***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***


On Wednesday 15 November 2006 07:16 am, Zhou, Tongqing (NIH/VRC) [E] wrote:
> 
> I am refining several structures (all the same except point mutations,
> 2.0-2.8 A resolution) with Refmac5 (TLS+ISOT). One data set goes to 2.0
> A and the refinement seems fine with 
> 
> REMARK   3   R VALUE            (WORKING SET) :  0.20075
> REMARK   3   FREE R VALUE                     :  0.23986
> REMARK   3   MEAN B VALUE      (OVERALL, A**2) :  12.944

But that "MEAN B VALUE" is really the "mean value of the Biso residual
after applying the TLS model", if I understand your described experiment.
So without inspecting the TLS model itself, we can't say whether it is
reasonable or not.

> The low B also worries me, 

It shouldn't worry you in general.  If the TLS model were perfect
(which of course it never is) then this number would be 0.
To find the "real" mean B value you need to average over the
values in the file output by TLSANL.

> so I ran TLSANL, the total B of some residues went negative. 

That was the correct test to make, and it sounds as if something
has indeed gone awry.

> I then submitted the output pdb to PARVATI server, it 
> came back horrible, essentially listing all atom with "Non positive
> definite ellipsoid" ..., the mean anisotropy is 0.15, all my other
> structures have mean anisotropy of ~0.45 and normal average B about
> 40-50. 

That is also a good way to check, and again indicates that your
refinement has gone bad.

> I am not sure what I did wrong. Can anybody  advise how to generally
> deal with this?

I suggest resetting all the Biso values to the Wilson B, and re-running
refmac with the same input TLS groups as before.  Sometimes starting the
refinement off at a different starting place is all that it takes to 
overcome a local minimum or other refinement artifact.  Usually it
behaves better if you start it off with the Biso values from TLSMD
analysis, but not always.   

You could also try setting the ISOR restraint higher, to prevent the
ellipsoids going non-positive definite. But I have to admit that I'm
not 100% certain how strongly refmac uses this restraint in the case
of TLS refinement.

-- 
Ethan A Merritt
Biomolecular Structure Center
University of Washington, Seattle 98195-7742

Reply via email to