Re: [ccp4bb]: CNS composite SA omit - summary

[William Scott <wgscott@chemistry.ucsc.edu>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

Another option that you might wish to try is EDEN:  
http://www.edencrystallography.com/index.html

You can run this in "correction" mode to look for model errors and there 
is also an option for randomly perturbing the map (and then re-refining 
it) as a further mechanism for minimizing model bias.  It also works 
quite well when large chunks of the model are missing, when run in 
"completion" mode.  It does not do structural refinement per se, but 
rather real-space improvement of the maps.  As a consequence, there are 
no difference maps -- you just have to contour lower to see density for 
the missing (or wrong) parts of the model.


On Friday, July 26, 2002, at 08:00 AM, Phoebe Rice wrote:

> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
>
> And the verdict is ... general disillusionment?
>
> individual comments:
> --------------------------------------------------------
> In my opinion, SA refinement and maps are never any better than those 
> that I
> get from the normal minimization runs in CNS.  This is for structures 
> where I
> have data to roughly 2A, and also to higher resolution.  SA refinement 
> tends
> to randomize well-built structures, and does as well, but not better 
> than
> simple minimization on structures that have errors.
>
> So you're not the only one.
> ---------------------------------------------------------------
> I'm interested as my observations have been similar to
> yours.  Basically, very little differences and I've
> stopped doing the omit maps.  Did anyone suggset a
> magic variable that we've got set incorrectly?
> -------------------------------------------------------------------
> In my opinion, SA-omit maps are quite useless because:
> 1- they take too long to compute;
> 2- in your case you are using the "wrong" coefficients (2mFo-DFc).
>
> Much more interesting are modified SIGMAA maps (3mFo-2DFc for acentrics
> and 2mFo-DFc for centrics) and Bhat's OMIT maps computed using these
> coefficients. Also reduces calculation time, the maps clearly indicate
> where errors are.
> ----------------------------------------------------------------
> It is my (our) experience as one moves towards the end of a refinement 
> that
> composite
> omit maps calculated via the sa_omit script in CNS and "normal" cross
> validated
> 2mfo-dfc maps tend to track each other quite well.   The only potential 
> issue
> is that
> the composite omit maps may be a bit more noisy than the 2mfo-dfc maps 
> which
> means
> that some care should be taken in adding solvents based on peaks in the 
> omit
> maps or perhaps
> in dealing with disordered regions in structures which give rise to 
> broken
> density in the
> normal 2mfo-dfc map.  In fact most differences we have noticed have 
> involved
> regions
> of weak density in the 2mfo-dfc map where the composite omit maps 
> sometimes
> tend
> to be a bit more broken (at a given contour level) than the sigmaa 
> weighted
> map.
> ------------------------------------------------------------------
> ..... anyway, if you load both maps into mapman you can at
> least check *how* similar the they are. if they are
> *identical* i suggest that something has gone awry
> with the map files at some stage
> .....
> > I get a correlation coefficient of 0.998.
> .....
> not sure if that
> is good news or not. well, it saves running eternal
> composite sa omit map jobs
>
>
>
> Phoebe A. Rice
> Asst. Prof., Dept. of Biochemistry & Molecular Biology
> The University of Chicago
> phone 773 834 1723
> fax 773 702 0439
> http://molbiol.bsd.uchicago.edu/faculty/pRice/index.html
> http://bmb.bsd.uchicago.edu/
>
>