Predicting Residence Time and Drug Unbinding Pathway through Scaled Molecular Dynamics

Abstract

Computational approaches currently assist medicinal chemistry through the entire drug discovery pipeline. However, while several computational tools and strategies are available to predict binding affinity, predicting the drug–target binding kinetics is still a matter of ongoing research. Here, we challenge scaled molecular dynamics simulations to assess the off-rates for a series of structurally diverse inhibitors of the heat shock protein 90 (Hsp90) covering 3 orders of magnitude in their experimental residence times. The derived computational predictions are in overall good agreement with experimental data. Aside from the estimation of exit times, unbinding pathways were assessed through dimensionality reduction techniques. The data analysis framework proposed in this work could lead to better understanding of the mechanistic aspects related to the observed kinetic behavior.

The structural and biochemical impacts of monomerizing human acetylcholinesterase

“…However, it is generally extremely challenging for proteins exceeding 70 kDa unless the system being studied has unique properties . Additionally, larger macromolecules can affect the root‐mean‐square deviation (RMSD) of structures or atomic coordinates, a measure of similarity that is often used to analyze MD trajectories, modeling, and docking . MD simulations run on larger macromolecules can result in the RMSD losing the ability to discriminate conformation differences between structures and dynamics …”

“…34,35 Additionally, larger macromolecules can affect the root-mean-square deviation (RMSD) of structures or atomic coordinates, a measure of similarity that is often used to analyze MD trajectories, modeling, and docking. [36][37][38][39] MD simulations run on larger macromolecules can result in the RMSD losing the ability to discriminate conformation differences between structures and dynamics. 33 As the subunit molecular weight of this recombinant hAChE is~60 kDa, dimeric and tetrameric forms of hAChE pose problems for these NMR or MD approaches.…”

Accelerating Dissociative Events in Molecular Dynamics Simulations by Selective Potential Scaling

“…Similar results were obtained for α = 3. 25 Table 2. Mean unbinding times (τ sim unbind ) for inhibitors 1, 2, and 3 computed from ssMD simulations carried out with α set to 3.00, 3.25, 3.50, and 4.00, respectively.…”

“…Typically, during scaled-MD simulations, the potential energy surface is uniformly scaled by a scaling-factor, α, where 0 < α < 1. Recently, scaled-MD simulations have been used to accelerate ligand unbinding during simulations and to estimate the relative ligand residence times or unbinding times (22)(23)(24)(25). However, because the entire potential energy surface is scaled by α during scaled-MD simulations, conformational restraints are needed to maintain the structure of the protein components in the complex; without these restraints, the individual protein components would unfold (22).…”

Transient states and barriers from molecular simulations and milestoning theory: kinetics in ligand-protein recognition and compound design

“…However, ligand binding/unbinding processes can be excessively longer than microsecond simulation lengths by using computer hardware available to most scientists [13][14][15] . Various methods such as aMD, metadynamics, weighted ensemble, scale-MD, and PaCS-MD have been used to accelerate sampling the ligand binding/unbinding processes [16][17][18][19][20][21][22][23][24][25][26][27] . In terms of the binding paths found, various algorithms such as umbrella sampling 28,29 and milestoning [30][31][32][33] have been used to estimate kinetic rates and free energy profile.…”