Hi Melania,

I have a bit of a problem applying NCS to my fairly low resolution structure. If I try to look for NCS symmetry automatically - phenix doesn't find any. If I specify sections that I want to be related by NCS, the script usually crashes.

What is the error message?

I have a dimer in the asymmetric unit, however it appears that one domain is related by 3-fold symmetry while the other is related by 6 fold symmetry. I can "cheat" phenix and generate operators for each of the domains by feeding a stripped pdb file (one that contains only a dimer of one domain or the other) to phenix.simple_ncs_from_pdb and then it easily identifies the ncs operators... unfortunately the simple_ncs_from_pdb.ncs file generated that is meant to be fed into phenix.refine contains only the chain selections, not the actual operators.

The operators are recomputed each time the NCS is used, based on the current coordinates. There is not way to enter the operators externally.

Is there a way to specify different NCS operators (rotational matrix and translational vector) for specific chain selections in the .eff file or under GUI? or if there isn't, how can I still take advantage of my NCS symmetry?

Also, is there any obvious reason why NCS operators that I generated with Phenix do not match the operators that I found using CCP4 Lsqkab? (yes I already accounted for chain A moving onto B or chain B moving onto A)

Not sure.

Can I choose to apply NCS symmetry to just my backbone? How would I specify that?

For example:

selection = "chain A and backbone"

You may want to try Jeff's new torsion angle NCS, which often gives better results than the Cartesian NCS and is very easy to use. I hope Jeff will chime in.

Ralf

P.S.: You probably figured this out already: you want to set ncs.find_automatically=False if you give you own selections for Cartesian NCS.