Note that pdb codes refer to cristal structures and not to proteins. For example, some pdb files contain more than one protein. Plus, there it may be more than one pdb id for the same protein. So, it's better if you use the Uniprot codes directly, they should be included in the pdb file.

To get all the known interactions for a certain protein, there are many databases:

• STRING
• GeneMania
• Mint

It is important to note that these databases return predictions based on the current annotations, and that they may contain a lot of false positives or negatives. You may want to manually check at least some of the interactions found, and read the literature referred in each prediction. It will take some time, but may save you from working on a wrong network.

Protein-protein interaction data compiled in the resources mentioned by Giovanni will be a good start. But the interaction data in those databases are derived using multiple methods that includes experimentally validated interactions and computational approaches (text-mining, database search). Such database also provides interactions based on co-expressions and cellular co-localization. You may refer to their original papers for full list of experimental or computational methods. As you are looking at structural interactants of your protein of interests you may look at subset of experimentally validated interactions.

I would recommend the following databases that utilize structural information to derive protein-protein interaction

• iPfam
• 3DID

For protein-complex based interaction data, you can use:

• CORUM

You have to ID mapping using SIFTS to map all interactions to a common ID space (say interactions in PDB level or in level of conserved protein domain in each structure etc.)