[phenixbb] difference in estimated error between phenix and refmac

Pavel Afonine pafonine at lbl.gov
Thu Aug 12 11:49:20 PDT 2010

  Hi Ling,

I have no idea what the error estimate based on R means. In 
phenix.refine the ML based coordinate error estimate is done as 
described in formula #19 in

/Acta Cryst./ (2002). A*58*, 270-282

with the ML parameters determined as described here:

Acta Cryst. (1995). A51, 880-887.
J. Appl. Cryst. (1996). 29, 741-744.

This number is sensitive to many (irrelevant) thins such as bulk-solvent 
model, etc.. So I wouldn't take it too literally. Model to map fit, map 
quality, local and global, geometry stats and Rfactors are all much more 
useful model quality characteristics than this number.


On 8/12/10 10:04 AM, Ling Qin wrote:
> Dear Colleagues,
> I am refining a 2.15 A structure that was previously refined with 
> refmac by someone else. Without changing anything in the 
> refmac-refined model and using all the default phenix.refine settings, 
> I am getting pretty much the same R and Rfree (from 19.3%/21.7% to 
> 18.9%/21.8%), with an ML estimated error of 0.22 A. In the 
> refmac-refined pdb file, the estimated error is much lower: 0.16 
> (based on R), 0.144 (based on Rfree), and 0.096 (based on ML). I am 
> using phenix.refine version 1.6_289, but also tried 1.5_2 with the 
> same result. Is there an underlying reason for this difference? Thank 
> you very much for your insightful opinions in advance.
> Cheers,
> Ling
> _______________________________________________
> 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/20100812/aa49d142/attachment-0002.htm>

More information about the phenixbb mailing list