Hello, I'm interested in determining which residues in a protein are most likely to be involved in protein-protein interaction. Does anyone know of any tools/methods to gauge the likelihood that a surface residue will, or will not, be involved in protein-protein interactions, to any degree?
I have a list of structure based protein-protein interaction site predictors:
- http://pipe.scs.fsu.edu/meta-ppisp.html
- http://haddock.chem.uu.nl/services/CPORT/
- http://pipe.scs.fsu.edu/ppisp.html
- http://sparks.informatics.iupui.edu/PINUP/
- http://bioinfo.weizmann.ac.il/promate/
- http://bmbpcu36.leeds.ac.uk/ppi_pred/
- http://sppider.cchmc.org/
The first and the second are metaservers using some of the ones listed below. It will give you also per-residue scores.
As for sequence based you must search yourself, I don't know them, I don't trust them. You'd better try to construct 3D model first, if you don't have a structure.
The answer to this kind of question is always "yes, there are lots". Here is some more general advice.
I suggest that you start by searching the literature (either PubMed or bioinformatics journals) and the web (Google), with some appropriate keywords. This will make it apparent to you just how much existing literature exists on this topic.
There are tools which require protein structures as input (e.g. the KFC server) and others which make predictions based on sequence alone (e.g. ISIS, see also accompanying publication). These are just two examples from many: see, for example, a web search results for "protein-protein interaction web server".
To narrow down your options, consider the following:
- what input data do I have (sequence, structures or both) ?
- what kind of output do I need ?
- do I need a user-friendly tool (e.g. web server) or can I install local software ?
- is the tool developed and actively maintained by a group with a good reputation in structural bioinformatics ?
- have they performed benchmarking and validation of their predictions ?
The last point is very important and should be reported in any publication that describes the tool.
For another list of protein binding site (interface) prediction servers (partially overlapping) check-out: http://rosettadesigngroup.com/blog/58/10-protein-protein-interface-prediction-servers/
I think it depends if the protein protein interaction is between two domains or a domain and a smaller ligand. For the first you would need good 3D structures before using the tools above but in the second case you only need a structure for the domain, and then you could use http://pepsite.russelllab.org/ and ANCHOR (http://anchor.enzim.hu/) for guessing the ligand residues.