[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ccp4bb]: Free-Rflag and NCS

***  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 Tue, 26 Dec 2000, Hurley, Thomas D. 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    ***
> Hello,
> I have two cases of high-order NCS that we are ready to refine, one 10-fold
> and another 12-fold.  
> Question 1 - How to assign Free-Rflags that account for the known NCS
> operations?

There is no ideal solution. The best that I know of today is the thin shell
Rfree selection in SFTOOLS. Probably better than random selection in such high
NCS cases as yours. However, the shells are really too thin but you can't make
them much thicker and still sufficiently sample your resolution range. If you
have true point group symmetry (10-fold, 12-fold, or 52, 62 point groups) then
you could select NCS related spheres of Rfree reflection. My personal version
of SFTOOLS has this option but it hasn't been properly tested. I don't know of
other programs that implement this.

The most serious problem with high NCS is that you will overestimate Sigmaa
which will lead to non-ideal weighting during your maximum likelyhood
refinement. But how large this effect is in practice has, to my knowledge, not
been quantified. Perhaps the maximum likelihood gurus could give some feedback
on this. Your Rfree will also be underestimated making your structure look
better than it is. I don't think this is a problem. Just state that the Rfree
is most likely somewhat underestimated and I don't think any reviewer would
argue with you.

> Question 2 - What is the "best" way to treat 3.4 A, data in terms of
> conversion to F's?  As the resolution is less than optimal for anything that
> resembles an accurate Wilson plot, should we use the TRUNCATE NO option and
> accept ~4% loss of data (99% complete at this point)?  Another possibility
> is to use the MLI option in CNS and import the I's for refinement.  Anyone
> have experience using the MLI option under these conditions?  (To forstall
> the obvious responses - Yes, we do intend to use NCS-strict).

The truncate option only uses the Wilson plot to determine the expected
intensity distribution at a given resolution. I don't expect that the result
will be very sensitive to the Wilson scaling parameters. You can test this by
running the program several times. Once letting it figure out the Wilson
parameters itself and some other runs where you specify slightly different
scale values (with the SCALE option) and then use SFTOOLS or some other
program to compare the results. Remember that the largest differences will be
for the weakest reflections.



Dept. of Medical Microbiology & Immunology
University of Alberta
1-15 Medical Sciences Building
Edmonton, Alberta, T6G 2H7, Canada
phone:	1-780-492-0042
fax:	1-780-492-7521