- Posts: 210
Severe inhomogeneity of large DNA+PW CG simulation
- visvaldas
-
Topic Author
- Visitor
7 years 3 months ago #6848
by visvaldas
Severe inhomogeneity of large DNA+PW CG simulation was created by visvaldas
I am running a large simulation of linear 200bp ssDNA + Polarized Water + Ions, dodecahedral box, at 372 K. The system is very large (>1000Angs, >4 mln CG particles). I am using 4.6.7 GROMACS. Something very odd happens. (Notably, the system is quite hard to equilibrate without crash, but I managed to do that by increasing number of steps in EM step). At the end of energy minimization, there is a visible thinning of the density at the boundary between (it seems) the periodic cells, perhaps several angstrom wide. When I start equilibration, the thinned region starts expanding, and in the end of 500psequilibration ( I get a "bubble" of very dense PW+Ions in the (I guess) center of the cell , and the rest of the space is "rarefied", almost to the point of vacuum. DNA is not affected directly by this. The VMD illustration with +X,+Y+Z copies of the cell added will explain better than words:
postimg.org/gallery/20n5ygsrc
(sorry, the sequence of pics is in the opposite order).
My thinking is that the pressure gets very unevenly distributed but the program thinks it's fine because the overall average pressure is correct, or perhaps there is something wrong with the communication of the particles across cell boundaries.
I tried to do the simulation in GROMACS 5.1.4 but I couldn't manage to carry on even several md steps without crash. I guess next thing to try a pressure control instead of Berendsen.
Does anybody has a clue what's going on?
The equilibration mdp file was taken from DNA Martini tutorial and is the following:
integrator = md
dt = 0.010
nsteps = 50000
nstxout = 0
nstvout = 0
nstfout = 0
nstlog = 10000
nstenergy = 10000
nstxtcout = 10000
xtc_precision = 100
nstlist = 10
ns_type = grid
pbc = xyz
rlist = 1.4
coulombtype = Shift
rcoulomb_switch = 0.0
rcoulomb = 1.2
epsilon_r = 15
vdw_type = Shift
rvdw_switch = 0.9
rvdw = 1.2
tcoupl = berendsen ;v-rescale
tc-grps = System
tau_t = 2.0 ;0.5
ref_t = 372
Pcoupltype = isotropic
Pcoupl = berendsen ;parrinello-rahman
tau_p = 3.0 ;4.0
compressibility = 3e-4 ;4.5e-5
ref_p = 1.0
gen_vel = no
gen_temp = 372
gen_seed = 473529
constraints = none
postimg.org/gallery/20n5ygsrc
(sorry, the sequence of pics is in the opposite order).
My thinking is that the pressure gets very unevenly distributed but the program thinks it's fine because the overall average pressure is correct, or perhaps there is something wrong with the communication of the particles across cell boundaries.
I tried to do the simulation in GROMACS 5.1.4 but I couldn't manage to carry on even several md steps without crash. I guess next thing to try a pressure control instead of Berendsen.
Does anybody has a clue what's going on?
The equilibration mdp file was taken from DNA Martini tutorial and is the following:
integrator = md
dt = 0.010
nsteps = 50000
nstxout = 0
nstvout = 0
nstfout = 0
nstlog = 10000
nstenergy = 10000
nstxtcout = 10000
xtc_precision = 100
nstlist = 10
ns_type = grid
pbc = xyz
rlist = 1.4
coulombtype = Shift
rcoulomb_switch = 0.0
rcoulomb = 1.2
epsilon_r = 15
vdw_type = Shift
rvdw_switch = 0.9
rvdw = 1.2
tcoupl = berendsen ;v-rescale
tc-grps = System
tau_t = 2.0 ;0.5
ref_t = 372
Pcoupltype = isotropic
Pcoupl = berendsen ;parrinello-rahman
tau_p = 3.0 ;4.0
compressibility = 3e-4 ;4.5e-5
ref_p = 1.0
gen_vel = no
gen_temp = 372
gen_seed = 473529
constraints = none
Please Log in or Create an account to join the conversation.
- peterkroon
-
- Offline
- Gold Boarder
Less
More
7 years 3 months ago #6858
by peterkroon
Replied by peterkroon on topic Severe inhomogeneity of large DNA+PW CG simulation
Polarized water should be used with either reaction-field or PME. Could you try either to see if it solves the problem?
Please Log in or Create an account to join the conversation.
- visvaldas
-
Topic Author
- Visitor
7 years 3 months ago #6869
by visvaldas
Replied by visvaldas on topic Severe inhomogeneity of large DNA+PW CG simulation
Thank you for the tip about PW, I will try that too! Just now I also realized epsilon_r should be set to 2.5 for PW.
I was able to fix the "bubble" formation, however, by changing 2 things simultaneously in the above mdp file (I suspect that just one of them really is important, perhaps it's the velocities):
1) I set "tcoupl=V-rescale"
(I thought this will be less prone to crashes compared to pcoupl, why don't I start with tcoupl before messing with pcoupl)
2) I set "gen_vel = yes"
(I realized that perhaps the particles are in some very bad initial speed distribution when coming from EM step; I am not sure why the program doesn't complain, but hey! stranger things have happened)
And, actually, the run was much more stable than before, with no LINCS warnings, etc.
Perhaps this will be useful to other users.
I was able to fix the "bubble" formation, however, by changing 2 things simultaneously in the above mdp file (I suspect that just one of them really is important, perhaps it's the velocities):
1) I set "tcoupl=V-rescale"
(I thought this will be less prone to crashes compared to pcoupl, why don't I start with tcoupl before messing with pcoupl)
2) I set "gen_vel = yes"
(I realized that perhaps the particles are in some very bad initial speed distribution when coming from EM step; I am not sure why the program doesn't complain, but hey! stranger things have happened)
And, actually, the run was much more stable than before, with no LINCS warnings, etc.
Perhaps this will be useful to other users.
Please Log in or Create an account to join the conversation.
Time to create page: 0.103 seconds