it sounds like this is a fairly ambitious application of NEB- not that it
won't work, but we don't really have a lot of experience with the kind of
things you're asking. I think you need to monitor the path (using some
geometric variables that you choose based perhaps on visual analysis of the
structure change) and see if it is varying during the annealing. You want
conditions that will allow sampling of alternate paths. Once again, this is
not an easy problem and NEB only provides the tool, you still need to do
things like check convergence and so on. Along those lines, I think that
generating paths with just the endpoints and again with some specific
intermediates can be a good way to see if the results are truly independent
of your initial guess at the path.
We apply NEB forces to everything that looks different between the
endpoints- but it tends to be easier to make sure the path is continuous in
all variables if you apply the springs to everything but solvent (and ions,
etc).
good luck!
carlos
On Wed, May 25, 2011 at 10:49 AM, patrick wintrode <pat_wde2.yahoo.com>wrote:
> Hi.
>
> We are getting ready use the NEB method for a large conformational change
> in a large system. We have completed the NEB tutorial and prepared the
> starting point/endpoint input files. I have few questions.
>
> Our total system size protein + solvent is ~45,000 atoms and our protein is
> 373 residues. The conformational change is complex and involves, among other
> things, the separation of 2 beta strands and the movement of a loop across
> nearly 60 angstroms. The tutorial was obviously optimized for a much smaller
> system.
>
> 1. What's a good estimate for how long we should equilibrate at each temp.
> during the simulated annealing? Can we use the annealing schedule in the
> tutorial as a guide, or do we need to try something totally different for a
> system this size?
>
> 2. Is it recommended to use apply NEB forces to the entire protein?
>
> 3. We were thinking of starting with half the replicas at the starting
> point and half at the endpoint and letting the NEB method find the path.
> Might it be better to generate a candidate path with targeted MD first?
>
> Thanks.
>
> Patrick L. Wintrode
> Assistant Professor
> Department of Physiology & Biophysics
> Case Western Reserve University
> Cleveland, OH 44106
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed May 25 2011 - 09:00:02 PDT
