CHARMM36 MD Jobs Result in a Protein Dimer Falling Apart/Coming Back Together Events

GROMACS version: 2020.2
GROMACS modification: Yes

I used CHARMM36 for a MD job of a homodimer protein (no cofactors). Everything looks fine in terms of the preparation (em, neutralization, nvt and npt equlibrations), but after 50 ns of the md simulation the dimers repeatedly separate and then come back. This is after corrections with trjconv (gmx trjconv -s md.tpr -f md.xtc -o md_processed.xtc -center -pbc mol -ur compact). Is this something to do with trjconv no processing properly or my system is really unstable with this force field? The same system is well behaved even at 200 ns with OPLS-AA/L. Any suggestions on what’s happening here?

my md parameters:

integrator = md
nsteps = 100000000
dt = 0.002
nstenergy = 5000
nstlog = 5000
nstxout-compressed = 5000
continuation = yes
constraint_algorithm = lincs
constraints = h-bonds
lincs_iter = 1
lincs_order = 4
cutoff-scheme = Verlet
nstlist = 40
vdwtype = cutoff
vdw-modifier = potential-shift-Verlet
coulombtype = PME
tcoupl = V-rescale
tc-grps = Protein Non-Protein
tau_t = 0.1 0.1
ref_t = 300 300
pcoupl = Parrinello-Rahman
pcoupltype = isotropic
tau_p = 2.0
ref_p = 1.0
compressibility = 4.5e-5
pbc = xyz
DispCorr = no
gen_vel = no

This is just a PBC effect. Center on one chain rather than the complex.

Thanks jalemkul. For creation of an index file (for trjconv) for one of the chains I guess I should use the em.gro structure?


For creating an index file, the choice of coordinate file is irrelevant.

Of course that makes sense (that was a silly question on my behalf). So the result is pretty great after centering on one of the chains. However I still have few short events when the rmds goes into nm. Any further suggestions on how I could remedy this?

Have you visualised the trajectory to see what is happening at those periods where you see the RMDS dramatically change?

I would guess that you need to do some further processing to ensure that things in the trajectory aren’t jumping around due to the PBC.

I was able to fix everything by using first the -pbc nojump option followed my normal usage of trjconv. Thanks for your help.