I made a movie demonstrating the effect of leaving out low-res data on
an electron density map once:
http://bl831.als.lbl.gov/~jamesh/movies/index.html#lores
But it is interesting I think that the brightest reflections are
actually what is important. Yes, the average spot intensity increases at
low res, but the variance is high and there are a surprising number of
bright spots at intermediate resolution. This is demonstrated by what
happens if you have too many overloads:
http://bl831.als.lbl.gov/~jamesh/movies/index.html#overloads
In retrospect, it makes sense that larger structure factors represent a
larger fraction of the total electron density in the map. After all, the
map is just the sum of all the Fs. If you have fewer Fs, then individual
(bright) ones will become "more important".
Still, I have to admit I have not done too many systematic tests on the
success/failure of an MR or site-finding program as I throw out more and
more bright spots. I'm sure, like everything else in crystallography,
there will be some cases where it doesn't matter and other cases where
it matters a lot. It all comes down to signal and error. Missing
observations can be thought of of as a form of "systematic error". If
the signal is huge strong diffraction and/or big anomalous differences),
then you can introduce large amounts of all kinds of errors and still
solve the structure, but if the signal is weak, then you have to be very
careful to reduce ALL sources of error to be less than the signal. Any
one error that is bigger than the signal will bury it. Examples of this
are small crystals with high background (weak spots, high
photon-counting noise) or trying to measure a weak anomalous signal
(small % change in spot intensity) in the presence of something that
introduces a small % error into the spots, such as a vibrating crystal,
a lousy shutter or radiation damage.
-James Holton
MAD Scientist
Richard Gillilan wrote:
>>
>
> Several times I have heard that low order (small angle) reflections
> are more important when solving low-resolution structures. I presume
> it is more than just a question of obtaining greater number of
> reflections.
>
> Does anyone know why low-order reflections are so important in these
> cases?
>
>
> Richard Gillilan
> MacCHESS
|