I am using Rosetta local refinement to calculate the interface scores of known protein complexes (in a bound form).
In the docking protocol, it says "Typical values for I_sc of godd decoys are in the range of -5 to -10.". However, for some complexes, rosetta local refinement gives me I_sc as low as -5988.1. This number doesn't seem realistic to me. Could you please explain why we see such small numbers? How can we interpret this result, if the good binding score is in the range of -5 and -10?
As an example, I tried local refinement for D and F chains of 1n86.pdb (after prepacking) and it gave me I_sc of -5988.1. You can find the pdb file (with only D and F chains) attached.
I would be very glad if you could help me.
You have a disulfide between the two chains. (Chain D 161 and Chain F 135).
When you move the two chains apart, Rosetta still thinks there is supposed to be a disulfide between the two residues. As there's some 500+ Angstrom between the two sulfurs, though, the separated complex scores really poorly. Thus when you do the [together-apart] calculation for the interface energy, the really positive 'apart' energy makes the interface score really negative.
Rosetta doesn't currently have a good way to turn off disulfides specifically in the apo context (and the energies thus obtained wouldn't necessarily be useful anyway), so your best bet is to calculate the interface energy of the non-disulfide bonded form. (So no disulfides in either the bound or unbound state.) You should be able to turn off the disulfides by passing the option `-detect_disulf false` - Note that this will turn off all disulfides, not just the ones in the interface.