On Thursday, 05 June, 2014 17:51:33 Carles Palanca i GarcĂa wrote:
>
> P { margin-bottom: 0.21cm; }
>
>
> Hello,
>
> I'm almost done refining a protein-DNA
> complex at 3A, but when I apply TLS (Refmac+TLSanl to include the
> ANISOU component) the B factors increase from 50-60 to 100-110 A2.
> The crystal is a bit special since it has 80% of solvent and the
> Wilson B-factor is 87 A2. When I compare my Bfactors with other
> structures of similar resolution using Phenix Validation, my data is
> clearly an outlier.
>
>
> To perform the TLS refinement I have
> generated .tlsin files for 2,4 and 6 groups with the TLSmd server and
> used them in refmac jobs consisting of 20 cycles of TLS refinement+50
> cycles of restrained refinement. This way the R-factor and R-free
> values have drop 1-2% (before it was 22.3/24.1). The three different
> options suffer from the same problem regarding high B-factors after
> TLSanl. Am I doing something wrong during the TLS refinement? What
> can I do to preserve the original Bfactors?
Preserving the original B factors should not be your goal.
I can explain part of what you are seeing, but probably not the
entire effect.
When you refine using TLS, the various programs convert this back
to an "equivalent isotropic B factor". Sometimes this is marked by
the programs as "Beq", but it is placed in your output PDB
file simple as "B" with no warning that it is only an approximation.
But not only is Beq an approximation, it is invariably an overestimate of
the isotropic B. Beq can run as much as 40% larger than
the "true" isotropic B. A better estimate Best is available but so far
as I know none of the standard programs use it. Math, charts, and
example demonstrations of the too-large Beq are reported in
"Some Beq are more equivalent than others". Acta Cryst. A67, 512-516.
E. A. Merritt (2011).
Ethan
>
>
> Thanks in advance,
>
> Carles Palanca.
>
>
--
Ethan A Merritt
Biomolecular Structure Center, K-428 Health Sciences Bldg
MS 357742, University of Washington, Seattle 98195-7742
|