Hello Everyone
This isn't really a question, just a warning to check your
anisotropic temperature factors both after refinement and after
deposition at the pdb.
We've been having a lot of trouble with anisotropic temperature
factors lately. Here's what's been going on.
We have three different proteins, each diffracting better than 1.2 A
- the best resolution is ~ 0.8 A. All have been refined with refmac
(yes I know shelxl is great for such problems, but that's not the
issue here).
For structures refined with some versions of refmac (including the
ccp4-6.0.2 linux distribution from early this year), the anisotropic
components output on the ANISOU card are ridiculous. When we run
anisoanl, ALL atoms are flagged as non-positive definite. When I ask
coot to show anisotropic ellipsoids, they have only two dimensions
(coot doesn't crash, Paul has the negative numbers trapped).
Refinement of the same data set/coordinates with shelxl produces
anisotropic temperature factors which are reasonable as judged by
these two methods. I can "fix" the refmac "problem" by installing
the latest version of refmac from Garib's web page (not the version
in ccp4-6.0.2). The anisotropic ellipsoids are now good (as judged
by anisoanl and coot). (Repeating myself, these are all very high
resolution structures).
So, I thought the problem was solved.
We reviewed structures we'd deposited with the pdb but were as yet
unreleased and found one with unrealistic thermal ellipsoids, which
confused us since the depositor had done the sanity checks before
depositing the coordinates.
When we compared the file we submitted to the pdb and the file which
was returned to us, the numbers on ALL the ANISOU cards had been
rotated in this fashion:
ANISOU 1 N MET A 1 1612 1818 1492 2 -160
-14 A N
has become
ANISOU 1 N MET A 1 1612 2 -14 1818 1492
-160 N
In other words the aniso card, (u11,u22,u33,u12,u13,u23)
has been rearranged to be (u11,u12,u23,u22,u33,u13) (this doesn't
make any sense to me)
I can speculate that the refmac/pdb issues are interlinked and one
arises from an attempt to fix the other, but I don't know that that's
true. I did a quick check of structures released by the pdb in the
last year with resolution > 1.1 A and found at least one with the
same problems (I stopped looking after I'd found one).
Sue
Sue Roberts
Biochemistry & Biopphysics
University of Arizona
[EMAIL PROTECTED]
