Hi Matthias,
a few suggestions:
- use TLSMD server to define TLS groups;
- before using TLSMD do some group ADP refinement with one or two
refinable ADP per residue (you can do it in phenix.refine), so you get
less biased ADPs from previous restrained B-factor refinement;
- just out of curiosity try phenix.refine: what if it works better in
this particular case? No worries if it doesn't but I would appreciate a
feedback.
- *personally*, I don't like the idea of having residual B-factors in
ATOM record; have a look at
http://www.phenix-online.org/presentations/neutron_japan_2009/phenix_japan_part1.pdf
to see what phenix.refine does in this case (Sorry I don't remember the
slide number, it's somewhere in a middle).
Good luck!
Pavel.
On 10/29/09 7:08 AM, Matthias Zebisch wrote:
Dear BB!
I have a strange problem with TLS refinement and refmac.
In my AU there are 4 monomers of the same protein.
Based on structural subdomains I defined 3 TLS groups for each of
those monomers so that the overall TLS group number is 12.
When I do TLS and restrained refinement from the GUI using my TLSIN
file which has only the group definitions,
everything seems fine. I only get quite a few atoms with residul B
factor with the strangely exact value of 2.00. Is this normal?
I had this also for waters but could overcome this with the keyword
"TLSD waters exclude".
R factors go down to 18%/22.5% (mild anisiotropic diffraction to 1.8A
in the best direction).
Now, my real problem is, that when I continue with model building and
refinement (restrained with fixed TLS or TLS+restrained),
I would like to use the original TLSOUT as the new TLSIN file. When I
do so, the R factors rather go up ending with values higher
then when leaving out TLS completely (~25%/30%).
Can someone explain to me why this happens? I would rather like to
avoid rerunning TLS everytime again as it is very time consuming.
Any help is greatly appericiated.
Matthias
PS: I have this issue also for other proteins.
