Re: [ccp4bb]: Eigenvalue Filter- More questions..

[Martyn Symmons <martyn@cryst.bioc.cam.ac.uk>]

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

Dear Raji
         what phx said...
         but to clarify - the denzo manual is confusing on this point -
the eigenvalue filter is in fact automatic and cuts in when scalepack
notices that some of the parameters to be fitted are too closely
correlated - so it makes a decision not to fit them to begin with - I
think that if this strategy works for it then later in the refinement if
things are looking good for it converging better _then_ scalepack will
automatically start fitting more of the parameters
         But you can decide to fit fewer parameters - I noticed you fixed
B and did not fit beta* yet anyway. If you have a good idea of the best
cell axes lengths then you can always re-run denzo and fix the cell to
that value. Do "grep 'unit cell' *.x" in a unix shell to see whether denzo
is being naughty and letting the cell wander too much through the frames.
(Beware of if beta is actually close to 90.000 you can start out with
denzo choosing beta greater than 90.00 but then it wanders below 90.0 and
scalepack throws up its arms in disgust.) Also you should fix either
distance or cell if it is low resolution data (worse than 3.5 say)
   Another approach is to try restraining the variation of the batch
scales and b factors between adjacent frames. There are some examples of
this in the denzo manual.
   Another approach is to use mosflm/scala - and in fact if you fix the
cell in denzo and indicate 'york output' then you can use denzo output for
input to scala - there is very good scala documentation on this.
on this.
   Good luck
             Martyn (Symmons)

Martyn F. Symmons
Crystallography and Biocomputing Group
University of Cambridge
-----------------------
"Am fear a ghleidheas a theanga,
 gleidhidh e a charaid."
Seanfhacal Gaidhlig (Gaelic Proverb).


On Wed, 26 Jun 2002, raji wrote:

> Date: Wed, 26 Jun 2002 11:09:41 -0600
> From: raji <raji@lamar.colostate.edu>
> To: ccp4bb@dl.ac.uk
> Subject: [ccp4bb]: Eigenvalue Filter- More questions..
>
> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
>
> Hello Everyone,
> I am including my scaling input file as I did not provide enough details in my
> previous email.
> All suggestions/criticisms are welcome.
> Thank you.
> Raji
>
> /data1/xtal/HKL/1.97.2/scalepack8m > gh4.slog << eof-sh
> number of zones 15
> number of iterations 25
> estimated error
> 0.03 0.03 0.03 0.03 0.03
> 0.03 0.03 0.03 0.03 0.03
> 0.03 0.03 0.03 0.03 0.03
> error scale factor 1.0
> ignore overloads
> output file gh4.sca
> reference film 10
> fix B
> space group p21
> resolution 100 3.0
> add partials  10 to 250
> format denzo_ip
> [@reject]
> reject hkl
> postrefine 22
> fit crystal a* 10 to 250
> fit crystal b* 10 to 250
> fit crystal c* 10 to 250
> fit crystal mosxx 10 to 250
> fit film  rotx 10 to 250
> fit film  roty 10 to 250
> rejection probability 1.e-4
> write rejection file
> sector  10 to 250
> FILE    10 './x/gh4_1_###.x'
> eof-sh
>
> My previous email-
> Hi,
> I am trying to scale one of my datasets using 'Scalepack' and I get the
> following message -
>          "Warning - eigenvalue filter in action.
>          It detected that parameters are too strongly correlated.
>          Reduce number of parameters fitted simultaniously in the future.
>          Film # of Eigenvectors not fitted are printed below:"
>          Denzo manual suggests the use of the EIGENVALUE FILTER with
> great caution.
> Any help/suggestions on how to proceed will be of great help.
> Thank you very much.
> Raji
>
> Rajeswari Edayathumangalam
> Dept. of Biochemistry & Molecular Biology
> Colorado State University
> Fort Collins, CO 80523 USA
>
>