AMBER Archive (2005)

Subject: Re: AMBER: langevin, GB and simulated annealing - 2

From: Andreas Svrcek-Seiler (svrci_at_tbi.univie.ac.at)
Date: Wed Apr 20 2005 - 03:03:56 CDT


Hi,
For SA runs using Langevin (or Berendsen), I'd suggest you'd
give NAB (http://www.scripps.edu/mb/case/casegr-sh-3.2.html) a try.
With NAB you have much more freedom for trying out annealing protocols or
whatever comes to your mind. You could (just for example) write out your
minmized structure after each annealing cycle to a pdb file and
the more accurate coordinates to a restart-file. Then you could
open the pdb file and append e.g. the final Eamber as a remark,
which is quite useful if you have 1000s of structures.

As for SA and Langevin in general:
For rapid heating, a large friction-constant is ok. However,
with a large friction-constant overall motion is "diffusion like"
and slowed down ( compared to NVE or a berendsen thermostat ).
So one might try alternating not only target-temperature,
but also gamma_ln as e.g.
TEMP1 - high friction - short
TEMP1 - low friction - longer
TEMP2 - high friction - short ("pull the brake")
TEMP2 - low friction - longer ("some equilibration/relaxation")
....
This is what I *think*, I never studied such protocols with respect to
their efficiency in detail.

good luck
Andreas

-- 

))))) ((((( ( O O ) -------oOOO--(_)--OOOo----------------------------------------------------- o Wolfgang Andreas Svrcek-Seiler o (godzilla) svrci_at_tbi.univie.ac.at .oooO Tel.:01-4277-52733 ( ) Oooo. -------\ (----( )-------------------------------------------------------- \_) ) / (_/

----------------------------------------------------------------------- The AMBER Mail Reflector To post, send mail to amber_at_scripps.edu To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu