[phenixbb] Refinement differences?

[Tim Gruene]

Hi Pavel,

it may be interesting to see percentage of outliers in the input model. If the 
input model contains a large percentage of outliers, phenix cannot do magic, 
and rev 2155 may be better than 2037 in the sense that it better balances 
between data and geometry. Hence, if it does not force rotamer outliers to be 
corrected against the data, there will be more outliers in the output. To me 
this notion is supported by the improved gap R-work vs. R-free in the later 
version.

This depends, though, on the input model.

Best,
Tim

On Wednesday, February 10, 2016 08:53:48 PM Pavel Afonine wrote:
> Hi Mark,
> 
> to me statistics from both refinement runs look essentially the same in
> a sense that if you run say 100 refinement jobs each one starting with
> slightly perturbed structure (shake in coordinates with rmsd~0.3-0.5A,
> shake in B-factor) you will get ensemble of essentially the same but
> still slightly different refined models. Variation in R-factors my be
> ~0.5-2% between models in the ensemble. Geometry may vary slightly as
> well. With this in mind I would say different numbers you get are within
> the spread and I doubt the differences are really significant.
> 
> The strangest thing to me is that you still get ~9% or rotamer outliers:
> 2155 is supposed to be much better in eliminating rotamer outliers
> (although it is still limited to without H included).
> 
> I agree with Eckhard: we made a lot of changed recently, such as re-do
> NCS how NCS works, using CDL (conformation dependent library for
> restraints), etc..
> 
> Of course there may be other reasons for differences between these two
> runs but this is all I can tell given amount of information provided!
> 
> All the best,
> Pavel
> 
> On 2/9/16 13:54, Mark A Saper wrote:
> > Hi,
> > I used Phenix, build 2037 to refine a a new structure.  I reran the
> > refinement (with the identical parameters) in build 2155 (v1.10.1) and
> > obtained different statistics and slightly worse geometry.  Any idea why?
> > 
> > In *2037*:
> >                  start         final
> >   
> >   ---------------------------------------
> >   R-work:           0.2488        0.2244
> >   R-free:           0.3003        0.2851
> >   RMS(angles):      1.67           0.80
> >   RMS(bonds):       0.020          0.004
> >   
> >                Ramachandran outliers: 1.2% (Goal: < 0.2%)
> >                
> >                   Ramachandran favored:  94.5% (Goal: > 98%)
> >                   
> >                       Rotamer outliers:   8.7% (Goal: 1%)
> >                       
> >                        C-beta outliers:   0    (Goal: 0)
> >                        
> >                             Clashscore:   3.14
> >                          
> >                          Overall score:   2.19
> > 
> > *2155*
> > 
> >                    start         final
> > 
> > ---------------------------------------
> > 
> >   R-work:           0.2488  0.2331
> >   R-free:           0.3003  0.2856
> >   RMS(angles):      1.77 0.72
> >   RMS(bonds):       0.020  0.004
> >   
> >                  Ramachandran outliers:   1.4% (Goal: < 0.2%)
> >                  
> >                   Ramachandran favored:  93.4% (Goal: > 98%)
> >                   
> >                       Rotamer outliers:   9.4% (Goal: 1%)
> >                       
> >                        C-beta outliers:   0    (Goal: 0)
> >                        
> >                             Clashscore:   4.10
> >                          
> >                          Overall score:   2.36
> > 
> > _________________________________
> > Mark A. Saper, Ph.D.
> > Associate Professor of Biological Chemistry, University of Michigan
> > Ann Arbor MI  48109-1055 U.S.A.
> > saper at umich.edu <mailto:saper at umich.edu>     phone (734) 764-3353
> > fax (734) 764-3323
> > 
> > 
> > 
> > 
> > 
> > _______________________________________________
> > phenixbb mailing list
> > phenixbb at phenix-online.org
> > http://phenix-online.org/mailman/listinfo/phenixbb
> > Unsubscribe: phenixbb-leave at phenix-online.org
-- 
--
Paul Scherrer Institut
Dr. Tim Gruene
- persoenlich -
OFLC/102
CH-5232 Villigen PSI
phone: +41 (0)56 310 5297

GPG Key ID = A46BEE1A

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 173 bytes
Desc: This is a digitally signed message part.
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20160211/75d8a327/attachment.sig>