Single-molecule spectroscopy has proven to be an ideal tool of investigating protein-protein interactions by (according to protein folding methods described previously) applying the Bell model (or several derivations) to a multidimensional energy landscape which includes the protein bound to its ligand as well as the unbound state.
A model target of investigation is the two-domain calcium sensor Calmodulin, which binds to more than 300 known target peptide sequences in varying manners. Via molecular cloning techniques we create a construct, where Calmodulin is sandwiched within ddFLN domains (which act as handles for pulling), and where we fuse a target peptide sequence in that manner, that it binds to CaM or one of its two binding domains.
By applying force we can now steer the CaM-peptide complex through the unfolding energy landscape and thus distinguish different binding modes of targeting sequences which enable Calmodulin to work as a very fine-tuned Calcium sensor system in that variable way. [1,2]
Another example of protein-protein interaction occurs in mechanosensitive systems, as for example human Filamin or the Titin-Telethonin complex in the z-disc of the muscle. Here we discovered one of the most stable protein complexes (with forces >700nN) when force is directed in the N-C terminal direction (as it is occuring through muscle stress). We further could observe a strongly direction-dependent stability of this complex, so protein interaction cannot be explained profoundly with only taking into account thermodynamic without considering directional informations. 
1) Junker, Ziegler & Rief, Science, 2009 Jan 30;323(5914):633-7.
2) Junker & Rief, Proc Natl Acad Sci U S A, 2009 Aug 25;106(34):14361-6. Epub 2009 Aug 10.
3) Bertz, Wilmans & Rief, Proc Natl Acad Sci U S A, 2009 August 11; 106(32): 13307–133310.