*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Le 11 janv. 07 à 17:30, Mischa Machius a écrit :
On Jan 11, 2007, at 10:05 AM, Marc SCHILTZ wrote:
I think this is clear to all of us. But just as as we should
refrain from basing our models (or certain parts of them)
exclusively on diffraction data, we should also abstain from
basing them (or certain parts of them) exclusively on information
about what we expect the structure to look like. I believe that we
should be conservative and only include features in our models
that are both chemically sensible (i.e. reflect our prior
knowledge about protein structures) AND supported by the
diffraction data.
As is obvious from the discussion, there are those who believe that
coming up with a useful description for disordered side chain atoms
is fully compatible with this notion, because the diffraction data
reveal where C-alpha and C-beta are located, and those who believe
that this is not enough information.
Still one might consider that high B-factor (or more precisely ADP)
are NOT a sensible way to parametrise the kind of disorder of the
last atoms of a side chain (eg. end of a Lys side chain). ADP
consider that an atom's position is distributed with a 3D Gaussian
probability density (somehow ellipsoid in shape) around a centroid
point. In the case of 'disordered' side chains, it is more likely
that the disorder is coming from torsion around the previous bound,
producing a probability density in a shape of a torus (or part of it
as certain angle might be preferred).
If you agree with this reason of 'missing' density on terminal atoms
of a side chain, then you can not model the disroeder by hight B
factors, and you NEITHER are entitled to use ADP restraints during
refinement for these atoms....
You'd much better remove the atoms altogether or put a few
conformations (but then how to choose them, and choose there
respective occupancies and ADP)? Finally in the multiple
conformations case you are likely to see somehow anisotropic ADP
(large movement allowed in the direction compatible with torsion
movements), maybe this can be done by assigning some kind to TLS to
the side chain (TL might be enough in this case)?
In short, given the current used parametrisation (the one which are
currently available in refinement software), you should better remove
the atoms...
Serge.
Looks like there won't be a consensus.
Obviously ;-)
Maybe we should start to really start working on how to model those
side chains...
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.3 (Darwin)
iD8DBQFFpnCo5EPeG5y7WPsRArL9AJ9ZWsr4DO1g7KBykDQ71SwG2F/3cQCg50Sd
62GTHf+0IcrMDuuWR1nGuGk=
=0+LP
-----END PGP SIGNATURE-----