If you have information on which part of the protein are interacting, and its sequence, you can search for homologous or similar sequences in other proteins. This would give you a list of potential targets and probably a similarity score which you can also use. Then you would have something to run simulations on.
And I think that with 3D structure you should be able to run simulations in silico in this case (eventhough it is probably more common to look for the ligands to the protein, not proteins to the ligand).
In silico simulations are computationally intensive, but I´ve also seen some being run on a laptop, so you don´t necessarily need access to HPC-clusters, or supercomputer.