Dear Gregory,
You need a sufficient number "n" of free reflections to get a
statistically valid Rfree value. 500 should be ok. I normally select 5%
free reflections. However, I always hate it when programs decide that
they know better than the user. Nobody prevents you from using your
original mtz, which was used as an input for BALBES, for further
refinement and ignore the mtz produced by BALBES. The only result from
molecular replacement you really need is the positioned and oriented pdb
file. You may have to set the space group in your mz to the space group
found by BALBES.
Good luck!
Herman
________________________________
From: CCP4 bulletin board [mailto:[email protected]] On
Behalf Of Gregory T Costakes
Sent: Friday, April 15, 2011 4:20 PM
To: [email protected]
Subject: [ccp4bb] Question about BALBES R(free) fraction
My name is Greg Costakes and I am a graduate student for Cynthia
Stauffacher at Purdue University. I am attempting to solve the structure
of an 18kDa protein domain using molecular replacement to obtain proper
phasing (44% identical / 78% homologous to an existing structure).
Crystals diffracted to roughly 2 angstroms and are in the space group
P212121. There are 10,200 unique reflections with 5.7 fold redundancy. I
chose to make an R(free) set containing 10% of the reflections so that I
will have just over 1000 in the set.
After failing to obtain a valid solution from Phaser and Molrep,
I turned to BALBES which was able to successfully perform the molecular
replacement and give me final R/R(free) values of 0.3/0.34. However, I
noticed the MTZ file that BALBES gave me contained an R(free) set of 5%
containing only 500 reflections. I was wondering why BALBES changed the
percentage of reflections pegged for the R(free) set. Is there a way to
prevent BALBES from changing your R(free) fraction size? Also, is 500
reflections enough for an R(free) set? I had previously been told that
R(free) should contain between 1000-1500 reflections. Please let me
know. Thank you.
------------------------------------------------------------------------
-------
Greg Costakes
PhD Candidate
Department of Structural Biology
Purdue University
Hockmeyer Hall, Room 320
240 S. Martin Jischke Drive, West Lafayette, IN 47907
------------------------------------------------------------------------
--------
