Re: [ccp4bb]: Contour

[Bart Hazes <bhazes@ualberta.ca>]

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


I think in practice Andrew and Gerard are doing the same thing. You DO look at
5 sigma peaks just as you look at outliers in the Ramachandran plot, too close
contacts etc. Not because these must be errors, but because they are
suspicious and you want to visually make sure that they are not errors or to
fix them. Yes, a five sigma peak in a "perfect" map with very low rms in terms
of e/A**3 is meaningless but since the rms of the map is based on
statistics, these will be extremely rare. What I would recommend is to use a
peaksearch of the Fo-Fc map and look at the peaks sorted by peak hight.
Starting at the strongest peak work your way down until you have had a whole
row of peaks that you feel are not telling you anything. Important: don't
forget to look at the biggest negative peaks. A -5 sigma peak is just as
suspicious as a +5 sigma peak.

Wrt to putting a water in any >3sigma peak, this should clearly not be done.
Interpreting density means you want to find a CHEMICALLY PLAUSIBLE explanation
for any density feature. For a water that means that it should have at least
one decent hydrogen bond with the protein and no too close contacts. You
should always judge both density and "geometric/energetic sensibility" and if
your density is poor you want to give more "mental weight" to the geometry.
In many cases you will end up in Andrews situation where you have the feeling
that the difference density does indicate a problem but you can't figure out
how to interprete it in which case you better leave it alone. Really the
perfect model doesn't exist, you just want to get as close to it as possible
given your data quality. Try never to overinterprete your density since others
that look at the structure without seeing the density will blindly believe
what you have build even if they shouldn't.

Bart

On Thu, 6 Sep 2001, Leslie A. wrote:

> ***  For details on how to be removed from this list visit the  ***
> ***          CCP4 home page http://www.ccp4.ac.uk         ***
>
>
> > so you suggest that we should go back to the dark
>  > ages and put a water in every 3 sigma Fo-Fc peak ?
>  > (that's what people may be tempted to do after
>  > reading this)
>  >
>  > and again: "5 sigma" is a meaningless concept in an
>  > Fo-Fc map - you must talk in absolute terms. we're
>  > talking about electrons here, not statistics. saying
>  > that 'N sigma' is a significant feature no matter what
>  > is like saying that a beer should cost one currency
>  > unit in every country, i.e. a pound in the uk and a
>  > lira in italy. you need to put them on an absolute
>  > scale (e.g., euros ;-)
>  >
>  > --gerard
>
> I made no such suggestion. I clearly stated that one might not be able
> to interpret such features.
>
> On reading your own original statement, people might be tempted to
> believe that a 5 sigma peak in their final difference map was not
> significant. I assert again that in the case under discussion
> (difference maps for a macromolecular structure) this is not the case.
> I also assert that for macromolecular structures, it is actually NOT
> necessary to calculate the difference maps on an absolute scale, as
> you are suggesting, simply because there are always sufficient errors
> in the model to ensure that a 5 sigma peak is not just "noise".
>
>   Andrew Leslie
>

===============================================================================

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

===============================================================================