Formally, the "resolution" of any image is the minimum distance between
two objects in the image that can be "resolved", or separated. Think of
two Gaussian-shaped peaks that are 4 A apart. If you "blur" the image
enough, eventually the two peaks merge into one, and there is no longer
a valley between the two peaks. The point where the two peaks become
one is when the "resolution" is 4 A.
This should not be confused with the ability to know that there are two
peaks! For example if you know that all the peaks in your image are
round (like atoms) and you see one peak that is highly oblong and twice
the total intensity that it should be, then you can be pretty confident
that there are two things stuck together there. You might even be able
to accurately compute how far apart they are by looking at how oblong
the peak is, and perhaps simulating images until you get something that
looks like what you see. This application of prior knowledge has been
given the name "super resolution" in some circles. Others might call it
"centroiding". The super-resolution comes from the fact that if you
know something about the shape of the curve, you can usually get an
error bar for the center of a peak that is much smaller than the width
of that peak.
So, the "resolution" of an image tells you the minimum feature size you
can see, but only if you know nothing else about the image. Prior
knowledge gives you "super-resolution". So, when it comes to hypothesis
testing, the resolution of the image is only one component of the
information you have at hand. To answer questions you need to define
controls and assays, and to "prove" something statistically you need a
well-defined "null hypothesis". AKA: if not in a salt bridge, where
else could the side-chain be? These are the games you play when you
don't have a beautifully clear image.
It is true, however, that if the crystal doesn't diffract at all, then
you can't use data from it to draw any conclusions.
-James Holton
MAD Scientist
On 5/24/2015 2:54 AM, Smith Liu wrote:
Dear All,
In order to acceptably explain the salt bridges, hydrophobic
interactions and H-bonds among subunits in the crystal structure of a
protein complex, is there a threshold resolution of the crystal, for
example, if the crystal is poorer than 4A or 5A, the crystal structure
solved cannot be used to acceptably explain the intersubunit
interactions at the non-covalen bond level?
Smith
