One of the method is to cross-linking followed by LS-MS/MS analysis. Cross-linked your protein of interest with DSS (arm length around 11.4 angstroms and interact with lysines only). You can use DMA and DMS as well. Now perform LS-MS/MS of cross-linked and non-cross-linked samples and look for peptide which are exclusively present in non cross-linked samples. Those are the amino acids which were responsible for self-association.

I agree with Zeyaul Islam that a crosslinking experiment would be valuable. However, I would point out that although the experiment will identify residues that are close together, those residues are not necessarily responsible for the protein-protein interaction. Moreover, such an experiment is not guaranteed to identify residues that are close together because it relies on the presence of 2 primary amino groups (lysine side chains or N-termini) in close proximity. Finally, there are additional crosslinkers that can be considered with different linker lengths and different chemistries. I refer you to this web site:

https://www.thermofisher.com/us/en/home/life-science/protein-biology/protein-labeling-crosslinking/protein-crosslinking.html

Hi Liyana, yes, a cross-linking experiment would work, as outlined above. We wrote software (Byonic) that expressly searches for X-linked pairs, as suggested by Adam above.

But there are more experiments: if you have SEC separations, you can separate the fractions to see which of those peaks contain 'n' multimers. Usually by Intact Mass LCMS (a relatively simple experiment)

Secondly, we also work with the Uni of Washington who are using our 'Footprint' software to work out which parts of the protein are solvent-exposed by using 'oxidative footprinting' followed by mass spec, and therefore which parts of the protein are NOT attached to each other. The advantage is that you can also work out the relative 'strength' of the interactions (but we would hesitate to translate that immediately into binding information without other info)