Hello, I have been trying to validate interface energies computed using Rosetta 2.3, but have been having some difficulty obtaining agreement between two expressions that should be equivalent.
I start with a protein with two chains in a single PDB file, ab.pdb.
Then I run Rosetta++ with the following flags:
-interface -ddg_bind_only -s ab.pdb -intout ab.txt -output_structure -Wpack_only -relax_unbound false
This gives me three files:
ab_nat_0001.pdb (A chain only)
ab_nat_0002.pdb (B chain only)
ab_nat_0003.pdb (A+B chains together)
along with the following energies:
dG = E(AB) - E(A) - E(B) = (1253.7) - (672.7) - (639.4) = -58.4 final dG
Then I take each of the ab_nat_000*.pdb files, and rescore them using the following flags:
-s ab_nat_0001.pdb -score -fa_input -intout score.txt -overwrite -scorefile a.score
This time I obtain
E(AB) - E(A) - E(B) = (1253.7) - (669.5) - (634.4) = -50.3
which is different by 8 kcal/mol, because the energies of A and B are now different when scored separately.
The inconsistency persists despite trying many changes:
-- using softened potentials
-- relaxing structure beforehand
-- setting relax_unbound to false
Does anyone have any idea what could be causing these energy differences? Thanks!