sigma_cutoff

Yu Wai Chen ywc at mrc-lmb.cam.ac.uk
Tue Mar 14 08:57:35 EST 2000


Matthew Meyer wrote:
> 
>     I've been re-refining some mutant structures.  This time I'm
> throwing away peaks where I/sigI < 2.0.  R and Rfree are much improved
> (by typically 4 %) over the previous refinements when sigma_cut = 0.0.
> However, after expansion to full resolution, far fewer waters are picked
> before further picking pushes R back up. I notice that many waters in
> the old model are not within hydrogen-bonding distance of anything.  I
> recently read that was bad in an article by Kleywegt
> (http://journals.iucr.org/d/issues/2000/03/00/gr0949/gr0949bdy.html#BB100).
> My question is which is right? Would the reporting of only 3 waters in a
> ~140 aa protein refined to 2.45 A be an indication for refinement
> errors?

For I/sigI = 1, your "signal" is only as high as your noise; i.e. it is
not a signal statistically.  For I/sigI < 1, you are including pure
noise.

So think about it, what you saw, by introducing noises (or signals that
cannot be distinguished from the noises) into your data, can this be
right?  The noises in data produce fluctuations in your maps and the
program picked up high noise peaks and assign these to be waters.  This
is classical "over-fitting".  The most important thing is to watch for
Rfree.  If it didn't improve, what is the point of introducing the many
waters?  And also, for 2.45A, you really don't expect to see many
ordered waters.

-- 
===================================================================
Yu Wai CHEN, Ph.D. ..................   email:ywc at mrc-lmb.cam.ac.uk
 Centre for Protein Engineering,             tel:+44-(0)1223-402148
 MRC Centre, Hills Rd, Cambridge CB2 2QH, UK fax:+44-(0)1223-402140
 WWW homepage: http://www.mrc-cpe.cam.ac.uk/~ywc




More information about the X-plor mailing list