Data reproducibility

Plan your small angle X-ray or neutron scattering measurements
Post Reply
Message
Author
uan
Member
Posts: 11
Joined: 2009.03.22 18:32

Data reproducibility

#1 Post by uan » 2009.05.25 14:28

Hi there,

I have a problem with curves shifted along the scattered-intensity-axis: it seems, as if I had some strange offset.
I have these four SAXS-curves, after subtracting the buffer. Each curve corresponds to one capillary. The buffer was measured in all cases in the same capillary as the protein solution. All four measurements were done with same capillary-diameter, same protein-solution/concentration and all other parameters, as we could think of.
Still, as you can see, the four curves do not overlay. Is this a buffer-subtraction problem? Have you experienced measurements like that, too? How else, in a more satisfying way could the buffer-subtraction be done?

many thanks in advance!
Attachments
SAXS_curves.pdf
(28.27 KiB) Downloaded 833 times

User avatar
AL
Administrator
Posts: 686
Joined: 2007.08.03 18:55
Location: EMBL Hamburg, Germany
Contact:

normalization problem

#2 Post by AL » 2009.05.25 19:01

It looks like a normalization problem. The concentration is never measured 100% accurately so you always get some offset along the I axis in log scale.
Besides on some beamlines normalization against beam intensity is not done properly.
(I suppose the exposure time was the same.)

Scale all curves (by multiplying) to see if they overlap - i.e. the shape is the same. Sometimes it is necessary to subtract a constant to make the curves overlap perfectly.

If you suspect a buffer subtraction problem - check on linear scale if there are many negative values in the higher q range. Try to subtract different buffer measurements or their average to decrease the amount of negative I values. (normally one should measure a buffer before and after the sample).

uan
Member
Posts: 11
Joined: 2009.03.22 18:32

Re: normalization problem

#3 Post by uan » 2009.05.26 10:04

Thank you AL. I have some comments and further questions:
AL wrote:The concentration is never measured 100% accurately so you always get some offset along the I axis in log scale.
Besides on some beamlines normalization against beam intensity is not done properly. (I suppose the exposure time was the same.)
Sure, exposure time was also the same. The protein in all four cases was taken from the very same Eppi, so even if the concentration was not well indicated, I would expect the curves to look the same...
AL wrote:Scale all curves (by multiplying) to see if they overlap - i.e. the shape is the same. Sometimes it is necessary to subtract a constant to make the curves overlap perfectly.
You mean I can do this "by eye"? But then the problem is, that if they overlay at the lower angle region, they do not at higher angles and vica versa.
AL wrote:Try to subtract different buffer measurements or their average to decrease the amount of negative I values. (normally one should measure a buffer before and after the sample).
The buffer measurements of the four separate capillaries are also slightly shifted on the Y-axis, therefore I would have expected that the protein_solution - buffer differences should give the same. I don't see why the average of the four buffer-measurements would help...
Measuring the buffer before AND after the sample might help: you mean, I should take the average of those?

User avatar
AL
Administrator
Posts: 686
Joined: 2007.08.03 18:55
Location: EMBL Hamburg, Germany
Contact:

Re: normalization problem

#4 Post by AL » 2009.05.27 16:17

The protein in all four cases was taken from the very same Eppi, so even if the concentration was not well indicated, I would expect the curves to look the same.
From my experience the concentration can change for various reasons - evaporation of the buffer (especially if it is volatile), dilution of the sample if the sample cell was not rinsed/dried properly etc.
You mean I can do this "by eye"?
No, you can do it using Primus from the ATSAS package or a similar program. In Primus use the Scale button.
Measuring the buffer before AND after the sample might help: you mean, I should take the average of those?
At EMBL's X33 we always measure one buffer before the sample and one buffer after. Try to subtract the buffer before, the buffer after, the average of both and check which worked best (at X33 this is done automatically for every measured sample).

Some advices: measure water several times to see if the scattering pattern changes from measurement to measurement.
Since you measure proteins it is a good idea to do radiation damage checks by measuring the same sample twice (without reloading). Compare the two unsubtracted measurements: if they are identical - there is no damage; if they slightly differ in a particular angle range - radiation damage is present; if they differ a lot - there is a normalization problem, talk to your beamline scientist.

Post Reply