Side-chain packing with multiple chains

No, Rosetta packs them together. Yes, Rosetta is residue centric, and Rosetta doesn’t care which chain residues are made of (unless your using some specific application like InterfaceAnalyzer). The default score function in the weekly releases is talaris2013. We are currently working on improvements to our main documentation, but tilll then, you can find a description here:

https://www.rosettacommons.org/node/3508

And here is a description of the individual energy terms that make up talaris2013, pulled from the in-progress documentation:

fa_atr – lennard-jones attractive
fa_rep -lennard-jones repulsive
fa_sol -lazaridis-jarplus solvation energy
fa_intra_rep – lennard-jones repulsive between atoms in the same residue
fa_elec – coulombic electrostatic potential with a distance-dependant dielectric
pro_close – proline ring closure energy
hbond_sr_bb – backbone-backbone hbonds close in primary sequence
hbond_lr_bb – backbone-backbone hbonds distant in primary sequence
hbond_bb_sc – sidechain-backbone hydrogen bond energy
hbond_sc – sidechain-sidechain hydrogen bond energy
dslf_fa13 – disulfide geometry potential
rama – ramachandran preferences
omega – omega dihedral in the backbone
fa_dun – internal energy of sidechain rotamers as derived from Dunbrack’s statistics
p_aa_pp – Probability of amino acid at phipsi
ref – reference energy for each amino acid
METHOD_WEIGHTS – not an energy term itself, but the parameters for each amino acid used by the ref energy term

I don’t think you would want to pack each chain individually, but it really depends on your purpose. It depends on the protein itself though too. In a multi-chain protein its really the interface you probablywant to think about in this case – this is why you want to pack them together as the side chains from one chain are interacting with the side-chains of the other. If you only pack one here, you are putting a constraint on that chain due to the fixed positions of the side-chains in the other chain. If your interested in some surface far away, then repacking only that section or only that chain should be ok. It really depends on what you are trying to accomplish and what the PDB is. Is it an X-ray structure or NMR? A homology model or a frame of an MD trajectory? Are you interested in energies, rotamers, hbonds, C-alpha RMSD, or something else entirely?