A Benchmark for Accurate GPCR Ligand Binding Affinity Prediction with Free Energy Perturbation

G protein-coupled receptors (GPCRs) are among the most important drug targets in the pharmaceutical industry. Free energy perturbation (FEP), which can accurately predict the relative binding free energies of drug molecules, is now widely used in drug discovery. With the development of structural biology tools such as cryoelectron-microscopy (cryo-EM), the structures of a large number of GPCRs have been resolved, which provides the basis for FEP calculations. In this study, we developed an FEP protocol for GPCR FEP calculation. We performed calculations on 226 perturbation pairs of 139 ligands against 8 GPCRs, spanning 12 datasets (A2A , mGlu5 , D3, OX2 , CXCR4, β1, δ and TA1 receptors) and obtained promising results, particularly for agonist ligands in the TA1 datasets (R2, 0.58, RMSE, 1.07 kcal · mol−1 ). The average R2 is 0.61 and the average RMSE is 0.94 kcal · mol−1 , which is comparable to experimental accuracy(<1 kcal · mol−1 ). We also investigated factors that impact the accuracy of FEP results, including ligand binding pose, water placement, and protein structure. Our input structures for FEP calculation are publicly available as a benchmark dataset for future GPCR-FEP studies (https://doi.org/10.5281/zenodo.7988248). This represents the largest collection of GPCR FEP calculations known to us thus far. This work is expected to significantly contribute to the advancement of GPCR-targeted drug discovery.