Efficient Sampling of Free Energy Landscapes for the Calculation of Protein-Protein Binding Affinities in Membranes.

The accurate simulation of realistic biomembranes is a long-term goal in the field of membrane biophysics. Efforts to simulate increasingly complex lipid bilayers, consisting of multiple lipid types and proteins, have been hindered by the shortcomings of current force fields, both coarse-grained and all-atom, in the modeling of protein-protein and protein-lipid interactions. Due to the fundamental importance of protein dimerization to cellular signaling and protein trafficking, the study of protein-protein association and the related dimerization free energies has received significant attention in both simulations and experiments. Detailed comparisons of simulation results with NMR, crystallography, and FRET studies served as a test of the accuracy of the simulation methods and provided insights into the underlying structural distributions and thermodynamic driving forces defining the interactions. These comparisons have led to the conclusion that existing state-of-the-art simulation methods have failed to effectively sample the equilibrium between associated and dissociated states, leading to inaccurate estimates of binding constants and misrepresentation of the associated structural ensembles. Here, we discuss the drawbacks of previously used protocols and review our systematic development of effective computational methods for enhanced sampling simulations that exhaustively sample the native and non-native dimer conformations and provide precise estimates of the associated equilibrium binding constants. We conclude by identifying the most important current challenges to the field that must be met in closing the gap between simulation and experiment in the study of protein-protein association in the membrane.