Re: [ccp4bb]: Q: reduce size of Rfree set

[R J Read <rjr27@cus.cam.ac.uk>]

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

On Mon, 10 Jul 2000, Laurent Maveyraud wrote:

> ***  For details on how to be removed from this list visit the  ***
> ***    CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html    ***
> 
> Dear all,
> 
> I refined a structure against data to 2.3 A, where the Rfree set
> included 2600 reflections (5 % of data). I now have data to a much
> higher resolution, and putting 5 % of data in the Rfree means  11000
> reflections, which is quite a lot. I would like to use a smaller
> fraction for the Rfree set (say 1 %). For this, I need only to keep 1
> every 5 reflections of the "low resolution" free set.
> 
> Any hints on how I can do this ?
> 

If you follow Eleanor's suggestion, I don't know if you even need to shake
things up.  With more than 4 times as much data, at about 1.4A resolution,
any overfitting to 2.3A will probably disappear quickly.

However, if you do want to keep 1 in 5 of your old test data, you can do
it fairly easily in sftools.  I'll assume the working data have a 1 in
column RFREE and the test data have a 0.  I'll also assume you have a
column called F_Old which contains the original amplitudes.  Then do the
following in sftools:

read my.mtz
select col rfree = 0          ! only old test data
calc col rfree = rfree(0.2)   ! keeps 1 in 5 of old test data
select all
select col F_old absent       ! only new data
calc col rfree = rfree(0.01)  ! flag 1% of new data
select all
write my_new.mtz

Perhaps a bit inelegant, but it should do what you want.

Randy J. Read
Department of Haematology, University of Cambridge
Wellcome Trust Centre for Molecular Mechanisms in Disease
Cambridge Institute for Medical Research
Wellcome Trust/MRC Building     Tel: + 44 1223 336500
Hills Road                      Fax: + 44 1223 336827
Cambridge CB2 2XY, U.K.         E-mail: rjr27@cam.ac.uk