[phenixbb] NCS restraints

[Pavel Afonine]

Hi Nat,

> On Thu, Jun 30, 2011 at 9:07 PM, Yuri<yuri.pompeu at ufl.edu>  wrote:
>> I have some fundamental questions about NCS:
>> 1- I think its somewhat of a consensus that at higher res. it should not be
>> used to avoid missing out "real" detailed discrepancy. Is this true?
> Usually, yes.  There is some risk in the use of NCS restraints,
> because (depending on how the NCS relates to crystallographic
> symmetry) it can bias R-free due to symmetry inherent in the data.  At
> moderate resolution (maybe 2.0A) some people will start with NCS
> restraints at the early stages of refinement, and release them later.
> As always, it's difficult to give an exact rule of thumb, but I think
> Pavel has cited 2.0A as the cutoff before, and that's pretty
> reasonable.

this is based on my ~4 or 5 years old tests, when I re-refined all 
structures in PDB that have NCS with and without applying NCS 
restraints, and very roughly at resolution about 2A or higher I saw more 
cases where using NCS was doing more harms than good (judged by Rfree, 
and Rfree-Rwork only). However, with improved  NCS restrains (applied in 
torsion angle space) this limit may be pushed farther towards higher 
resolution. We need to re-run these tests over the whole PDB again to 
see what's the current state of the art...

Pavel.