Two points:
1. B-factors tend to differ lots between NCS copies, so you want to set
those restraints rather low (at least, I always do, by default)
2. NCS groups tend to need a far more fine-grained description than just
plonking in the whole domain. For structures in my lab, we often see
that tight NCS "does not work" -- until the group definitions are
selected more carefully.
Cheers
phx
Nicholas Keep wrote:
I am refining a low (3A) resolution structure of a 3 domain protein.
There are 4 copies in the ASU. I have been applying tight NCS
restraints by domain in refmac and have pulled the weak MR solution
down to Rfree below 30 (just).
However my question is that in 2 of the 4 copies one of the domains is
very poorly resolved. I can lower Rfree by around 0.5% by omitting
the domains from the PDB entirely or not applying the NCS restraints
to these copies of the domain. Clearly they are there and should
resemble the moderately well resolved copies by coordinates but the
way Bfactor restraints are applied between NCS copies seems to be the
issue. If tight restraints are included the B factors are much lower
(30-40) rather than 60-80 for the poor domains.
I was wondering if there is a theoretically correct way to treat this?
Would applying TLS scaling to each domain lead to the residual B
factors being more balanced?
Can a B factor offset be applied to the NCS restraints or could I only
apply a coordinate restraint not a B factor restraint between certain
copies?
Comments welcomed especially from Garib.
Happy New Year
Nick
