Hi guys,

I've just started looking at the use of cross-validation for dipolar 
coupling refinement of protein structures in XPLOR-NIH.  I can see how 
the R-factor is incorporated into an anneal script, and have all my 
couplings normalised to DNH, allowing me to report a working R(dip) for 
the total set.  I would like to know whether there is an automated 
process for cross-validation?  I'm assuming that I understand how to do 
it manually, that is:

*  remove a reasonable percentage of couplings and re-calculate a 
working set of structures
--> R-factor is generated for each structure by the anneal script and is 
R-work.
*  use only the couplings left out above and re-calculate a set of 
structures
--> R-factor is generated for each structure as above but is now R-free.
* do this for a few sets and compare the appropriate R-factors for 
cross-validation.

Can this be done in the one script? I was thinking of joining 2 anneal 
scripts  one after the other but I was interested to see if there is a 
more efficient way?  

Assuming I'm correct in my understanding of the cross-validation (I've 
read the JACS 121, 1999, paper a lot this weekend in between structure 
calculations and thesis writing and I think I've got it!), should I 
report R-factors as "factors" between 0 and 1, or as a percentage?  I'm 
leaning toward the latter as a couple of recent publications report 
percentage, but I just want to make sure this is the standard way.

Thanks in advance,

Regards,

Haydyn

--
----------------------------
Haydyn Mertens
PhD student
Dept. Biochem.
University of Melbourne
Australia

Reply via email to