Evaluating the accuracy of the umbrella sampling plots with different dissociation paths, conformational changes, and structure preparation

Abstract

The kinetics of ligand dissociation has been found to be crucial for a good drug candidate. Therefore, examining the underlying free energy profile of the dissociation that governs the kinetics becomes important. Umbrella sampling (US), a widely used free energy calculation method, has long been used to explore the dissociation process of ligand-receptor systems. The potential of mean force (PMF) computed from US seems to always produce binding affinity and energy barriers that more or less agree with experiments. However, such PMFs are influenced by many practical aspects, like the method used to generate the initial dissociation pathway, collective variables (CVs) that used to describe the reaction coordinate (RC), and how intensive the sampling is in the conformational space restrained by the CVs. These critical factors were rarely studied. Here we applied US to study the dissociation processes of β-cyclodextrin (β-CD) and p38α complex systems. For β-CD, we used three different β-CD conformations to generate the dissociation path manually. For p38α, we generated the dissociation pathway using accelerated molecular dynamics (AMD) followed by conformational relaxing with short conventional molecular dynamics (MD), steered molecular dynamics (SMD) and manual pulling. We found that even for small β-CD complexes, different β-CD conformations will alter the height of the PMF and different dissociation directions result in appearance/disappearance of local minima. SMD poorly samples the residue sidechain movement, leading to overestimated height of PMF. On the other hand, the AMD pathway relaxed by short conventional MD sampled more accurate structures, resulting in reasonable PMF.