On Sun, Jan 15, 2017 at 4:39 AM, Andrew Schaub <aschaub.uci.edu> wrote:
> Good Afternoon,
>
> I am setting up a simulation for constant pH replica exchange MD in
> explicit solvent. The two tutorials on Jason's blog are great:
> http://jswails.wikidot.com/ph-remd
> http://jswails.wikidot.com/explicit-solvent-constant-ph-md
>
> I am trying to set up a constant pH REMD in explicit solvent using the two
> tutorials as a guide.
>
> My system is larger than that described in the tutorials. I am studying
> conformational changes in a system in the range of pH 5.5 to 9, so I am
> doing 8 replicas (pH 5.5 to 8.0, in 0.5 stepsize).
> Atom Size: 38986
> Protein Residues: 376
> Titrable Residues: 50
> Waters: 11001
>
> How often should I attempt replica exchange swaps?
How often should I attempt to change protonation states?
>
> The tutorial has replica exchange swaps every 100 steps (200 fs), and
> protonation state changes every 5 steps ( 10 fs) in pHREMD and every 100
> steps (200 fs) for the explcit. I am looking to speed up the simulation, if
> possible
​Attempting exchanges less frequently will probably improve things a little
in explicit. But the more frequently you exchange, the better your
sampling will be. Attempting replica exchanges every 200 fs is probably a
decent compromise.
For implicit solvent CpHMD, the more residues you have, the more often you
need to try to exchange (since each protonation state change attempt
involves one or two neighboring residues only). For explicit solvent,
replica exchange attempts become more expensive because it attempts *each*
residue in random order. You might be able to increase this to 200 as well
to speed up the simulation a little bit, but it will come at a cost of
reduced protonation state sampling.
The choices for the ntcnstph parameter differ greatly between implicit and
explicit solvent due to the way that proton exchange attempts are handled.
nstlim vs. numexchg behaves similarly in both situations. Outside of
possibly increased simulation efficiency (more pronounced for explicit
solvent, but probably minimal for both), there is no benefit to using a
larger value of nstlim.
HTH,
Jason
--
Jason M. Swails
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jan 16 2017 - 18:00:02 PST
