Hi,
Double check that you've really got the energies from MDOUT sorted by
temperature; as Koushik said that really does look like replica space.
Another possibility is that your temperature spacing is too small, but
you'd still expect a more Gaussian shape if that were the case. One
thing you could do to check would be to sort the trajectories
themselves (with 'ensemble') and then get energies via post-processing
the sorted trajectories (using either sander imin=5 or cpptraj
'energy' etc).
Sorting generic replica datasets is on my to-do list for cpptraj;
actually cpptraj will sort constant pH replica data now, just needs to
be extended to other data sets.
-Dan
On Wed, Aug 5, 2020 at 8:35 AM Diego Morone <diego.morone.irb.usi.ch> wrote:
>
> Dear AMBER users,
>
>
>
> I am trying to set up replica exchange on AMBER. Using explicit solvent
> (ff14SB + TIP3P water) and 16 geometrically spaced replicas I get a good
> range of accept ratios (from 5 to 16 %) and 10-20 round trips/replica, but
> weird histograms (see attachment) which don't look like gaussians. The
> simulation is run for 40ns/replica on 2 GPUs with pmemd.cuda.MPI on Amber
> 19. After heating at 300K, the equilibration was performed first NPT at 300K
> 1atm, then NVT at 300K. I get a similar plot with implicit solvent and less
> replicas. Could you help me in spotting the problem?
>
>
>
> Many thanks
>
> Diego
>
>
>
> TREMD
>
> &cntrl
>
> imin = 0,
>
> irest = 0,
>
> ntx = 1,
>
> ntt = 3,
>
> gamma_ln = 1.0,
>
> temp0 = TEMPERATURE,
>
> cut = 9.0,
>
> ntc = 2,
>
> ntf = 2,
>
> nstlim = 10,
>
> dt = 0.002,
>
> ig = RANDOMNUM,
>
> numexchg = 2000000,
>
> iwrap = 1,
>
> ntpr = 5000,
>
> ntwr = 5000,
>
> /
>
> &wt TYPE='END'
>
> /
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Aug 05 2020 - 06:30:03 PDT
