- The US calculation consists to simulate several windows along
the chosen restrained variableζ, in order to correctly sample
and explore the conformational space of every system state
during the unbinding event. To make sure that the potential
of mean force (PMF) obtained is correct, the sampling of each
window must overlap their neighboring sampling distribution.
This is very important when the system is sampling energy
states, such as transition states and/or energetic barriers. This
overlap allows the validation of sampling exhaustiveness along
the path, and then continuity of the free energy profile. - Another parameter of importance in this protocol is the step
size between windows. As described earlier, the sampling along
the restrained variableζis made with the definition of several
windows, centered on specific values of the bias. In order to
have a good overlap between windows, the step size, and
consequently the number of windows, must be well defined
(seeNote 17). In our case, the first value of the constrained
distance, calculated from the cMD, is about 6.2 A ̊, but a smaller
initial distance (beginning the calculation at a distance of
ζth¼5.2 A ̊) is chosen in order to correctly sample the global
minimum of the PMF. We will consider that the inhibitor is
totally separated from the protein once it is surrounded by two
water shells, corresponding in this case to a constrained dis-
tance of 36.2 A ̊, and the step between each window is set to
0.5 A ̊, representing a total number of 62 windows. - The last parameter to consider is the simulation time in each
window. The simulation time must be long enough to equili-
brate the system around each new fixed valueζthfrom the
currentζ. An equilibration step needs to be performed at the
beginning of each window, to allow the system and thusζto
sample aroundζth. In our example, for each window, 200 ps of
equilibration are considered, followed by 2 ns of production.
The total simulation time for the 62 windows is 136.5 ns.
3.5.2 Results for the
Unbinding Process of
p38—Biaryl-
Triazolopyridine Ligand
- Calculations were performed on a Quadro K5000 GPU unit
from a standard workstation, with AMBER14 program com-
piled with CUDA v6.0. Performance is estimated to 20 ns/day,
meaning that the whole calculation take about 6.5 days. After
the end of the calculation, cpptraj analysis module of Amber-
Tools has been used to center, align, and analyze the trajectory.
RMSD analysis, combined with distance measurement and
solvation estimation, could thus be performed to check the
stability of the protein kinase domain and estimate the progress
of the unbinding event. Then, the trajectory could be checked
with the VMD visualization software in order to verify if the
process is complete, meaning that the ligand is positioned in
the solvent, with no interactions left with the protein kinase.
Enhanced Molecular Dynamics 419