Hi G,
Dear all,
I have several similar models which can be superimposed. (looks like NMR
solved structure) Then I made those superimposed models to be a single
ensemble in phaser.
My question is:
what would be the difference of running phaser with this kind of ensemble
and with an ensemble which include only one model? As I have B factor
information in PDB, the uncertainty in model has already been considered
even if I just provide one model. Is that right? Would that really be more
helpful to test an ensemble with multiple models superimposed?
Although I am not aware of any systematic study, there are indications that
an ensemble can be a better model than a single model with B-factor
weighting (check out the phaser MR tutorial with toxd). For one thing, an
ensemble will weight up the structurally conserved parts of the model,
while this is not guaranteed with B-factors (although both tends to weight
up the core and weight down the surface). Also, B-factors are normally
restricted to describe an isotropic uncertainty, but ensembles can describe
complex motions of loops, etc.
What would happen if those superimposed models are not quite similar, or
they can not superimposed well( some distance between them can be
observed)?
It is possibly a good idea to do a weighted superposition, so that only the
structurally similar parts are superposed (this is automatically done by
the SSM algorithm). If there are still major deviations, you can consider
trimming these residues away. This is something that can be done
automatically with ensembler (currently in PHENIX, but eventually to be
included in CCP4 as well).
BW, Gabor
Many thanks,
G