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

Re: [ccp4bb]: A simple question of resolution



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

Dear All,

I would suggest to do very small experiment (basically for those who favor
cutoffs):
Collect data to highest resolution. A little bit higher than you can see by eye.
Then process all data and
do refinement with various resolution cutoffs. You can predict that if R-merge is
around 50-60, I/sigmaI is around 1 (preferably
a little bit higher than 1) then your
crystallographic R-free is around 20-30 (at high resolutions, at low resolution it
might be higher related with model
errors) suggesting that those reflections do contain information about structure.
I have seen this tendency in few cases.
Sometimes including higher order reflections may make difference between solvable
and unsolvable cases (for instance Tassos's
example of automatic tracing).

Two important points are that we are dealing with Fourier transformations and
contribution of errors from coordinates
(due to non-modelled atoms or atoms with error) is in general larger than errors
in reflections. (Refinement and model building
programs are much better than before but they are not that good yet)

In refinement when you remove a reflection you assume that it is precisely what is
calculated from coordinates. When it is
weak and included you assume that it is weak.

What resolution to report is old problem. Optical resolution seems to be best
available at the moment.

Regards
Garib

P.S. I believe Gerard Kleywegt has done some experiment on this and results are
interesting and seems to support message
that use more reflection.


"Edward A. Berry" wrote:

> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
>
> Phil Evans wrote:
> >
> > Why use a cut off in refinement? This is unecessary and a Bad Thing to do
> >
>
> If one does not use a cut-off on F/SigF, why use a resolution cutoff?
> There is nothing intrinsically objectionable about High Resolution Data,
> the only reason I can see for excluding them is if they are weak (nonexistent).
>
> Is it better to use a resolution cutoff which excludes a small number of
> 5-10 sigma reflections in the last shell while keeping 0.5 sigma reflections
> at lower resolution, or a sigma cutoff which rejects weak reflections
> wherever they occur?
>
> (this just to elicit more discussion- I am a firm believer in a zero F/sigF
> cutoff AND using resolution as far as truncate will allow). How to
> report the resolution without misleading is then a problem. I would
> favor Bart Haze's/Gerard Kleywegt's "effective resolution" based on the
> number of reflections above some cutoff.
>
> Ed

--
------------------------------------------------------------------
Garib N. Murshudov, Chemistry Department, University of York, U.K.
Tel: Home +44 (1904) 42 62 87, work:  +44 (1904) 43 25 65
------------------------------------------------------------------