-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
You already have excellent suggestions for how to proceed with your
refinement, but I wish to add some details which, hopefully, clarify
your situation.
The average B factor is a measure of how rapidly your diffraction
intensities fall off with resolution. Since you could only collect
data to 4 A resolution they must have been falling fast so your model
will (and must) have a very high average B factor. An average in the
region of 280 A^2 sounds find to me.
The problems with refining B factors at this resolution are
two-fold. First, the value of the average B factor becomes tightly
coupled to the average B factor of the bulk solvent model (Bsol). A
poor value for Bsol can cause your average B to differ greatly from
its proper value. The program documentation should guide you in
deciding if your Bsol is reasonable.
Second, with 4 A resolution data it cannot be decided if two atoms
closer than that distance have similar B factors or wildly differing
ones. Say you have two atoms 3 A apart. The 4 A X-ray data does say
that their average should be 280 A^2, but are the individual values
both 280 A^2 or is one equal to 1 A^2 and the other 560 A^2? The 4 A
data set is perfectly happy with either model and an uncontrolled
refinement will likely result in crazy and wildly changing B factors,
that when averaged together give the right average.
Refinement programs have either restraints to damp these local
fluctuations, or TLS and/or group B's to forbid them. These
restraints and constraints shouldn't affect the average B factor, only
the amount of variation about the mean.
A lower average B factor is in no way an indication of a more
correct model. A model based on 4 A resolution data must have a very
high average B.
Dale Tronrud
On 7/22/2015 7:06 AM, Appu kumar wrote:
> Dear All, I am facing troubles in refining structure (C2221 space
> group) at 4A resolution . The current Rwork/Rfree stalls at
> 35.8/37.7.Given low resolution, the total number of unique
> reflections accounts to *19356 while number of atoms in pdb
> corresponds to **3893. *At this very low resolution, the number of
> available observations are insufficient to support the refinement
> of four parameters (x, y Z, Biso) and therefore i think the
> modeled atomic displacement (B-factors) are excessively high
> ranging from 100-280 which is scary at this moment.
>
> Is there any illustrative parameters to work on which might be
> useful for putting control over the shooting up of B-factors during
> refinement. I tried incorporating the group B-factor refinement
> with TLS group but seems like is not much useful.
>
> I would appreciate any incoming suggestions and thoughts on
> adopting crucial refinement strategies pertaining to this
> resolution.
>
> Thank you Appu
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.22 (MingW32)
iEYEARECAAYFAlWv1zEACgkQU5C0gGfAG12GYwCePIOIGGjn8lBixmuouSSffqN8
iqQAnj+MEpe5z091ATk9qSVUQC4krjyU
=FKpb
-----END PGP SIGNATURE-----
|