Predicting the thermodynamics and kinetics of protein-protein interaction from spontaneous association simulations

Despite the biological and pharmaceutical importance of protein-protein interactions, accurately predicting dissociation free-energies and rates remains an outstanding challenge. Here, we predict these quantities from all-atom molecular dynamics simulations of ribonuclease barnase with its inhibitor barstar. We characterized the detailed contribution of different residues at the interface, along with the role of water, identifying three distinct binding pathways. Using Langevin models, we inferred the free-energy and diffusion landscape of the process, hence the kinetic properties, systematically evaluating the model's quality across different collective variables, and obtaining robust dissociation rate predictions. Our computational scheme, based on unbiased molecular dynamics trajectories initialized in the dissociated state, is significantly simpler than previously proposed approaches, and it is very data-efficient, requiring tens of sub-microsecond simulations. The results open the way to the systematic study of protein-protein interactions, including atomic-level details on the mechanism and kinetics that are paramount for biomedical applications.