On the accuracy of molecular simulation-based predictions of k_offvalues: a Metadynamics study

Abstract

Molecular simulations have made great progresses in predicting k off values-the kinetic constant of drug unbinding, a key parameter for modern pharmacology-yet computed values under-or over-estimate experimental data in a system-and/or techniquedependent way. In an effort at gaining insights on this issue, here we used an established method to calculate k off values-frequency-adaptive metadynamics with force fieldand a subsequent QM/MM descriptions of the interactions. First, using force fieldbased metadynamics, we calculate k off of the Positron Emission Tomography (PET) ligand iperoxo targeting the human muscarinic acetylcholine receptor M 2 . In line with previously performed in silico studies, the prediction (3.7 ± 0.7 · 10 −4 s −1 ) turned out to differ significantly from the experimentally measured value (1.0 ± 0.2 · 10 −2 s −1 ).Next, we use DFT-based QM/MM simulations to show that this discrepancy arises from erroneous force field energetics at the transition state. It turns out that this discrepancy is partly caused by lack of electronic polarization and/or charge transfer in commonly employed force field. We expect these issues to arise also in other systems where charged portions of the system play a pivotal role, such as protein-or DNA-protein complexes. Graphical TOC Entry-0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 1 2 3 4 5 6 7 8 9 10 11 12 zpath spath 0 4 8 12 16 20 24 28 >32 Free Energy [kcal/mol]

Multisecond ligand dissociation dynamics from atomistic simulations

“…The resulting value for K D = 1.8 × 10 −8 M underestimates the experimental value 48 7.1 × 10 −8 M by a factor of~4. Considering that we attempt to predict unbinding times on a time scale of half a minute from sub-μs MD simulations, and that a factor 20 corresponds to a free energy difference of about 3 k B T (i.e., 15 % of the barrier height in Hsp90), we find this agreement remarkable for a first principles approach which implies many uncertainties of the physical model 51 . We attribute the larger deviation in comparison to trypsin to issues with the sampling of the correct unbinding pathways: especially unbinding rates in the range of minutes to hours fall into the same timescale as slow conformational dynamics of host proteins 48 , requiring a sufficient sampling of the conformational space of the protein as a prerequisite for dcTMD pulling simulations.…”

Probing the Suitability of Different Ca2+ Parameters for Long Simulations of Diisopropyl Fluorophosphatase

“…It may result in a simulation of a system transitioning between two minima via an unphysical route that therefore produces artificially heightened barriers. Another point of consideration was recently broad to attention regarding the lack of polarization and charge transfer in common MM force fields leading to the erroneous energetics even at the geometrically adequate transition state [ 31 ].…”

Predicting ion channel conductance via dissipation-corrected targeted molecular dynamics and Langevin equation simulations

“…12,15,75 A similar issue has recently been reported for nonequilibrium simulations of ligand unbinding from receptor molecules. 84 Further application of the dcTMD-LE simulation combination to ion channels may therefore require the usage of polarizable force fields.…”