Searching For Proteins That Have Similar Type Of Residues In Binding Pocket

Question

[Flow]

I am working with some protein that has 6 polar basic residues on its binding site. I would like to find similar proteins on PDB (or other database) that have this type of residues (polar basic residues (ARG or LYS or HIS)) at similar 3D positions (given some threshold). It would be something similar to ligand pharmacophore similarity search, but for parts of proteins. How can I perform this query?

PS: Given the unavailability of such software, I am thinking about developing a GPU based software that would scan whole PDB database for finding related proteins given user's input. If somebody is interested, we could collaborate. This is serious, please contact me.

Answer 1

[Pawel Szczesny]

In the past I have used SuMo by Martin Jambon (also a BioStar user - Martin, where are you? :) ). It will require some modification of your PDB file so you can easily select which residues you want to take into account (for example by marking required residues as belonging to chain X, etc.). See the server for details.

Alternatively, you could try to download one of many programs designed to analyze protein structures for binding pocket similarity (ten or so are cited in the introduction of this short paper: http://stat.fsu.edu/~jinfeng/publication/TIPSA_ACMBCB2011.pdf ) and then substitute program's database with your own definition of binding pocket. However, I can imagine it's not going to be the most straitforward way.

[Flow]

very interesting, I will have a look, please look at my question's update

Answer 2

[Bilouweb]

Hello,

Your question make me think of my application for 3d motif search in protein structures : RASMOT3D

In rasmot, you give a definition of a reference 3d motif which contains :

• a pdb file
• the residues involved in the motif
• for each residues, the list of equivalents residues.

Given this reference motif, rasmot3D search through the pdb to find similar sites (geometricaly).

The webserver is a light version of the application, if you need, we can collaborate for a more complex search through the PDB.

[Flow]

please give me your email, if possible, thanks

[Bilouweb]

gcollet at ouvaton.org

Answer 3

[Khader Shameer]

I would approach this problem from the perspective of sequence-structure analysis. I will use pattern search using amino acids and 3D spatial search to develop such a tool.

1. I will compile binding site data from public resources of protein binding sites. Both sequence data (residues + structural positions) and structural data (full-PDB files) will be compile in to a database

2. I will apply filtering conditions (for example classification of amino acids (polar residue, hydrophobic, charge etc) to the sequence dataset and get initial list of proteins for structural analysis

3. Once you did the filtering, you can define a distance shell around c-beta atoms and compute distances between active site residues. You can use K-d tree or similar data structure for implementing such a program on CUDA-type architectures.
   

4. You may use such distance shells to define and identify related structures. I think active site residues will be close (<= 7 Angstroms in distance).

I have used similar concepts of distance-shell searches to define higher order residue interactions in protein structures.

PS. How to contact you ? No contact details in your profile :) !

[Flow]

werner.schroedinger at googlemail dot com

[Khader Shameer]

Thanks flow, you've got a mail.