[AMBER] MD stops due to the distortion of a molecule

From: Sergey Samsonov <sergey.samsonov.ug.edu.pl>
Date: Thu, 17 May 2018 10:23:57 +0200

Dear AMBERs,

I'm encountering the following problem. I'm trying to simulate O- and
N-phosphorylated saccharides combining GLYCAM06 and phosphate parameters
developed for the phosphorylated amino acids from

First, I introduced the parameters for the phosphates into the .dat
file, which I read in leaprc. Then, I added missing parameters from
GLYCAM06 and then gaff in case some parameters were still required.
These parameters corresponded to the part of the molecule, where the
phosphate group was connected to the sugar. The final file with the
additional parameters is attached (extra_phosph.dat). In case I simulate
molecules with non-protonated phosphate groups: PO3-O-(sugar ring) or
PO3-O-NH-(sugar ring), everything is fine. The geometries look correct,
and there are no problems with multiple (around 50) simulations of 100
ns already carried out. The same applies for the protonated
O-phosphates: HO-PO2-O-(sugar ring). However, when I simulate a
protonated N-phosphate HO-PO2-NH-(sugar ring), the simulation goes well
for certain time, which is always different and could vary from hundreds
of picoseconds to tens of nanoseconds, but then it stops. In order to
investigate, what happens, I wrote the frames more often, and I found
the error related to the exceeding vmax limit, which corresponded to
distorted geometry when it was visualized in VMD. The corresponding
error message was:"ERROR: Calculation halted. Periodic box dimensions
have changed too much from their initial values. Your system density has
likely changed by a large amount, probably from starting the simulation
from a structure a long way from equilibrium." This error message is to
be expected because all of a sudden bonds got unphysically distorted,
while their lengths were huge. I tried the same simulations on CPUs and
GPUs, with sander and with pmemd, and the effect is always the same.

Since the only parameter, which differs for the molecules without errors
and with errors is the dihedral HO-OQ-P-Ng, which is however defined for
a general angle

X -OQ-P -X    3    0.75    0.0    3.    from X -OH-P -X, I tried to
define this parameter explicitly for HO-OQ-P-Ng using the parameter from
gaff for X-oh-p5-X 3 1.6 0.0 3. The effect did not change meaning that
the minimum energy value does not affect the event of simulation

Similar problem was already discussed in the mailing list:

The solution was GPU-related, which is not my case here.

Do you have any idea how to deal with this problem? Maybe some
additional restraints could be used in order to prevent the geometries
leading to such a behaviour (f.i. to avoid possible planarity of
something, but the question is what exactly)?

Thank you very much in advance and best regards,


Sergey Samsonov, PhD
Project Leader at Laboratory of Molecular Modeling,
Department of Theoretical Chemistry,
Faculty of Chemistry,
University of Gdansk
ul. Wita Stwosza 63
80-308 Gdańsk, Poland
phone: +48 58 523 5166

AMBER mailing list

Received on Thu May 17 2018 - 01:30:03 PDT
Custom Search