UC Irvine

Binding free energy calculations estimate ligand binding affinities in a physically rigorous manner. Therefore, these calculations play an important role in drug discovery, estimating ligand potency of compounds before made for experimental testing to help guide discovery efforts. Different approaches to calculating binding free energies have been developed, each with their benefits and weaknesses. Absolute Binding Free Energy (ABFE) calculations calculate the binding free energy of a single ligand to its protein. ABFE calculations can be computationally expensive as they require sampling (potentially slow) conformational changes that happen upon ligand binding. This has thus far limited widespread use of the method in the pharmaceutical industry. We investigated the performance of two different approaches to calculating absolute binding free energies, a non-equilibrium and an equilibrium approach, on two different test systems. We highlighted common pitfalls or sampling problems of these calculations and identified ways on how to shed light into and assess slow degrees of freedom that impact calculated binding free energies. Relative Binding Free Energy (RBFE) calculations are an alternative approach and calculate the difference in binding free energy between two ligands. RBFE calculations have some benefits over ABFE calculations, namely that slow conformational changes of the system upon ligand binding do not need to be sampled since one of the ligands always occupies the binding site. In common RBFE approaches, one ligand is mutated into the other ligand, and the application of the method is restricted to the comparison of structurally related ligands. We implemented and tested an alternative approach for RBFE calculations, Separated Topologies, that allows for larger ligand transformations compared to standard RBFE methods, breaking the chains of ligand similarity restrictions in RBFE. Finally, we investigated the impact of protein conformational sampling on calculated binding free energies. We find there are often many different metastable protein conformations. Falling into the pit of one of these by failing to sample the other relevant states leads to inaccurate and biased results, highlighting the importance of adequate conformational sampling to obtaining converged calculated binding free energies.