I am using the relax.linuxgccrelease application to relax a protein complex. I simply send the pdb containing all of the fomed complex to relax.
The score outputted by relax represents the sum of the score of folding plus complex formation (in Rosetta arbitrary units, of course)?
Pretty much. The Rosetta total score of a monomer is usually taken as a(n inexact) proxy for the energy of folding. Also, the standard way to compute interface energy is to calculate the energy of the bound state, separate the interacting partners some large distance (so they no longer interact) and then calculate the score again. This energy is taken to be the binding energy. Or in equations:
complex_total - separated_total = binding
complex_total - ( partnerA_total + partnerB_total ) = binding
complex = partnerA_total + partnerB_total + binding
Stochastic effects and differences in repacking/minimization details means the numbers might not be precisely equal, but that's at least the conceptual idea.