Dear Israel,
I wonder why you do not see anything in your Fo-Fc (mFo-DFc?) maps,
since the limit for a (n+1) * mFo - n * DFc map, when n approaches
inifinity, is the mFo-DFc difference map. There is nothing wrong in
going further like with 4mFo-3DFc maps, but you should bear in mind that
they are becoming more and more difference map-like and less and less
regular maps.
What I think is happening is that since you added your ribosomal factor
to preformed crystals, you have incomplete occupancy, say the factor
bound only to 50% or less of the ribosomes. This means that you have to
go down in contour level from the 3 sigma which is the default for
difference maps to maybe 1.5 or even 1 sigma. If the factor is bound
indeed at lower level you will see relatively nice density, if nothing
is there, you will see only noise.
To prove a conformational change at a single residue, I would just leave
it out and refine. Since the refinement program has no reason to bias
either conformation, the result should be unbiased. However, in the case
of partial occupancy, the resulting map will be a superposition of both
conformations. Also to reduce model bias, you might want to try the
Buster program.
Best regards,
Herman
________________________________
From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On
Behalf Of Israel Sanchez
Sent: Thursday, October 04, 2012 9:17 PM
To: CCP4BB@JISCMAIL.AC.UK
Subject: [ccp4bb] On maps and doubts
Hello everyone,
I would like to share my experience with one dataset and request
some advice on which is the best way to prove a conformational change
seen in a density map.
The first issue arose when we were looking for an extra
ribosomal factor added to a crystalized ribosome. After careful data
collection and refinement (I/sigma last shell 1.2, 3.1A and CC1/2 around
22%) the sigma-A-weighted maps 2mFo-DFc and Fo-Fc does not show any
clear difference density that we could interpret as the expected factor.
Interestingly, a computed map with coefficients 3mFo-2DFc started to
show some features that clearly could be explained as a fragment of the
factor. The density improved even more with a B-sharpened map. We have
seen this behavior before and I was wondering if someone else is using
this kind of maps and may could explain the reason behind this density
improvement. Is it a crazy idea to go even higher like 4mFo-3DFc?
The second query has to do with which is the best way to prove
that a conformational change is present in an specific residue (in this
case and RNA base) in your structure. To my knowledge, a classic omit
map with simulated annealing would do the job regarding removing the
model bias. Actually, I found an interesting alternative in PHENIX
called a Kick map, were a series of maps computed from a ramdoinised set
of models yields a averaged map ideally free from model bias. Does
anyone has a preference for any of those schemes? Are there more
alternative to prove a conformational change in a model phased with a
molecular replacement solution?
Thank you very much in advance.
--
Israel Sanchez Fernandez PhD
Ramakrishnan-lab
MRC Laboratory of Molecular Biology,
Hills Road, Cambridge, CB2 0QH, UK
