How to predict Protein-Protein Interactions from a pair of gene symbols?

Question

I have a list of pairs of gene symbols who interact (putatively) and would like to assign each pair a score (and record other details) based on the predicted Protein-Protein Interaction (PPI). The existing PPI webservers I've looked at (ClusPro and HADDOCK) require that I use PDB ID's as input. Unfortunately, PDB ID's don't always exist for the gene symbols I'd like to compare. I believe I need either:

1. a way to generate PDB files myself (perhaps with a folding program?) and then connect/upload them to a PPI webserver.

or

2. a way to find the most similar sequence that does have a PDB ID and upload it to a PPI webserver.

Am I understanding my problem correctly? Are there resources that can help with 1 or 2?

Answer 1

This may or may not be possible, depending on what proteins you are considering. Generating a PDB file means predicting the structure of the protein. There are no methods for predicting protein folding accurately from plain sequence data, so you will need some experimental data on the structure of your proteins.

If your proteins have not been structure determined, the next best thing is some kind of homology model, that is, a prediction of your protein structure based on a known structure of a homologous protein. These can be more or less accurate, depending on how close the homolog protein is. If specific domains are interacting, it may be enough to create a model of the domain only.

There are online tools available for homology modeling, for example swiss-model from the expasy site. This is not an automated process though --- creating an accurate model requires some skills. You should read the publications describing these tools and make sure you understand the process and error souces, and perhaps contact an expert if in doubt. Good luck!