Hi all,
I've been exploring the use of pure TLS models at various resolutions.
By "pure" I mean that there is no individual atomic Biso component for
the protein atoms. The net B factors for each atom are described only
by the TLS group it belongs to. I have had really good success with
this for refinement at low resolution, say 3-4 Angstroms.
More recently I have been comparing the model quality and R factors
from pure TLS and TLS+Biso refinement at resolutions between 2 and 3 A.
But I have run into a glitch in using refmac. What I want to do is
this:
1) refine structure with conventional Biso model
2) use TLSMD to generate a multi-group TLS model
3) reset protein B factors to 20 (tickbox in ccp4i interface to refmac)
4) refine TLS model with no individual Biso for the protein atoms,
but with the usual positional refinement
The problem is that at stage (3) refmac resets the B factors for all
atoms, not just the protein atoms. This wipes out the previous values
stored for water molecules and ligands. So at the end of (4), even if
the protein ADPs are well described by the pure TLS model, all the
waters and ligands are still stuck with B factors of 20.
Suggested fix:
Add a variant of the command "temp set 20.0" that applies only to
atoms that belong to some TLS group. Atoms that are not in any TLS
group should retain their current Biso value.
Ideally there would also be a way to refine the Biso values for
these non-TLS atoms. Perhaps using the "MIXED" keyword?
Or maybe I'm overlooking some existing way to achieve the same goal.
cheers,
Ethan
