a conformational search method that can reliably find the energetically favorable conformations of a heterogeneous system. Both of these prerequisites represent daunting challenges. Consequently, the routine docking of small molecules or peptides to proteins in their correct binding modes, and the reliable ranking of binding affinities remain unsolved problems. Nonetheless, computational techniques are continually evolving so as to broaden the range of feasible applications, and the accuracy of predictions and theoretical approaches can often be of great help in guiding and interpreting experiments. We discuss the energetics of protein-ligand systems and survey conformational searching techniques. We illustrate how molecular modeling of a protein-ligand complex sheds light on the observed resistance of a mutant dihydrofolate reductase to the antibiotic trimethoprim. In another example, we show that relaxation of side chains in different crystal structures of the same complex, benzamidine bound to trypsin, is needed to draw sensible conclusions from the calculations. The results of these relatively simple conformational searches underscore the importance of incorporating protein flexibility in simulations of protein-ligand interactions, even in the context of relatively rigid binding pockets.