This new paper in Nature Methods...
thinks that if we mine the existing data and use all the proteins that have phylogenetic relationships with one another we can get to the answer of who interacts with whom.
The results are impressive. And it is worth noting that even though they didn't even cite the BioPlex resource, 57% of the data points they incorporated came from direct experiments with human proteins.
BTW, InWeb_IM is their resource in the Venn Diagram above, so even if they aren't right about all of them, it is a whole lot more protein-protein interactions for us to look through!
You can access InWeb_IM's nice web interface directly here.
You might not want to directly Google it, cause the site it took me to isn't the one in the paper -- and looks a little scammy!