[phenixbb] DEN refinement

[Jeff Headd]

Hi Morten,

On Fri, Nov 30, 2012 at 6:50 AM, Morten Groftehauge <
mortengroftehauge.work at gmail.com> wrote:

> Hi guys,
>
> Two quick questions about DEN refinement:
> 1) If gamma refines to 0 does that mean that I should run DEN refinement
> again with the output model? Or should I do a refinement and then another
> DEN? Since gamma = 0 the refinement should be allowed a possibility of
> moving the structure further towards the reference restraints in my mind.
>

A gamma value of 0 corresponds to a non-deformable network, meaning that
the model strictly follows the distances derived from the reference. It
could be that model should have the same basic conformation as the
reference, and that has then given the best result. It's only necessary to
try it again if you don't think the overall conformation of your model has
converged, or if you wanted to try some different settings. If you haven't
tried running with different random seeds, I would suggest trying that
first.


> 2) If gamma refines to 1 it means that my reference model is
> inappropriate, right? Not really much of a question, I know.
>

 Correct, this means that the diffraction data that you are refining
against drive your model to a conformation that differs from that of the
reference model, and that conformation is preferred. I would also suggest
trying with different random seeds, just to see if you arrive at the same
answer.

Jeff




> And I know that DEN refinement is in beta but it would be nice if you
> outputted the optimal gamma and weight for the refinement to the GUI.
> Actually it would be nice if the did that for geometry weights and xray
> weights too when one optimizes them.
>
> Cheers,
> Morten
>
> --
> Morten K Grøftehauge, PhD
> Pohl Group
> Durham University
>
>
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://phenix-online.org/mailman/listinfo/phenixbb
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20121130/183dbae2/attachment.htm>