I've done a fair bit of REM now and how often you choose to apply the
test is one that has been passed over many times. If you test more
frequently (fewer steps between tests), your replicas will be more
mobile as they are more likely to change temperature (remembering that
the test is semi-random). More frequent tests therefore mean replicas
can be spaced further apart, for the same level of mobility. This saves
on computational expense.
However, if you test too frequently then the temperature of the system
is not equilibrated and the question must be asked about the driving
force that the thermostat provides.
Personally I have not yet used the AMBER code, but space my replicas
according to how quickly I believe the thermostat method equilibrates to
the new desired temperature. Of course this can take a long time, so a
pinch of salt is required! Even if the velocities are rescaled every
step, thus keeping the temperature perfect, the potential energies take
time to match that of the new temperature. I would suggest exploring the
thermostat method and watch the equilibration of your potential energies
after a temperature jump similar to the distance between your replicas.
Using a tight thermostat, I test every 2ps.
Hope this is of some small help!
Adrian
>
> Dear Amber users,
>
> When using replica exchange MD in amber8, I am confused at what the appropriate
> value of NSTLIM is, i.e. the parameter that controls the exchange frequency.
>
> Several published values of the exchange frequencies can be found, such as:
> 1) 46-residue three-helix 0.375ps pnas(2003)
> 2) 12-residue beta-hairpin 5 ps jmb(2004)
> 3) 5-residue peptide 0.01ps Sugita (1999)
> 4) amber user guide 100-1000 MD steps
>
> So, it can be seen that the value varies in a broad range. Any one could give
> some suggestions about how to choose it? Any disadvantages of too small and
> too large values?
>
>
> Best Regards,
>
> ---
> J. Zhang, Dr
> Institute of Biophysics
> Nanjing University
>
>
>
>
>
>
>
> -----------------------------------------------------------------------
> 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 Wed Jun 02 2004 - 10:53:00 PDT
