Significant evidence exists that accurate relative binding energies can be computed using free energy perturbation (FEP) methods for systems with high-resolution crystal structures that are bound to a ligand from the chemical series of interest. However, the ability to apply FEP to homology models and crystal structures without a ligand from the chemical series of interest would greatly increase the domain of applicability of FEP. Here, we present the first large-scale validation study of FEP applied to homology models. We explore a number of target classes (kinase, bromodomain, GPCR, and a protein-protein interaction) and degrees of homology. We describe the homology modeling protocol and investigate the sensitivity of the results to the homology (full protein and binding site only) between the template and target. Encouragingly, we find that the accuracy of the results degrades little, if any, when FEP is run on a homology model as compared with a high-resolution co-crystal structure. The key to success seems to lie in the ability to generate a reasonable initial pose and binding site arrangement such that the FEP simulations can refine the structures into the correct energy basin. These results have significant implications on the domain of applicability of FEP.

On the use of homology models for binding free energies predictions

251st National Meeting (2016)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.