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


On Friday 19 May 2006 10:27 am, George M. Sheldrick wrote:
> Possibly some future version of a refinement program will apply TLS
> restraints rather than constraints, this could be programmed in a way
> that would refine the Uij values directly and so avoid this problem.

We tried this.  It was very easy to add an additional input file
to Refmac, one that contains target values of the anisotropic ADPS
to be used as restraints. We then spend 2 years working out how
best to generate those restraints. What we ended up with was the
TLSMD algorithm.

The surprise was that using the multi-group TLS model as a set
of restraints to an anisotropic refinement was no better than
using it for TLS + isotropic refinement.  That is to say, the
R factors achieved were essentially the same for either refinement
mode.

However, the TLS+Biso model contains many fewer parameters than
the TLS-restrained Aniso model.  Thus is is statistically preferable.

Now it may well be that if you have true atomic resolution data
then the fully-aniso refined model, restrained or otherwise, will
deviate significantly from the TLSMD+Biso model.  But in this case
I think the existing SIMU/DELU/ISOR (shelxl) or SPHE/RFAC/RBON (refmac)
restraints are sufficient.


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

Reply via email to