Sounds like a pretty successful refinement, given the resolution!
The fact that releasing NCS improves R-free suggests there are
real ncs violations (otherwise releasing ncs increases R-free).
But they may be confined to a few residues in contact areas.
You can locate these violations by comparing the structures
before and after release to see what moved, or by superimposing
the different monomers after release and finding outlier residues.
Then release NCS restraints only in these limited areas, keeping
the advantages of NCS wherever it is valid.
Paul Paukstelis wrote:
Hi all,
I'm refining the structure of a complex at low resolution (4.5).
Certainly refinement at low resolution will become more common, but
there isn't a whole lot out there now to use as a guide. I've
incorporated most of the suggestions from DeLaBarre and Brunger, but I'm
looking for any other suggestions and I have a couple of specific
questions.
I have independent, high resolution structures for both molecules of the
complex, and I got a nice solution from Phaser. There is 4-fold NCS
(45,000 atoms in the ASU). Because of the low resolution I have also
added manual restraints based on secondary structure of the individual
structures to help the observation/parameter ratio. So far, I have been
using tight NCS restraints in Refmac along with Babinet scaling for bulk
solvent, fixed B for the solvent, and isotropic refinement for B
factors. Currently, R=25.7 Rfree=30.0 and the difference between R and
Rfree has stayed pretty constant from starting values in the 40's. I
just loosened up the NCS and got a fairly significant drop in R
(21.3)but a less significant drop in Rfree (28.8). The maps seem
somewhat better with some possible ion sites getting stronger in the
difference map. Should I worry about the large R change relative to Rfree?
I also have some very strong density for a portion of a molecule
(missing in the original structure) that I can't model unambiguously at
this resolution. It is very likely in multiple conformations. I'm
guessing this is going to cause the R factors to bottom out and I'm
worried trying to push the R factors down without building this will
lead to a some distortion around this area. Any suggestions on how to
handle this?
Finally, what would you like to see in a paper as validation for the
correctness of a model at this resolution and phased by MR? Omit maps of
key features?
Yes, I am trying to get higher resolution data.
Thanks in advance.
--paul
