Assessing quality of high-res refinement?

[Darren Nickerson]

In article <Pine.SGI.3.91.980107164808.4295A-100000 at kali>,
Flip Hoedemaeker  <fhoedem at oci.utoronto.ca> wrote:
>On Wed, 7 Jan 1998, Darren Nickerson wrote:
>
>> 
>> Greetings,
>> 
>> I am looking for pointers towards accepted practice in assessing the
>> quality of models refined against high-resolution (1.35A) data, in
>> particular using SHELX97. Has concensus yet emerged from within the
>> high-res enthusiasts as to what constitutes an 'acceptable' distribution
>> of omega angles, for instance?
>> 
>> Thanks in advance for any info, by e-mail or in this group.
>> 
>
>Procheck, which is part of CCP4, will assess the quality of your model as 
>a function of resolution. Ramachandran plot, peptide bond planarity, bad 
>contacts, tetrahedral distortion and hydrogen bond energies will be 
>plotted against an accepted bandwidth ranging from 1.0 to 4.0 Ang resolution.

Of course, thanks ;-) But is PROCHECK really suitable for data approaching
atomic resolution? In broad terms it is, of course, but isn't it based on
a library of structures of "at least 2.0 Angstroms"? Perhaps it includes its
own bias which may not be appropriate to really high-res work?

-Darren