I am looking for a solution to the following problem. Any insight will be greatly appreciated. A membrane receptor (A) has an extracellular domain (AE), transmembrane domain (AT) and intracellular domain (AI).
A bacteria (B) binds to (AE) leading to dimerization of (A) at (AT) and subsequent downstream signaling through (AI).
(A) has no known natural ligands.
(A) has one known inhibitor (I) binding to (AE)
The structure of (AE) bound to (I) is available from PDB.
How do we identify which protein in (B) binds to (AE).
The obvious solution is to dock all proteins with known structures in (B) to all known pockets in AE. Any better solutions?
Labels: docking, protein, structure
I am looking for a solution to the following problem. Any insight will be greatly appreciated.
13 Comments:
An informative reply I got from bionet.molbio.proteins
Not going to work. Without extra biochemical info on
the interaction interface, protein docking does not work.
And even if it did, what you are proposing woul take forever.
A better solution is to approach this as what it is - purely
experimental problem. Use zero length cross-linker to
link A to B, pull down A, digest and identify bacterial
peptides. This will likely result in a short list of potential
candidates, which can be expressed and their binding
to the receptor measured.
DK
This post has been removed by the author.
Reply 2:
have you tried to do something funtional? like an IP of receptor A,
incubated with bacteria or even bacteria proteins and then do mass
spectrometry to determine the receptor A ligand?
Reply 3:
The inhibitor (I) that you are talkin about , is it a peptide?
if yes then u could go for those proteins or peptides having similar
sequence. assuming that they are homologous you can dock the bacterial
protein.
Thats what i think.
Satish Balakrishnan
If I is a peptide as suggested by Satish, you'd have a more sensitive than search whole-sequence similarity by using InterPro domain, motif, regex or HMM across the protein-binding domain.
If I is not a peptide sequence (small molecule inhibitor or otherwise), there's a strong possibility that it isn't bound to the same site on AE as the protein in B, and therefore the AE:I PDB structure won't help too much. There are in silico approaches for predicting binding proteins but they are still slow and unreliable; you're best of using an experimental technique.
Thank you very much Satish & Luke for highly stimulating replies.
(I) is a peptide indeed, but it is a synthetic antibody. Hence Luke’s approach should work. Is there any easy way to separate the protein binding domain of an antibody? What do you think about a structure similarity search like VAST?
Bell
With respect to the above problem, I have done the following.
As (I) is an antibody it has heavy and light chains (IH & IL)
" Opened the structure (A + IH + IL) in Swiss-PDB.
" Applied different colors to (A), (IH) & (IL).
" Selected Chain (A).
" Selected neighboring aminoacids within a distance of 12Å on (IH) & (IL).
" Found 2 regions of more than 7 consecutive aminoacids on (IH) and 1 region on (IL) (Hypervariable binding regions?).
" Other discrete aminoacids within that distance were ignored. (Approximately 4 on IH and 2 on IL).
" Did a BLAST & Domain search against B genome database.
Any comments on the methodology especially on the arbitrary choice of 12Å distance or chain length of 7 aminoacids.
Read the full story here: http://www.gulfdoctor.net/bioblog/2008/01/protein-protein-docking-problem.html
Though I was excited to see a hit from one of the membrane proteins on the list of hits, I have made a fundamental mistake. Due to certain biological reasons, I am pretty sure now that (B) does not bind on the same pocket on (A) as (I). Hence let me redefine the problem:
(A) has two related receptors (A1) and (A3).
(A1) and (A3) have a high affinity ligand (E)
Though A has no high affinity ligands, pockets in (A) may be similar to (A1) and (A3) because of high sequence similarity. This assumption is experimentally validated.
I did a BLAST search of (E) against (B) and got a SINGLE HIT.
(E) has a conserved domain signature in PROSITE.
+-------------------+ +-------------------------+
| | | |
x(4)-C-x(0,48)-C-x(3,12)-C-x(1,70)-C-x(1,6)-C-x(2)-G-a-x(0,21)-G-x(2)-C-x
| | ************************************
+-------------------+
'C': conserved cysteine involved in a disulfide bond.
'G': often conserved glycine
'a': often conserved aromatic amino acid
'*': position of both patterns.
'x': any residue
How do I search (B) genome for this signature?
A = ErbB2 (HER2)
B = M. Leprae. Causative organism of Leprosy
I = Herceptin
E = Epidermal Growth Factor
A-I structure = pdb: 1N8Z
Read More about Leprosy here: http://www.gulfdoctor.net/blog/2007/09/hansens-disease-leprosy.html
Here is the first publication related to this thread:
http://www.ncbi.nlm.nih.gov/pubmed/18583808?dopt=Abstract
The second one is in print in Leprosy review.
Dear Bell
Though I am still naive Bionformatician But would like to share my views. Correct me if I go wrong.
Protein docking softwares dont actually deal with some flexible regions o the proteins. I would therefore suggest you to use MD Simulations instead.
There is also a possibility that the interacting protein B doesnt have a solved structure. Can some program be implimented to find complementary regions at sequence level.
By the way I appreciate your Blog and invite you to mine which is only a few days Old : http://bioinfo-horizon.blogspot.com/.
Suggestions are invited.
Cheers!
Meenakshi
Hi
Thanks for all your comments. I have published the article leading from this thread in leprosy review ( http://leprosy-review.org.uk/ ) in sept 2008 edition. (Freely available online). The title is
335 Schwann cell invasion by M. leprae : the probable Trojan horse B. R. EAPEN
Post a Comment
Links to this post:
Create a Link
<< Home