I want to dock two proteins, A and B. I have pdb files for both of them, A.pdb and B.pdb. Now, A.pdb is an X-ray structure, while B.pdb is an NMR ensemble. Is it possible to run a docking simulation with these inputs? How can I do it?
How exactly do you want to use the NMR ensemble in your docking?
If it's just running multiple separate runs with each of the entries in the NMR ensemble, that is certainly possible. You would just need to separate out the different ensemble members into their own files, and then (manually) launch runs with each of them as a starting structure.
If you want Rosetta to automatically switch/sample between the different members of the ensemble during the run, that's a little more difficult. There is a protocol called EnsembleDock in Rosetta, but it's not particularly well-used (or well-documented) - I'm also not sure that it's *exactly* what you want to use. Take a look at the papers http://www.ncbi.nlm.nih.gov/pubmed/18640688 and http://www.ncbi.nlm.nih.gov/pubmed/20535822 for more information.
At any rate, there shouldn't be an issue with using NMR models as Rosetta input, or even a mixed model with some parts X-ray and some NMR. The one possible concern is that sometimes NMR ensembles have structural deficiencies which X-ray structures typically don't. These can normally be remedied by running relax on the structures before using them in Rosetta the first time. See http://www.ncbi.nlm.nih.gov/pubmed/23565140 for one such protocol. For NMR models in particular, you might want to relax the strength of the constraints to allow more movement. Also, a particular problem of NMR models is bad backbone angles. Relax includes a "-ramady" flag (pun intended) which can help fix up backbone angles if your structures have particularly bad rama scores.
I was wondering if there was a published protocol to do this.
A related question is: Is there a way to convert an NMR ensemble into a "best" structure? Then I can use that structure for the docking.