Docking an atomic protein computer model in SAXS

Calculation of SAXS and SANS profiles (CRYSOL, CRYSON), superposition of models (SUPCOMB, DAMAVER, DAMCLUST), database DARA
Post Reply
Message
Author
Nyshae
Active member
Posts: 24
Joined: 2015.11.03 14:38

Docking an atomic protein computer model in SAXS

#1 Post by Nyshae » 2015.11.03 15:04

Hi everyone!

I am working with this protein and trying to get an idea of its shape. A model has been generated by computing various parameters and the protein has also been measured in SAXS. When I fit the model vs the SAXS data (with CRYSOL) I get a fit that is not optimal but not horrible either. Chi2 value is around 40-50 which I think is way too high, right ? However, the model visually fits really nicely in the SAXS DAMMIF model in PyMol...

So my questions are:

a) When I do primary analysis with Primus, the I(0) and Rg are quite consistent between Guinier and P(r). However, the value of I(0) is way too small (we expect 3 times that). But we get an ok value from the Porod volume, is that possible ? (the protein is quite elongated) How to know when the data would be acceptable by the SAXS community (from statistics point of view)?

b) Is there an automated way to refine the computer model to make it fit better the SAXS data ? Or, considering that it is already a computer model and not a crystallographic structure, is the fit considered ok ?

c) The protein is constituted of sub-domains A and B. The computer model is much better for A. We measured the SAXS data only for AB protein. Is there a way to fit the A part only in "half" of the SAXS model (other than manually with PyMol)?

d) Is it possible to turn the bead model (obtained from Dammif) into a "sequence model" like these guys have done (http://www.rcsb.org/pdb/explore/explore ... ureId=3GAU)?

Sorry for the long post and thanks a lot for any input you might give me ;)
Cheers !

ckerr
Active member
Posts: 86
Joined: 2015.03.25 09:03
Location: EMBL Hamburg

Re: Docking an atomic protein computer model in SAXS

#2 Post by ckerr » 2015.11.10 10:12

a) Not sure about this one

b) That depends on your model. If by 'model' you mean 'protein structure' then you could try using SREFLEX (part of ATSAS 2.7).

c) I'm not sure if I understood your question properly, but if I did then the answer is: Of course not! There is no way for dammif to know, when it is generating a bead model, which part is domain A and which part is domain B.

d) Looking at the paper for this structure, the authors already had crystal structures for all the domains and for several groups of two or three domains. They used SAXS and MD to come up with possible structures for the whole protein. There is no way to transform a bead model into an atomic resolution structure.

Nyshae
Active member
Posts: 24
Joined: 2015.11.03 14:38

Re: Docking an atomic protein computer model in SAXS

#3 Post by Nyshae » 2015.11.10 12:36

Hi,
Thanks a lot for your answers !
For question b), the model in question is purely computational (based also on literature references).

ckerr
Active member
Posts: 86
Joined: 2015.03.25 09:03
Location: EMBL Hamburg

Re: Docking an atomic protein computer model in SAXS

#4 Post by ckerr » 2015.11.11 10:12

What I meant was: does the model represent the protein as a sequence of amino acid residues with their positions in 3D space (in which case SREFLEX can be used) or does it use a simplified representation e.g. sphere, cylinder, SASHA envelope, DAMMIF dummy atom model etc.?

Nyshae
Active member
Posts: 24
Joined: 2015.11.03 14:38

Re: Docking an atomic protein computer model in SAXS

#5 Post by Nyshae » 2015.11.11 10:33

Sorry for the misunderstanding. Yes, the model represents a protein 3D structure with amino acids positions. So yes, I could try SREFLEX ! I just had the older version of ATSAS so didn't see that.
Thanks again !

Post Reply