Hi all,
I found that it is possible (and to some extent advised) to refine the
cell-scale (aka magnification or "pixel size") of a cryo-EM map using
the phenix program PHASER. This program is originally intended to use an
input atomic model as a search template to assign phases to X-ray
amplitudes, but it has the capacity to refine the cell-scale and even
mentions EM data as an application.
There are refinement macrocycles to iteratively refine the fit and
cell-scale, but these seem to be rigid-body searches of the template
against the map, not conventional refinement macrocycles that permit
alteration of covalent geometry. This strikes me as problematic; one
would need an external ground-truth structure that agrees well with my
density to get a reliable cell-scale estimate. If I, on the other hand,
have no such structure and e.g. build a novel protein into my map, I
already suffer a scale-bias from fitting against this map. That is, a
cell-scale refinement of map A using a model built against map A is not
unbiased. In fact, even using an external model would result a refined
magnification of map A which suffers potential bias from that external
reference, right?
In my mind, we want macrocycles that permit alteration of covalent
geometry during cell-scale refinement , in which case the cell-scale
will only be "biased" by the force-field used to refine covalent
geometry and secondary structure, right? What methods to people
currently use? I've heard of the "rescale density in chimera by
changing pixel-size to maximize cross-correlation with reference
density", but I don't like it. It seems a less rigorous way to end up
with exactly the same bias as above.
Cheers,
/Björn
########################################################################
To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=CCPEM&A=1
|