From Affinity to Activity: Predicting Drug Efficacy through Binding Thermodynamics

The initial step in rational drug design is identifying molecules, called ligands, that bind to a target protein associated with a disease. A particularly powerful and rigorous computational approach to predict this protein-ligand binding affinity is Free Energy Perturbation (FEP). Recent advances have allowed FEP calculations to achieve near-experimental accuracy in predicting binding free energies, making it a cornerstone of modern drug discovery pipelines. However, while strong binding to its target protein is crucial for a molecule's drug efficacy, it is often insufficient alone. To produce a particular functional response, drugs need to either block the proteins' functions or modulate their activities by changing the conformational equilibrium. Unfortunately, the time scales for protein conformational changes are prohibitively long to be efficiently modeled via physics-based simulations. Thermodynamic principles, however, suggest a shortcut: binding free energies of the ligands with different receptor conformations —- each associated with a different function — may infer their efficacy. This allows us to repurpose FEP, developed to predict a ligands binding affinity, to also predict its functional activity. We introduce a robust protocol and an exhaustive study demonstrating that binding thermodynamics to active versus inactive states of a target provides a strong predictor for the efficacy of a ligand. We achieve unparalleled accuracy when validating our method on eight G protein–coupled receptors and a nuclear receptor whose three-dimensional structures in the active and inactive state are known from experiments. Motivated by these successes, we are also expanding this protocol to other target classes like integrins and ion channels. By combining it with artificial intelligence–based structure prediction, we can even use FEP on targets without experimentally determined structures. Overall, our insights have significant implications for the drug discovery process as they allow detailed predictions of a ligand's effects on its target.