[gmx-users] REMD in explicit solvent

[Mark Abraham]

On 19/06/2012 5:55 PM, Esben Jannik Bjerrum wrote:
> Hi Gromacs Users
> 	I'm new to list, but hope to get a little help from someone experienced
> with setting up some replica exhange simulations in explicit solvent. It
> keeps exploding after app. 50 to 200 ps simulation time. It runs fine up
> to there, and then suddently one or a couple of atoms get unstable and
> move to much between timesteps.

All the advice of 
http://www.gromacs.org/Documentation/How-tos/Steps_to_Perform_a_Simulation 
and http://www.gromacs.org/Documentation/Terminology/Blowing_Up pertains 
to each individual system for REMD. In particular, you need to 
equilibrate each replica under its conditions and not expect them to 
tolerate a sharp jump in T and to pressure-coupling at the start. The 
fact that you are using gen-vel = yes is proof that you have not 
equilibrated yet :-)

> Heres what I have done.
>
> Build the peptide in extended configuration in pymol.
> Ran a 200 ps simulation with implicit solvent model to compact the
> structure.
> Generated a solvent box.
> Equilibrated the system for 200 ps.
> Generated a range of MDP file with varying temperatures and made
> individual .tpr files
> Ran the simulation with the switches -multi 16 -replex 1000
>
> I tried to troubleshoot a little with lowering the dt and equilibrating
> and minimizing more. Also tried different temperature coupling, as I'm
> not sure what is the most appropriate for REMD simulations.

Here's where your background reading comes in ;-) What's good for normal 
simulations is generally good for REMD. There is a school of thought 
that NVT REMD is not good because the pressure is artificially high (but 
I haven't seen anyone demonstrate artefacts from that), but NPT costs 
extra because the volume change affects your PME load balance and accuracy.

> Is there something wrong with the parameters of the mdp file?
>
> title                    = Yo
> cpp                      = /lib/cpp
> include                  = -I../top
> define                   =
> integrator               = md
> dt                       = 0.002
> nsteps                   = 10000000
> nstxout                  = 10000
> nstvout                  = 10000
> nstlog                   = 10000
> nstenergy                = 1000
> nstxtcout                = 1000
> xtc_grps                 = Protein
> energygrps               = Protein  SOL
> nstlist                  = 10
> ns_type                  = grid
> rlist                    = 0.8
> coulombtype              = PME
> rcoulomb                 = 0.8

Most would regard this set of PME parameters (most of which are coming 
from defaults) as too cheap to be effective.

Mark

> rvdw                     = 0.8
> tcoupl                   = Berendsen
> tc-grps                  = Protein      SOL
> tau_t                    = 0.1  0.1
> ref_t                    = 300  300
> Pcoupl                   = Berendsen
> tau_p                    = 1.0
> compressibility          = 4.5e-5
> ref_p                    = 1.0
> gen_vel                  = yes
> gen_temp                 = 300
> gen_seed                 = 173529
> constraints              = all-bonds
>
> Best Regards
> Esben
>
>


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20120619/5e64c96d/attachment.html>