[phenixbb] on Rwork and Rfree

[Pavel Afonine]

Hi Dialing,

for example, it may mean the following:
we typically refine macromolecular structures using maximum-likelihood 
(ML) target function and not the R-factor. Maximizing the ML target does 
not necessarily mean minimizing the R-factor (although free R is 
correlated with the phase error and the latter should improve with the 
optimization of ML). So what you observe is slightly odd but not too 
unexpected.

This is discussed to some extent here (for example):
http://www.ccp4.ac.uk/newsletters/newsletter37/08_ml.html

The other 10+ potential reasons include:
- suboptimal refinement strategy;
- target weights may need adjustment;
- ...

Pavel

> Dear All,
>
> What it means if the Rwork or Rfree increase slightly during refinement?
>
> Best regards.
>
> Dialing
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://phenix-online.org/pipermail/phenixbb/attachments/20120126/d8b40490/attachment.htm>