On Thu, Nov 03, 2011, Brian Radak wrote:
>
> Might I burden you with an additional question? I was under the
> (apparently incorrect) impression that SHAKE/SETTLE is not valid when doing
> a minimization (although I don't really know what is meant by that). Or at
> least this is apparently the case with CHARMM (or so the documentation
> says). I would suppose such an incompatibility would depend on the
> minimization algorithm and the implementation of the constraint method
> (projecting out forces vs resetting positions or something like that
> right?). What clever way does AMBER use to avoid these troubles (although
> I don't know specifically what they are)?
There is indeed a sense in which SHAKE is not compatible with the minimization
routines in sander and pmemd: the energy surface the minimizer "sees" does not
include SHAKE (which is a correction to dynamics). So the minimizer will get
"stuck" (giving LINMIN failures if you use the old-fahsioned conjugate
gradient routines). But if you are just using minimization to eliminate some
bad contacts prior to MD, things are usually fine.
You *can* do correct minimization with constraints in NAB/sff. Of course the
catch-22 is that NAB/sff doesn't know about periodic boundary conditions.
Most Amber developers can live with the limitations of our minimization
routines, and hence have never really had the drive to put a SHAKE-aware
minimizer into sander or pmemd.
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Nov 04 2011 - 07:30:02 PDT
