Structure Determination Of Protein Domain

Question

[Shweta]

I am trying to determine the structure of one domain of a multi-domain protein. This is what I have done so far: 1. Extracted the fasta sequence of the domain from uniprot. 2. Ran a psi-blast search. 3. Selected around 13 spp. from the blast results. 4. Meanwhile, searched the domain in pfam; selected around 18 sequences from there and performed MSA using clustalw.

I have two queries (as of now): 1. I want to know whether I have proceeded correctly so far; if so, then what next should I be doing. 2. The psi-blast results that I got show the sequence of the entire protein (despite me entering the query as the domain only). Hence, I want to know why this is showing up; and how I should chop the sequence correctly (i.e., to retrieve only the domain of interest).

Pretty long question, this, but hope I made sense.

Hoping for a reply soon. Thanks in advance

[Pals]

Did you extract the sequence of the domain of your interest or the multi-domain protein from UniProt?

[Pals]

Does it mean that you extracted the sequence for the domain of your interest or for the multi-domain protein from UniProt?? And do you want to model the domain structure?

[Shweta]

During all my searches, I have used the sequence of the domain of interest only, not the entire protein. Yes, I want to model the structure of the domain only

Answer 1

[Pals]

IMHAO you are in the right track. When you have the homolog for your domain, you can make an alignment with the best homolog so that you can see which residues in the template correspond the residues of the domain of your interest. After knowing this, you can edit the PDB file in order to remove the residues that are not of your interest. Finally, you can use that edited structure to model the domain using modeling programs (Modeller, SWISS-MODEL or ITASSER).

[Shweta]

Hi, thanks for replying.I did an Emboss Matcher local pairwise alignment, as per your suggestion. The results show identity as 26.8% and similarity as 46.4% with 7.1% gaps. What is your opinion on this? Also, regarding you mentioning about editing the pdb file, could you elaborate some more on that front, I'm unsure as to how I should be doing it

[Pals]

The alignment score tells that you can make the model but you need to be careful as the identity score is below the twilight zone. I suggested you to edit PDB file if the template contains other chains in addition to the one containing the domain of your interest. For example, if the template contains chain A, B and C and if chain A contains the domain of your interest, you can remove other chains and use only chain A to make model. You can delete those chains and hetatms using text editor (just delete the portion that is not of your interest).

[Shweta]

I tried to visualize the template A chain in Discovery Studio. There, I noticed that the first 30 residues are missing there as compared to the Uniprot sequence. Is this something to worry about?

The template that I'm using is actually a fusion of two domains; one, the domain of my interest, and the other is the adjacent domain. How should I proceed now? I'm confused!

[Pals]

It is not surprising to find the missing residues in terminal regions so you don't need to worry about that. Now you have two choices- either you can cut all residues in template that do not match the domain of your interest and use it in "Build Homology" protocol, or if you think you would not have problem to have another domain in your model, you can model using the whole chain. Also, it is possible that you remove another domain after generating the model. You should find it easy to use MODELLER's features within Discovery Studio.

[Shweta]

But if the domain of interest is a part of the template, why is the homology so poor? The sequences too are hardly matching

[Pals]

The results show identity as 26.8% and similarity as 46.4% with 7.1% gaps. What does this statement in your first comment mean then?

Answer 2

[DG]

Assuming you mean the 3D structure of the domain you want to see if any solved structures exist for proteins that contain that domain. For that you want to check in the PDB database. Both PFAM and Uniprot sequence records usually have information on any associated structures.

You can also use any of the sequences you have already downloaded and run a regular BLAST query against the PDB to see if any homologs have had their structure determined. You can then get the relevant file(s).

[Shweta]

Hi Dan, thanks for replying. I first did try with the regular blast, but didn't get good results, so went for psi-blast. I have also tried with pdb blast. Pdb blast, swiss-modeller and i-tasser all showed the same template, so I was happy with that