[phenixbb] question on the weight for adp

[Ed Pozharski]

Technically, it will affect validity of the model.  Let's say you adjust
weights so that geometry is weighted down to essentially zero - you are
very likely to get invalid model (unless it's ultrahigh res and you
don't need restraints anyway).  On the opposite, if you use restraints
that are extremely tight, you will tilt the balance towards idealized
model which will be not the best you can get from the data.

The question of correct weights is a tricky one, but the good ole
approach of minimizing Rfree (or maximizing log-likelihood) while
keeping the rmsd_bonds within 0.02ish will not lead you astray.

On Thu, 2010-09-30 at 17:49 -0500, fn1 at rice.edu wrote:
> Hi Kendall,
> 
> Thanks! Does this mean that I can always adjust the weights in order  
> to get beta map/stats? And whatever weights used will not affect the  
> validity of the model?
> 
> Fengyun
> 
> 
> Quoting Kendall Nettles <knettles at scripps.edu>:
> 
> > Hi Fengyun,
> > I would add to Pavel's comment that the answer is empirical, and I  
> > would first try the optimize weights option, or if you really want  
> > to have one set weight for all cycles then set up half a dozen jobs  
> > in parallel and see which gives best maps/stats. I would bet the  
> > optimize weights on each cycle would likely perform best, though.
> > Kendall
> >
> > On Sep 30, 2010, at 6:36 PM, "Pavel Afonine" <pafonine at lbl.gov> wrote:
> >
> >>  Hi Fengyun,
> >>
> >>> My case is that some part of the model will be refined to large B
> >>> values (about 100) if the default value of wxu_scale (1.0) is used.
> >>> The other part of the model will have average B value of about 40,
> >>> which the wilson B is about 35. The corresponding 2FOFC map has very
> >>> poor density for the region with large B values.
> >>
> >> "large" B-factors for atoms in poor density are expected. Also, the
> >> large B-factors is not always something wrong.
> >>
> >>> If the value of wxu_scale set to be 0.1, that part with large B will
> >>> be refined to about 80 for B. The corresponding 2FOFC map has some
> >>> better (but not good enough) density for this region.
> >>>
> >>> My question is that can I keep the wxu_scale at the small value for
> >>> the refinement?
> >>
> >> If it lowers B-factors, improves the R-factors and maps, then yes, keep
> >> whatever wxu_scale that does it. wxu_scale is just a number that
> >> balances target contributions arising from X-ray data and restrains and
> >> its absolute value doesn't have any particular meaning.
> >>
> >> Pavel.
> >>
> >> _______________________________________________
> >> phenixbb mailing list
> >> phenixbb at phenix-online.org
> >> http://phenix-online.org/mailman/listinfo/phenixbb
> > _______________________________________________
> > phenixbb mailing list
> > phenixbb at phenix-online.org
> > http://phenix-online.org/mailman/listinfo/phenixbb
> >
> >
> 
> 
> 
> _______________________________________________
> phenixbb mailing list
> phenixbb at phenix-online.org
> http://phenix-online.org/mailman/listinfo/phenixbb

-- 
Edwin Pozharski, PhD, Assistant Professor
University of Maryland, Baltimore
----------------------------------------------
When the Way is forgotten duty and justice appear;
Then knowledge and wisdom are born along with hypocrisy.
When harmonious relationships dissolve then respect and devotion arise;
When a nation falls to chaos then loyalty and patriotism are born.
------------------------------   / Lao Tse /