I am trying to estimate theses energies without the entropy component as a zero order approximation.

I suppose (in the rigid body approxumation) I could also select the atoms of both proteins and then ask for an interaction energy. This would avoid having to split the system and generating the energy difference a priori.