Two docking runs for the same protein complex with active sites that differ in their length. why does HADDOCK score differs?

Hello!
I made two docking runs between the same two proteins using the easy interface and writing different active sites for the two docking runs. I changed only the active site of one of the two proteins. The active site differ in length (4 a.a vs 11 a.a).
These are the results for the best cluster:

1- for the 4 a.a run:

HADDOCK score -143.5 +/- 4.4
Cluster size 145
RMSD from the overall lowest-energy structure 2.9 +/- 2.6
Van der Waals energy -44.6 +/- 4.3
Electrostatic energy -63.0 +/- 12.9
Desolvation energy -88.5 +/- 2.2
Restraints violation energy 23.0 +/- 13.19
Buried Surface Area 1241.2 +/- 93.1
Z-Score -1.5

2- for the 11 a.a run:

HADDOCK score -121.3 +/- 7.2
Cluster size 7
RMSD from the overall lowest-energy structure 21.4 +/- 0.4
Van der Waals energy -74.7 +/- 5.8
Electrostatic energy -214.5 +/- 21.4
Desolvation energy -13.9 +/- 12.5
Restraints violation energy 101.9 +/- 31.12
Buried Surface Area 2221.1 +/- 87.0
Z-Score -1.1

My question is: Why does the HADDOCK score differ between them? and another question, How can I compare betweent them (compare the numbers directly or what)?

Thanks in advance.

I made two docking runs between the same two proteins using the easy interface and writing different active sites for the two docking runs. I changed only the active site of one of the two proteins. The active site
differ in length (4 a.a vs 11 a.a).

These are the results for the best cluster:

My question is: Why does the HADDOCK score differ between them? and another question, How can I compare betweent them (compare the numbers directly or what)?

Different definitions of the interface will lead to different results indeed since the restraints are different.
You can see that the 11aa has a higher restraint energy.

If you want to compare different clusters from the two runs, you could submit the top model of each to the refinement interface.

In that way the interface will be further refined without restraints and you can directly compare the scores.

Hi Dear @amjjbonvin

According to this part of your answer “If you want to compare different clusters from the two runs, you could submit the top model of each to the refinement interface.”
Is it possible to compare the score of the best cluster result of two runs by following steps 1- haddock docking 2- refinement interface 3- quantitative comparison between the results ?

Because I need to compare docking score of two runs, one haddock run is protein-RNA and another is protein-DNA (protein is the same)? .
Thank you in advance,

In your case since the protein is constant, this might work, but remember that score is not equal to binding affinity…