*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Peter Adrian Meyer wrote:
The problem with an anomalous signal-to-noise ratio of 1 is that you
can't tell if you got anything just by looking at the data. That is, if
you have no signal at all you still expect your DANOs to have an average
value that is equal to the average value of SIGDANO. However, the
signal does tend to be stronger for low-angle data. The "DelAnom
correlation between half-sets" analysis done by SCALA does appear to be
a good way of detecting pretty weak anomalous signals. I highly
recommend that you look at it.
Why would the anomalous signal be stronger for low-angle data? f'' (as I
understand it) is roughly constant with respect to scattering angle
(resolution), while f0 falls off rapidly with increasing scattering angle.
delta f' (for anomalous) should behave similiarly.
Am I missing something here?
Two things: B-factors and background noise.
For error-free data (like the kind you read about in Stout and Jensen)
it is true that you can see the anomalous contribution is flatter than
your typical f with respect to scattering angle (the core electrons are
a sharper peak than the whole atom). However, core electrons are
subject to the same overall B-factor as the rest of the atom, and as
soon as you apply a realistic overall B-factor, the anomalous
differences at high angle become quite a bit smaller than the ones at
low angle. So, strictly speaking, I think I was not wrong to say that
the "anomalous signal" is larger at low angles. Although you have a
point that the strength of the anomalous signal increases RELATIVE to
the protein contribution at high angle.
All this becomes academic when you consider noise. High-angle data are
weak, and weak data are hard to measure. Nobody ever gets useful
anomalous differences in their outer resolution bin (unless you are
working with crambin). Perhaps it would have been more instructive for
me to say that the anomalous signal-to-noise ratio is stronger for
low-angle data. Sorry if I wasn't clear.
By the way, I may also have been unclear on the workshop page about the
program "cheat". Unfortunately, "cheat" is not a "real" heavy-atom
finding program. Sorry if anyone got excited. "cheat" is a jiffy
program I wrote that uses the solved structure to calculate a
model-phased anomalous difference Fourier map, and then I just pick
peaks in that map (this is explained a little further down on the
page). So, you can't run "cheat" until you have already solved the
structure. This was intended as a "positive control" for heavy-atom
finding since it is difficult to imagine a program that could do any
better than starting with the right phases (although I'm sure Gerard
Bricogne has some ideas).
I do find it interesting that no modern heavy-atom finding programs can
find the sites in "badsignal", but "cheat" can. This implies that there
may still be room for improvement in heavy-atom finding algorithms in
the future. The map from phasing "badsignal" with the right sites is
crappy, but not completely uncorrelated with the right structure. I
have now linked these maps from the workshop page:
http://bl831.als.lbl.gov/~jamesh/workshop/
O macros for looking at them are available in the "answer key" tarballs.
-James Holton
MAD Scientist
