Challenges of Absolute Binding Free Energies for Membrane Exposed Binding Pockets

Absolute binding free energy (ABFE) calculations are finding increasing use within drug discovery programs, particularly as they can be performed on very different ligand scaffolds and direct comparisons made between them. They also have the potential to provide a route to target selectivity prediction. To date, most work has focused on soluble proteins, with relatively little focus on membrane proteins. Structural studies are increasingly revealing the presence of lipid-exposed allosteric binding sites on membrane proteins. The exposure to lipids rather than water is presumed to make ABFE calculations considerably more challenging. To investigate this and highlight other potential difficulties associated with such binding sites, we explored the performance of ABFE calculations on two G-protein-coupled receptor systems with lipid-exposed binding sites; the P2Y1 receptor and the GCGR receptor. We show that despite both proteins being GPCRs with lipid- exposed binding sites, the performance of ABFE calculations differs significantly. ABFE results for P2Y1R are in excellent agreement with experiment (mean absolute error (MAE) of 1.21 kcal/mol), but the results for GCGR were significantly poorer with the MAE 4.74 kcal/mol being the best we could achieve with current resources. Nevertheless, the GCGR system serves as an excellent vehicle to demonstrate the more problematic issues which can exist, including vastly different apo versus bound state, the presence of lipids solvating the apo state and sampling problems exacerbated by the lipid environment. We discuss how some of these aspects can be dealt with, while others will clearly require more work. We hope this work will help drive the community forward in tackling these problems.