so far there isn't a single recipe for
getting good results from replica exchange.
it should be considered an advanced technique
that will require pretty extensive reading of
the literature to use properly.
there are articles out there that discuss
temperature selectino and exchange efficiency.
if you ahven't already, you should track them
down and read them. start out by trying to
reproduce what others have published, and work
with a small system until you understand it.
I can't make any specific suggestions since you
don't give us any details about what is happening.
how big is the system, is it GB or explicit,
why are you restarting, how did you choose the
temperature range, what is your T coupling constant,
how frequently you do exchanges, what the intial
structures are, how long the runs that you are
comparing are (how many exchanges) and so on.
All of these and more affect the results, so as
I said it should be considered an advanced technique
to be used only after you are very comfortable with
standard MD simulations.
If you use irest=1, tempi is ignored so I don't
think that it is likely that the value of tempi
is affecting your exchange acceptance ratio.
carlos
Scott E. Boesch wrote:
> Dear all:
>
> I run some replica exchange method (REM) simulations by using AMBER8.
>
> In the mdin.000 file, I change irest = 0 to irest = 1 and ntx = 1 to
> ntx = 5. And the program reads
> the restart file instead of inpcrd. I can restart REM simulations.
> Unfortunately, the acceptance
> ratio of exchange temperature decreases largely in the following
> simulations. I am wondering whether
> this problem can be solved by changing other input parameters.
>
> It seems a difficult job to obtain a good, even acceptance ratio (>10%)
> in REM. My simulations results indicate that they are largely dependent
> upon the initial temperatures (TEMPI), time steps between exchange
> attempts (NSTLIM), not only exchange temperatures (TEMP). However, even
> if I set some initial conditions to make a good REM simulations, the
> above restart problem make me feel discouraged.
>
> Thanks a lot
> Best
> Zunnan Huang
> znhuang.chemdept.chem.ou.edu
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Sat Apr 16 2005 - 13:53:00 PDT
