SAXS profiles & PDB: reasonal fit but very high chi values

Small angle scattering software (except ATSAS)
Post Reply
Message
Author
wai
Member
Posts: 10
Joined: 2011.07.07 10:34
Location: King's College London

SAXS profiles & PDB: reasonal fit but very high chi values

#1 Post by wai » 2013.10.10 17:19

Dear all, I have processed a set of 2 concentrations of a protein data, buffer subtracted and extrapolated to zero concentration. I ran Crysol with the known crystal structure. The fit looks reasonable but the chi is ~8!
crysol-fit.png
Fit of data and PDB file: note high chi value
crysol-fit.png (20 KiB) Viewed 7601 times
I ran Dammif-etc and came up with a nicely converging bead model which I used to make an envelope, and the PDB crystal structure can fit quite nicely inside the shape.

I don't know what is wrong? I also build all the missing residues with modeller and submit the whole ensemble to FoXS. The best model still has a chi of ~5.7. Please can you guys help? (Fig. 3)
Attachments
foxs-fit.png
Fig3: FoXS fitting.
foxs-fit.png (57.13 KiB) Viewed 7601 times

wai
Member
Posts: 10
Joined: 2011.07.07 10:34
Location: King's College London

Re: SAXS profiles & PDB: reasonable fit but very high chi va

#2 Post by wai » 2013.10.10 17:54

I also attach the Dammif envelope + PDB model to show the fit. Thanks.
Attachments
FGH-SAXS.png
FGH-SAXS.png (230.54 KiB) Viewed 7598 times

Alex
Active member
Posts: 692
Joined: 2007.08.09 21:10
Location: Planet Earth

Re: SAXS profiles & PDB: reasonal fit but very high chi valu

#3 Post by Alex » 2013.10.10 18:03

the high discrepancy at high angle may be due to buffer-sample mismatch. Why don't you post unsubtracted data for sample and buffers before and after.
Regardless of the buffer issue, you can try to use constant subtraction in CRYSOL or FoXS. I think in the former constant is more meaningful, in the latter - it is actually like an additional
fitting parameter (without any particular boundaries on its value - as I understood from the paper), which should help you get a very good fit.

HTH,
Alex

wai
Member
Posts: 10
Joined: 2011.07.07 10:34
Location: King's College London

Re: SAXS profiles & PDB: reasonal fit but very high chi valu

#4 Post by wai » 2013.10.15 12:12

Dear Alex,

Thanks for your reply. I'll post the plots after this one. For Crysol and FoXS, I have already played with the parameers and got the best fit, i.e. constant was used in Crysol. The following are raw data with buffer subtracted data at three different concentration.

wai
Member
Posts: 10
Joined: 2011.07.07 10:34
Location: King's College London

Re: SAXS profiles & PDB: reasonal fit but very high chi valu

#5 Post by wai » 2013.10.15 12:17

Here are the data, before and after buffer subtraction for 3 concentrations: "131" is 11 mg/ml, "133" is 5.4 mg/ml, "135" is 2.5 mg/ml. I did not use the lowest concentration as it has a slightly different bend towards a lower I0 than the other two, for scaling and extrapolation to zero concentration.
Attachments
135.png
2.5 mg/ml
135.png (21.88 KiB) Viewed 7572 times
133.png
5.4 mg/ml sample
133.png (17.28 KiB) Viewed 7572 times
131.png
11 mg/ml sample
131.png (16.66 KiB) Viewed 7572 times

HLee
Member
Posts: 11
Joined: 2012.05.04 21:37

Re: SAXS profiles & PDB: reasonal fit but very high chi valu

#6 Post by HLee » 2013.10.23 13:49

In my opinion,

About the first CRYSOL figure,
the big discrepancy between simulated (CRYSOL fit) and experimental data is appeared at the very small q-region.
I think this is the reason why the Chi value is so high, and and it is also the reason why the experimental Rg (RGE) is about 4 A larger than the simulated one (RGT).

In addition, this will be the reason why you get more elongated (larger Dmax) bead model compared to the crystal structure.

Post Reply