A Self-Adaptive Curved-Path Umbrella Sampling Approach for Protein-Protein Dissociation Energetics
Dhananjay C. Joshi1,2,3*, Jung-Hsin Lin1,3,4,5
1TIGP-CBMB, Academia Sinica, Taipei, Taiwan
2IBS, NTU, Taipei, Taiwan
3RCAS, Academia Sinica, Taipei, Taiwan
4IBMS, Academia Sinica, Taipei, Taiwan
5School of Pharmacy, College of Medicine, NTU, Taipei, Taiwan
* Presenter:Dhananjay C. Joshi
Protein-protein association and/or dissociation are central in almost all biological process. Computing statistical mechanics-based free energy difference between thermodynamically well-defined end-states of a reaction is central in biophysics. In this regard, the umbrella sampling method is widely adopted and considered as an effective pathway-based method for enhancing the samplings of the rare events of the reaction, which otherwise will not be possible by unbiased MD simulations. Construction of potential of mean force (PMF) using such sampled data to estimate the free energy difference is an established technique. Although umbrella sampling is a rather mature method, the choice of reaction coordinate in the PMF construction is still a subtle issue. In macromolecular association and/or dissociation reaction, the reaction coordinates were often assigned as a predefined vector between the interacting molecules, but such vectorial reaction coordinates are highly artificial and have their limitations for general applications. In this work, we proposed the curved-path umbrella sampling simulation approach. A reaction of protein-protein dissociation was simulated by conducting multiple independent simulations and the PMFs were constructed. The physical dissociation trajectories were traced from representative conformations obtained from clustering analysis on the sampled windows. To our surprise, a subset of these PMF profiles are observed to be nearly identical. Importantly, they seem to constitute a firm lower bound, even though the physical pathways are distinct. Finally, a scheme to discriminate decoys from the protein-protein docking is proposed, using our multiple-walker curved-path umbrella sampling approach, which demonstrates the potential of this method for the very challenging decoy discrimination problem.

Keywords: Molecular Dynamics Simulations, Protein-protein interactions, Macromolecular interaction energetics