On Mon, Apr 27, 2015 at 8:30 AM, lipengfei <lipengfei_mail.126.com> wrote:
> Dear all,
> When I run a T-REMD simulation using AMBER14, I get a file called
> "rem.log". But I don't understand how is the "success rate (i,i+1)" in the
> file "rem.log" calculated.
> After I refer to the p370 of the AMBER14 Reference Manual, it gives
> "
>
> Success rates are calculated as # successes/#tries × 2, where the factor
> of 2 is used
> because each pair of neighboring replicas attempts to exchange every other
> exchange attempt.
> In the beginning of the log file, this may lead to unusual success
> rates,but after a large number of exchange attempts the values will
> normalize.
>
> "
> What do the "# successes" and "#tries" mean respectively?
# successes means the number of time a replica exchange *attempt*
succeeded between replica i and i+1. # tries means the number of replica
exchange attempts that have been made. However, replica exchanges occur
between alternating pairs of replicas (for instance, the second replica
will first try to exchange temperatures with the first replica, then it
will try to exchange with the third replica). As a result, the
#successes/#tries will be off by a factor of 2, since replicas i and i+1
will only attempt to exchange 1/2 of the total exchange attempts (the other
half being between i and i-1
> How can I calculate the "# successes" and "#tries" in the each exchange ?
>
#tries just increases by one each time. You should see where it says
"exchange number" in the rem.log file. This is #ntries. As for the
successes, you have to keep track of each time an exchange attempt succeeds
(i.e., when the scaling factor is NOT -1), and add 1 each time the exchange
attempt succeeds.
However, you should be able to calculate the # of successes since you know
the success rate and the total number of tries (so use the formula in the
manual to solve for the # successes.
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Apr 27 2015 - 07:00:02 PDT
