Hi Jan, thanks for the questions.
- How does phenix.refine use ncs restraints?
It is harmonic restraints between NCS copies, applied to coordinates and individual isotropic ADPs; similar to CNS.
- Can one define the tightness of the ncs restraints?
Yes, indeed, it is missing in the manual. I will add it asap. Use : refinement.ncs { coordinate_sigma=None b_factor_weight=None } } to set up the NCS weights for all NCS groups. Or you can do it for each individual NCS group by setting same parameters inside of restraint_group{} scope. Note, you should be careful with changing NCS weights since NCS restraints added to all geometry restraints, so if you drastically change the NCS contribution you can easily distort the contributions from other restraints, like covalent bonds, angles ... For an extreme example, setting up huge NCS weight will result in perfect match of NCS groups (= small rmsd between them) with totally distorter stereochemistry.
Can one put different restraints on mainchain and sidechain atoms?
Currently no. Unless you split out your NCS group selections into main chain and side chain. The default for coordinate sigma is 0.05. If we do this split in the future what would be the default for main and side chain sigmas? Pavel.