normal production mdrun terminated with LINCS WARNING

  • rheamaz5
  • rheamaz5's Avatar Topic Author
  • Visitor
7 years 10 months ago #5629 by rheamaz5
Dear Martini users,

I am helping a friend out with martini membrane protein simulations. He is simulating a peripheral protein in a mixed lipid bilayer. The system was built using CHARMM GUI MARTINI Maker.There were 6 steps of equilibration performed of 1ns each using the below mentioned parameters and values :

define = -DPOSRES -DPOSRES_FC=50 -DBILAYER_LIPIDHEAD_FC=10
integrator = md
tinit = 0.0
dt = 0.020
nsteps = 50000

nstxout = 1000
nstvout = 1000
nstfout = 1000
nstlog = 1000
nstenergy = 1000
nstxtcout = 1000
xtc_precision = 100

ns_type = grid
pbc = xyz
nstlist = 10
cutoff-scheme = Verlet
rlist = 1.2
vdwtype = Cut-off
vdw-modifier = none
rvdw_switch = 1.0
rvdw = 1.2
coulombtype = pme
rcoulomb = 1.2


tcoupl = v-rescale
tc-grps = protein CHOL_POPC_DPSM_DPCE W_ION
tau_t = 1.0 1.0 1.0
ref_t = 303.15 303.15 303.15

; Pressure coupling:
Pcoupl = berendsen
Pcoupltype = semiisotropic
tau_p = 5.0 5.0
compressibility = 3e-4 3e-4
ref_p = 1.0 1.0

; GENERATE VELOCITIES FOR STARTUP RUN:
gen_vel = no

refcoord_scaling = all

This was followed by production run with the mdp file settings as below :
ntegrator = md
tinit = 0.0
dt = 0.020
nsteps = 2500000000

nstxout = 2000
nstvout = 2000
nstfout = 2000
nstlog = 2000
nstenergy = 2000
nstxtcout = 2000
xtc_precision = 100

ns_type = grid
pbc = xyz
nstlist = 10
cutoff-scheme = Verlet
rlist = 1.2
vdwtype = Cut-off
vdw-modifier = none
rvdw_switch = 1.0
rvdw = 1.2
coulombtype = pme
rcoulomb = 1.2

tcoupl = v-rescale
tc-grps = protein CHOL_POPC_DPSM_DPCE W_ION

; Pressure coupling:
Pcoupl = berendsen
Pcoupltype = semiisotropic
tau_p = 5.0 5.0
compressibility = 3e-4 3e-4
ref_p = 1.0 1.0

; GENERATE VELOCITIES FOR STARTUP RUN:
gen_vel = no
refcoord_scaling = all


He was able to execute the mdrun on his desktop with gromacs 5.1.0. However, frequently the mdrun is getting terminated with the following error :



Step 16714814, time 334296 (ps) LINCS WARNING
relative constraint deviation after LINCS:
rms 0.131115, max 3.385360 (between atoms 325 and 327)
bonds that rotated more than 30 degrees:
atom 1 atom 2 angle previous, current, constraint length
321 323 91.1 0.3343 0.1797 0.3300
323 325 90.0 0.3161 0.9815 0.3100
325 327 90.1 0.2994 1.3595 0.3100
327 329 90.2 0.3088 1.1087 0.3100
329 330 89.9 0.3161 0.4481 0.3100
330 332 89.9 0.3136 0.4520 0.3100
332 334 31.5 0.3100 0.3589 0.3100
Wrote pdb files with previous and current coordinates

Step 16714815, time 334296 (ps) LINCS WARNING
relative constraint deviation after LINCS:
rms 609002658.963345, max 15898352640.000000 (between atoms 329 and 330)
bonds that rotated more than 30 degrees:
atom 1 atom 2 angle previous, current, constraint length
321 323 88.8 0.1797 2.2707 0.3300
323 325 140.6 0.9815 191686.4531 0.3100
325 327 116.5 1.3595 191662.5156 0.3100
327 329 99.9 1.1087 49.1894 0.3100
329 330 41.5 0.4481 4928489472.0000 0.3100
330 332 82.0 0.4520 4928489472.0000 0.3100
332 334 94.0 0.3589 82.4154 0.3100
334 336 30.5 0.3118 7.7404 0.3100
336 339 90.0 0.3100 0.4559 0.3100
563 565 110.1 0.4909 178732.9062 0.4904
563 566 104.3 0.6018 178732.7031 0.6019
565 566 55.7 0.2718 0.8946 0.2719
591 593 76.2 0.4907 0.4245 0.4904
591 594 46.3 0.6018 0.5443 0.6019
593 594 39.7 0.2718 0.3656 0.2719
Wrote pdb files with previous and current coordinates
Segmentation fault (core dumped)

what could be causing this?

These atoms are from the protein that are causing LINCS warnings. Should we run a coarse grained simulation of just the protein in water and then the membrane in water and then take the simulated protein and membrane for alignment and further production run? Here, he provided the atomistically simulated structure of the protein for conversion to CG as an input for CHARMM GUI MARTINI MAKER and used the protein-membrane system option to generate the setup.

Please Log in or Create an account to join the conversation.

More
7 years 9 months ago #5646 by Pim
Others might have a better idea, but for any simulation it is possible that at some point an instability occurs. Sometimes that just happens. Short answer: It seems your box is blowing up after a small numeric instability that you can't really do much about.

It seems that your parameters are realistic (although if you do 50 microseconds you might want to save less frequently than every 2000 steps as the files will get really really huge) and only after 16 million steps you get an instability. At this point, there is no equilibration procedure that will prevent this. The Martini proteins should be stable using 20 fs (are you using Elnedyn/elastic network?), so I'm not sure there is anything you can do but restart the simulation. Which residue's beads are causing it (what are atoms 321-334?) and is it always happening in the same place? I hope the system isn't too small (I suspect it may be a bit small as you're running 50 microseconds on a desktop). In my experience it happens more often if you spread a small box over multiple processors.

If you give me more details about the system itself I might be able to suggest something else, but as I said it may be just bad luck.

Please Log in or Create an account to join the conversation.

Time to create page: 0.121 seconds