[gmx-users] REMD with 'bad contacts' error

[Justin A. Lemkul]

Sheeba Jem wrote:
> 
> Dear Gromacs users,
> 
>  I am having trouble running REMD for a system containing one peptide 
> molecule on the surface of a lipid membrane, the system contains the 
> following:
> 
> Protein   1
> POPC    128
> Water     4847
> Na+         9
> Cl-          15
> 
> The total number of atoms in the system is 21483. I have 50 replicas 
> with temperatures distributed from 250 to 400 K. After setting up the 
> system I minimized and  equilibrated the system for 14 ns at three 
> temperatures: 250 K, 300K and 350 K. I take the output file from the 250 
> K run and use that as the starting structure for the temperatures 
> between 250 to 300 K; similarly the output file from 300 K as the 
> starting structure for replicas between 300 to 350 K and for the 
> replicas between 350 to 400 K I use the output from the 350 K run. I 
> further equilibrate these structures at the replica temperature for 10 
> ns. The output from these 10 ns runs are then used as starting 
> structures for the replica exchange simulation. However when the replica 
> exchange simulation begins to run, it crashes after 2 ps with a bad 
> contact error:
> 

Your equilibration protocol doesn't make much sense.  First, you're not 
equilibrating properly under all conditions, and then (in your .mdp file) you're 
re-generating velocities so you just end up destroying any equilibration you had 
previously achieved.  Membranes are particularly sensitive to proper 
equilibration, so I suspect this is your root problem.  Equilibrate each system 
at the desired temperature, maintaining the ensemble by passing the appropriate 
.cpt file to grompp (with "gen_vel = no" so as not to override the existing 
velocities).

> 
> t = 2.008 ps: Water molecule starting at atom 21421 can not be settled.
> Check for bad contacts and/or reduce the timestep.
> Wrote pdb files with previous and current coordinates
> Jul  4 18:07:21 2011 22666 4 7.04 handleTSRegisterTerm(): TS reports 
> task <7> pid <15856> on host<cmp-13-9> killed or core dumped
> 
> 
> I use the Gromacs version 4.0.5 for all the simulations. Since the 
> starting structure for each replica has been well equilibrated I am not 
> sure how there could be hard contacts in the system. I looked at the 

The configurations may be somewhat equilibrated, but you're killing that by 
generating velocities.

> input structures and the trajectories from the equilibration runs and I 
> could not find anything strange with the system leading to hard 
> contacts. Also the membrane remains intact for the high temperature 
> replicas. Since I could not find anything to change in the system, I 
> tried running REMD reducing the time step from 2 fs to 1 fs which also 
> gave a similar error:
> 

If you get the same problem, it's likely still the thermostatting/velocity 
generation that's the problem.

> t = 2.020 ps: Water molecule starting at atom 18919 can not be settled.
> Check for bad contacts and/or reduce the timestep.
> Wrote pdb files with previous and current coordinates
> 
> I then tried with a timestep of 0.1 fs and got the same bad contacts error:
> 
> t = 0.101 ps: Water molecule starting at atom 20251 can not be settled.
> Check for bad contacts and/or reduce the timestep.
> Wrote pdb files with previous and current coordinates
> Jul  4 18:34:17 2011 25639 4 7.04 handleTSRegisterTerm(): TS reports 
> task <5> pid <17349> on host<cmp-4-5> killed or core dumped
> 
> 
> I am not sure if reducing the timestep further would help therefore I 
> looked at the temperature and pressure coupling. For all the above 
> simulations I had used nose-hoover thermostat and a  parrinello-rahman 
> barostat with semi-isotropic pressure coupling. I had previously 
> 'successfully' ran a REMD simulation of the peptide in water with 
> isotropic coupling, the difference in the two .mdp files were the type 
> of pressure coupling and changes in the bond contraint parameters (I 
> have attached both the mdp files). Since semi-isotropic pressure 
> coupling reproduces membrane properties well, I had used it for the 
> peptide-lipid system. To see if changing the coupling type made a 
> difference, with the 0.1 fs time step, I changed the coupling type to 
> isotropic and this time the job crashed with the lincs warning:
> 

Membranes are more sensitive, especially to temperature and pressure coupling 
and the state of equilibration.  Proteins in water are far more robust and take 
a lot of beating before they explode.  Therefore, the comparison is not a 
particularly good one.  Apples and oranges.

-Justin

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================