*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
On Sep 12 2005, Ashley Buckle wrote:
I have 3 questions:
1. What are the guidelines for the amount of superimposed models one
should use in an ensemble in PHASER?
We've found improvements with up to 5 or 10 models, but not much from
adding more. If you're using different homologues (as opposed to members of
an NMR ensemble), it's better not to add models that are significantly
lower in homology than the best model. For example, if there were a model
from a protein sharing 35% sequence identity, I might include other models
down to 30% sequence identity, but not models at 25% sequence identity.
2. If I have say 5 aligned models in an ensemble, is there any
advantage in PHASER to leaving the regions of poor alignment in the
ensemble? (eg lets say my aligned models look like a very bad NMR
structure - should I trim the bad regions, or does Phaser weight them
down sufficiently anyway?)
The averaging will tend to weight them down, but in order to be useful the
average density has to be in about the right place. If a loop runs through
the same volume, but in different conformations in different models, it's
probably useful to leave it. If it's all over the place, it's probably
better to trim it.
3. Using CCP4i, is it possible to ask phaser to search for, say, 50
different ensembles, one after the other? (eg do a complete
'automated search' phaser run on each successive ensemble. I't looks
like this is possible by selecting 'Add another search' under 'search
details' but I'm not sure....
Yes, but you have to be careful to distinguish between searching for
several molecules in the asymmetric unit (which is what "Add another
search" allows) and using alternative models in the search for one
molecule.
First, turn on the "Allow search with alternative models..." option. Then
you're given an opportunity to add alternative models with the "or
ensemble" pulldown. More alternatives can be added by using the "Add
alternative ensemble to test" button.
If you're going to do this for 50 possible models, this will get a bit
tedious, and it's probably better to use a script (perhaps starting from
one you get using the "Run&View Com File" option).
Randy Read
