On 17 Jun 2008, at 19:56, Roger Rowlett wrote:
Sampath Natarajan wrote:
Dear All,
I am refining a structure with 2.5A resolution by
refmac5. I could find the solution by MR using molrep. After
fitting the model, I refined the structure again with 0.3
weighting term, but the output PDB file shows many splits in the
residues. So I used 'auto' as a weighting term, then I didn't find
any splits. At the same time I noticed that figure of merits is
not increasing after model fitting. But the correlation
coefficient is more than 80%. Also while refining with auto
weighting term, the figure of merit goes down about 25% than
previous.
Finally I find that some of the residues show high B
factor. Could anyone give suggestions to reduce the B-factor?
Thanks in advance!
Yours sincerely,
Sampath
For 2.5A data, the refmac default X-ray weighting factor of 0.3 is
typically much too high, and does not appropriately apply geometric
constraints. (That is why your proteins chains are breaking up.)
***Restraints*** really. Constraints are something really different
(for one things they imply that they cannot be violated but must be
exactly reproduced, so a weighting term for a constraint does not
exists). I think we must avoid confusion, since implementation with
bond/angle constraints (torsional refinement) are available.
An alternative to the auto weighting function is to set your own
weighting factor and monitor rmsBOND and rmsANGLE deviations in the
final model. If the final rmsBOND is on the order of 0.015 A and
rmsBOND is on the order of 1.5 degrees then your choice of
weighting factor is reasonable. For medium resolution data in the
2.2-2.8A range, I would start with an X-ray weighting factor of
about 0.05 or so.
Although that the general advice above is of course valid and it
reflects common practice, it would be fare to iterate once more that
this remained for a while a somewhat controversial issue. In my
opinion it is getting settled lately, given the RMSZ in more recent
versions of refmac which might be a better quantity to follow - as
suggested in the papers by Ian Tickle, eg:
Acta Crystallogr D Biol Crystallogr. 2007 Dec;63(Pt 12):1274-81;
author reply 1282-3. Epub 2007 Nov 16.
Experimental determination of optimal root-mean-square deviations of
macromolecular bond lengths and angles from their restrained ideal
values. Tickle IJ.
I claim no expertise and the math is above my head mostly, but I got
pretty much convinced by that work for what I want to be doing and
its good reading even for the mathematically challenged, like myself.
Tassos
Cheers,
--
----------------------------------------------------------------------
--
Roger S. Rowlett
Professor
Colgate University Presidential Scholar
Department of Chemistry
Colgate University
13 Oak Drive
Hamilton, NY 13346
tel: (315)-228-7245
ofc: (315)-228-7395
fax: (315)-228-7935
email: [EMAIL PROTECTED]
