[gmx-users] about periodic images violation

[Anna Marabotti]

Dear gmx-users,
I'm inserting into the discussion about periodic images since I'm
experimenting a problem of minimum distance violation too. I'm doing
simulations on a dimeric protein (with no covalent bonds between the two
subunits) which derives not from a crystallographic structure but from a
model. I made several simulations changing the gen_seed in order to explore
deeply the conformational space of the protein. I used a triclinic box
(option editconf -bt triclinic -d 1 -c) filled with spc water and
neutralized with ions; in my opinion, it's quite a standard system. I
equilibrated the system with a minimization (emtol = 500 reached) and with
NVT+NPT in position-restrained mode (20+100 ps) at 310 K, then launched the
full MD for 30 ns, always at 310 K. I repeated this procedure
(NVT+NPT+FullMD) for each of the gen_seed assigned (random numbers),
starting from the same minimized structure. Obviously, in PR-NPT and full Md
simulations, I did not recalculated the initial velocity (it's a
continuation). Before starting with full MD I checked for energetic
parameters in .edr file (Pot, Kin, Tot, T, P) and all seem stable with no
apparent problems. I run the 30 ns simulations and now I'm checking for the
results.
Looking at the trajectories I see that in some (but not all) cases, the
protein moves from the center of the box to one of the edges, starting from
a time that is different in each simulation, when it happens. I run
g_mindist, and in some cases (generally when the protein doesn't move so
much) I see some "spikes" in the plot (that disappear if I apply trjconv
-pbc nojump to the trajectory), but apart from these "spikes" the minimum
periodic distance in the trajectory is at least 1.5 nm (I set van der Waals
cut-off at 1.4 nm). In other cases, however (essentially when the protein
starts moving towards the edges of the box), the minimum periodic distance
starts decreasing (in some simulations after 10 ns, in some simulations
after 20: there is not a common point after which you can see a sudden
decrease of minimum periodic distance of the system) and reaches a value
below 1.4 nm or even below 1 nm. Considering that the starting system is
more or less the same in all cases, I don't identify the reason why the
system "behaves" like this, and moreover what can I do to avoid this. My
questions are:
- do I have to enlarge the box? but I don't think that this would solve the
"motion" of the protein towards one edge of the box
- do I have to change the box? In his last message Tsjerk suggests to use a
rhombic dodecahedron box, could it be useful in my case?
- do you think it's a problem of stabilization of the system? Should I run a
deeper minimization, or a longer NVT+NPT in my system?
I'm quite puzzled especially because the system is the same in all cases,
the only thing I changed in my simulations is the gen_seed.
Could anybody give me some suggestions about it?
Thank you very much
Anna