A Multiscale Approach To Understanding Protein Ligand Binding Process

Abstract

Owing to the recent rapid increase in computational power and the improvement of the algorithms, it is now possible to reproduce the whole process of protein–ligand binding in a molecular dynamics (MD) simulation. However, because the ligand-binding process is a stochastic process, it is necessary to repeat the MD simulation many times to fully understand its statistical nature. Therefore, it is still difficult to apply the MD simulation to various protein–ligand systems and to elucidate the general properties of the ligand-binding process. To solve this problem, we are developing a multi-scale approach that combines the coarse-grained (CG) and the all-atom (AA) MD methods. Recently, we have shown that the ligand-binding process can be reproduced in the CGMD simulations with the MARTINI force field [Negami et al. J Comput. Chem. 35, 1835–1845 (2014)]. In this study, the CGMD simulations were applied to two different protein–ligand systems. For each system, 1–4 μs simulations were performed 50–100 times with different initial ligand placement and different initial velocities. The binding and unbinding rate constants and the dissociation constants calculated from the CGMD trajectories were consistent with the experimental values. Furthermore, the ligands tended to enter the ligand-binding pockets through specific pathways. In the present study, we optimized the pathway of a protein–ligand system toward the minimum free-energy pathway using the string method. The pathway was discretized with 32 images represented by a set of interatomic distances between the protein and the ligand. At each position of the image, free-energy gradient was calculated using the AAMD simulation and the pathway was optimized accordingly. The simulation was performed for 30×32 ns in total. We will discuss the ligand-binding process based on the free-energy profile calculated along the optimized pathway.