I believe the OP was asking how to best make an "element density map"
where the map value is proportional to the occupancy of not just any
anomalous scatterer, but a specific element of interest. For example,
suppose you have Zn and Ni in your protein, but you are not sure which
atom is which. If you have solved the structure and you have data
collected near the Zn and/or Ni edges, then you can make a "Zn density
map" and an "Ni density map" to resolve the "mysterious metal
problem". I don't think there is a "canned" procedure for doing this
because you have to bear in mind what the f" and f' are for each
dataset, but below I describe something that might work.
Making an "element density map" from conventional MAD data is a
slightly different procedure than what you would do to get the total
heavy atom contrtibution (FH) you would use to solve the structure by
MAD,SIRAS, etc. Specifically, you are not just interested in getting
the maximum anomalous signal, you are trying to isolate the signal
coming from a particular element. You must have more than one
wavelength to do this! For example, if you have "peak", "inflection"
and "remote", then you can obtain a map of "f'' from Zn" by computing
a phased anomalous difference Fourier map from the "peak", and the
"inflection" and then subtracting the "inflection" map from the "peak"
map. This will add a bit of noize, but it has the nice property of
eliminating peaks from S, Cl and any other anomalous scatterers you
might have. Another way to get "element contrast" is to use f'. Here
you would make a "regular" isomorphous difference map between two
wavelengths where the f' of the element of interest changed a lot.
This only happens near the edge. Once you have both kinds of "Zn
density" maps, you can then add them together (perhaps scaling them as
described by Matthews 1966, or using something like REVISE).
As for the rest of the discussion, when it comes to maximizing phasing
signal per unit dose, I prefer two wavelengths: split the difference
between the "peak" and "inflection" as "wavelength 1", and then use a
high remote as "wavelength 2". I flippantly call this "DAD" for
"dual-wavelength anomalous diffraction", but I suppose you could also
call it "Bifrucate Absorptive and Dispersive Anomalous Scattering".
Anyway, this gives you both kinds of anomalous difference (f" =
"absorptive", f' = "dispersive") for the price of two wavelengths.
The compromise from not being exactly "on" the optima isn't so bad in
practice. This actually makes sense when you consider that the "peak"
and "inflection" for Se are generally only ~3 eV apart and the
bandpass of a Si(111) monochromator is ~1.8 eV at 0.98 A wavelength.
In general, however, I cannot over-stress how important it is to
collect quasi-simultaneous wavelengths! If you just collect one
wavelength after another (OAA) and there is even a smidge of radiation
damage, then your f' will be completely lost in the non-isomorphism
generated by the damage. By changing the wavelength often (every
image is preferable, unless your monochromator is incapable of doing
this), all your f' differences will at least be comparing "apples to
apples" and "oranges to oranges" (in this analogy, X-rays turn apples
into oranges).
-James Holton
MAD Scientist
On Wed, Jul 6, 2011 at 12:03 PM, Pete Meyer <[log in to unmask]> wrote:
> It'll depend on your data, but you'd probably be better off using the
> inflection (rather than peak) and remote datasets for dispersive difference
> maps. This signal is usually fairly weak to begin with, and not using the
> infection datasets weakens it further.
>
>
> Pete
>
> FWIW - my understanding is that "dispersive" is often used to distinguish
> differences in f' from Bijvoet differences (in f'') during MAD, at least
> when treating MAD as MIRAS.
>
> Jacob Keller wrote:
>>
>> Dear Crystallographers,
>>
>> it seems to me that for clearly identifying/characterising anomalous
>> scatterers for a solved structure, one could make a map using two
>> datasets: one at the f" peak, one low energy remote. One would then
>> use the signal both from the Bijvoet differences in the peak dataset
>> plus the differences between the peak and low-energy datasets, which I
>> think I have seen called "dispersive" differences. I guess this would
>> be like a MAD map, but using pre-existing model phases--is there such
>> an animal in the software, or would it even be helpful?
>>
>> Jacob Keller
>>
>
|