[phenixbb] Real space implementation of DEN refinement?

[Pavel Afonine]

Hi Oliver,

> Just wondering whether there is any capacity to include DEN restraints 
> in phenix.real_space_refine, e.g. as implemented in the program DireX 
> (see ref below), or any possibility of implementing such in a future 
> release of PHENIX?

do you have a specific example of a model *before* and *after* running 
through DireX that exemplifies the benefits? I would be very interested 
to see it (if you have one and choose to send me, please send it to my 
email address directly, not the whole phenixbb).

phenix.real_space_refine has a number of options that allow to achieve 
very large convergence radius, such as model morphing, Simulated 
Annealing refinement, analog of MZ protocol 
(http://www.phenix-online.org/papers/dd5048_reprint.pdf) for rigid-body 
refinement, etc. These options are not enabled by default and so may 
escape user's attention.

> Also, on a different but related topic, does anyone have any 
> suggestions as to how to handle model validation (local and global) at 
> low resolution in real-space? It seems as though there is not a solid 
> consensus on how to handle this.

phenix.real_space_refine outputs block of text containing Molprobity 
statistics, as well as model to data (map) fit such as map correlation. 
You can also run Comprehensive Validation from Phenix GUI, although at 
the moment it would not accept map as a source of data.

Pavel