It's been many years since I last used PLUMED with gromacs and there bits that I have forgotten about and some new features etc., that I am a little unsure of.
For now, I would just like to sanity check a basic PLUMED input file. Imagine two proteins (a dimer) in water. I just want to output their bias along a straight distance collective variables (together-apart). I am a little uncertain about the PERIODIC and GRID features.
UNITS LENGTH=nm
cA: COM ATOMS=6763-6833
cB: COM ATOMS=6834-6904
d: DISTANCE ATOMS=cA,cB COMPONENTS
dz: COMBINE ARG=d.z PERIODIC=-5,5
UPPER_WALLS ARG=dz AT=4.8 KAPPA=150.0 EXP=2 EPS=2 OFFSET=0 LABEL=uwall
LOWER_WALLS ARG=dz AT=0.0 KAPPA=150.0 EXP=2 EPS=2 OFFSET=0 LABEL=lwall
metad: METAD ARG=dz SIGMA=0.1 HEIGHT=0.1 PACE=200 GRID_BIN=250 GRID_MIN=0 GRID_MAX=5
PRINT ARG=dz,metad.bias STRIDE=100 FILE=COLVAR
My concerns:
1) The only example given regarding periodicity as a function of distance was a unit cell of 20nm, so the example had PERIODIC=-10,10. I've used the same approach but for a unit cell of 10nm in z. How does this factor in fluctuations for unit cell dimensions under pressure?
2) Should the grid size (MAX, MIN) be defined by the possibly z-distance path that the two peptides can take? So should this then be 10nm? I kept it to 5 as my wall limit is just under 5.
3) Given that I defined a lower and upper wall of 5nm, how is this covered on the Grid?
4) Testing with the above parameters, I am getting negative dz (distance in z) values during the simulation run. How can I be getting a negative distance?
I am testing for a GRID_MAX = 10 at the moment.
Dear Anthony,
1) If you want periodicity in the Z-coord, you would probably need to fix the Z dimension of the box (this can be easily done with anisotropic pressure coupling and the respective compressibility of 0). From what I understand (or would expect), this should e.g. allow bias deposited at dz = -4.99 affect dynamics at dz = 4.99. If you allow your box size to fluctuate, this boundary region will be problematic. I still do not understand the physics of the PERIODIC keyword in sufficient detail, so am waiting for answers to your other question.
2) You have probably already noticed that your run will crash once the system steps outside (GRID_MIN, GRID_MAX). For that reason I usually take the "three-layer" approach - use the largest interval for grid, a smaller one for upper/lower walls and yet smaller with the INTERVAL keyword (this prevents metadynamics to push the system against the wall, generating potentially unphysical behaviour). If you want everything to be periodic, I would expect setting GRID_MIN = -5 and GRID_MAX = 5 (with no walls or interval) to do the job.
3) Now you have walls at 0 and 4.8 with a rather low force constant, so I guess this could get you into trouble - I imagine the system going from dz = 5 to dz = -5 and being affected by a (now sizeable) bias from the lower wall. However from what I know, the grid is only used for the deposited Gaussians, not for other biasing potentials such as walls.
4) Since your distance is a projection of the vector connecting COMs of the two groups on the unit vector in the Z dimension, it can clearly be either positive or negative - depending on which group is "higher" in terms of Z. This is quite useful e.g. when simulating permeation through asymmetric bilayers. If you just want the absolute value, you could take the root of the square by using COMBINE twice.
Hello Miłosz, thank you for your in depth answer. It has given me something to go away with and investigate. Thanks again.
It got a little further this time. But the system failed as it stepped outside the grid.
UNITS LENGTH=nm
cA: COM ATOMS=6763-6833
cB: COM ATOMS=6834-6904
d: DISTANCE ATOMS=cA,cB COMPONENTS NOPBC
COMBINE LABEL=dx ARG=d.x PERIODIC=NO
COMBINE LABEL=dy ARG=d.y PERIODIC=NO
UPPER_WALLS ARG=dy,dx AT=4.5,1 KAPPA=100000.0,100000.0 EXP=2,2 EPS=2,2 OFFSET=0,0 LABEL=uwall
LOWER_WALLS ARG=dy,dx AT=0.0,-1 KAPPA=100000.0,100000.0 EXP=2,2 EPS=2,2 OFFSET=0,0 LABEL=lwall
metad: METAD ARG=dy,dx SIGMA=0.3,0.3 HEIGHT=1.4 PACE=100 FILE=HILLS_AG_WT_MD BIASFACTOR=8.0 TEMP=312.0 GRID_BIN=325,250 GRID_MIN=-2,-3 GRID_MAX=6.5,3
PRINT ARG=dy,dx,metad.bias STRIDE=100 FILE=COLVAR_AG_WT_MD
It is ok to define a MASSIVE grid, assuming the system will find it impossible to every walk off it?
- Badges
- Science topic