Title: Probing the Kinetics of Protein-Ligand Binding Using Molecular Dynamics Simulation
Abstract: Protein-ligand interactions are pivotal to the functioning of biological processes inside cells. Molecular dynamics (MD) simulations can provide mechanistic insight of ligand binding and unbinding processes with atomistic detail. But most protein-ligand interactions are characterized as rare events for which the timescales can often go beyond milliseconds. They are difficult or sometimes impossible to probe using conventional MD simulations with the present day computational resources. Enhanced sampling methods like umbrella sampling and metadynamics can be used to accelerate the dynamics but the application of artificial biasing force makes it difficult to recover kinetic properties. Weighted Ensemble and Milestoning are two powerful path sampling techniques to study the kinetics of such rare events, although both require a significant amount of computational effort. We developed the Weighted Ensemble Milestoning (WEM) scheme, which combines the strength of these two methods to calculate the kinetics and the free energy profile from short and low cost MD trajectories. We study the unbinding and binding of 4-hydroxy-2-butanone (BUT) ligand to FKBP protein using the WEM protocol. We also propose an analytical diffusion model to calculate the binding rate constant and free energy, utilizing the WEM trajectories within the milestoning framework. Although the total computational cost of WEM simulation is less than 100 ns, the ligand binding and unbinding timescales, rate constants and the binding affinity results are within chemical accuracy to the literature values obtained using 30 μs conventional MD simulation. Additionally, the trajectories belonging to each milestone could be parallelized over multiple computing nodes leading to significant reductions in the wall clock time. Thus, WEM provides a computationally inexpensive approach to predict multiple important experimental observables regarding ligand-protein interactions and can find potential application in in-silico design of therapeutic agents.