On 19/02/15 11:00, Emilia C. Arturo (Emily) wrote:
Hello all.
Hello Emilia (Emily),
I'd like to understand what it is I'm looking at when I use Coot's
density fit analysis tool. I recognize that there was a post related
to this topic on the Coot bb a while ago --the discussion was on how
to interpret the red-ness or green-ness of the density fit plot
(https://www.mail-archive.com/coot@jiscmail.ac.uk/msg02995.html)
https://www.mail-archive.com/coot@jiscmail.ac.uk/msg02995.html%29--but --but
it doesn't seem the issue was resolved then (i.e. what does 'red'
really mean? is it ...bad?).
I failed to fix this for 0.8.1 I will add it to the list now for 0.8.2.
The colours were set for typical (~2.1A on the absolute scale).
Now I have more to ask that involves my using phenix-generated FEMs to
build in Coot.
So what I've done is the following: I adjust my model in Coot, using a
phenix-generated FEM as the map for fitting, then refine with phenix,
and using the refined pdb and reflections file, I use phenix to
generate a new FEM. Then I repeat. At some point I learned about
Coot's density fit analysis tool and took a look at how my model fits.
If the map that is selected in the sidebar of Coot is a FEM, then the
density fit analysis plot looks mostly green everywhere - fine. If,
however, I select as my map in the Coot sidebar the 2Fo-Fc that phenix
had generated along with the latest refined model--the one I'm
examining with the density fit analysis tool--then Coot's density fit
analysis plot looks red (with values ~ 0.3), with splashes of
orange, barely any green or yellow (with values ~ 0.3), almost
everywhere.
:-( yes - this doesn't work for FEM maps.
So these are my questions: What are the units of the density fit values?
electrons/A^3 with volume normalization.
I think FEM maps are normalized (so that the mean is 0 and standard
deviation is 1.0) and this is not typical for maps generated from
structure factors on the absolute scale - where you'd expect density
values smaller than this by a factor of 2 or 3 (or so)).
So the density values from FEM maps are higher and the validation graph
greener.
i.e. What is the calculation that's done?
The map is sampled at the atoms and weighted by occupancy.
I'm surprised that the FEM-dependent density fit graphs look so
different (i.e. so green) relative to the graphs generated if my map
is set to the 2Fo-Fc from the loaded model;
I know less about FEM maps than I should, but I can refer you to Section
7.9.1 of the documentation:
https://www2.mrc-lmb.cam.ac.uk/Personal/pemsley/coot/web/docs/coot.html#Validation-Graphs
After you've rescaled, I'd be interested in your reflections of the
differences.
Regards,
Paul.
