So I have two proteins - one is reasonably well-characterized in the complex (NMR results, mutagenesis), and the second is not so reasonably characterized (NMR-blind, a few results from mutagenesis). Using only the results we have does not appear to result in convergence. So, we decided to artificially add a few active residues in the less-well characterized protein. We know it’s alpha-helical, so we have an idea of which residues can be added, but the direction from the already known binders is not clear. So we “slid” the second residue and ran different simulations. Is the HADDOCK score from each run (2.4) worth comparing? Or is there a better way to compare results?
In this case, considering your are docking exactly the same system but with different restraints, you might indeed compare the HADDOCK scores.
But you should also check the scores after subtracting the restraint energy (10% of it) from the total HADDOCK score.