I am using HADDOCK 2.4 to dock a protein (containing 9 Cysteines coordinated with 3 Zn ions) with DNA. I am doing this for two proteins one for which DNA bound crystal structure is available and the other is the homologous (80% of sequence similarities) protein for which I want to predict DNA bound structure through docking. Even though the same DNA sequence is used while docking in both the cases, I find surprising that the docked model generated for the protein whose crystal structure is available docks perfectly as in crystal structure. Whereas, in case of the homologous protein, protein docks on the DNA in quite different orientations each time I give it for docking. Please note that I am not applying any restraints while docking in both the cases. Also, I have seen that on mutating the active site residues of the homologous protein to that of the protein with crystal structure, in the docked structure DNA binding still happens in different orientations each time I give it for docking. Hence, the question here is that does HADDOCK works by fetching information from the existing data bases about the protein structure and therefore docks protein and DNA in exactly similar manner as it is in the crystal structure? And, on the contrary, when a prior information about the pdb structure is not available HADDOCK works without taking any help from structure data base and hence binding orientations are different each time we run different HADDOCK runs (without applying any restraints) ?
HADDOCK does not fetch any external information but only uses what you are giving it.
In your case, when using the crystal structure, you effectively have a perfect case without any conformational changes.
It is therefore easier to find the “correct” solution.
You might still get multiple clusters with different orientations though.
Again always consider multiple clusters.
Docking to DNA without information is not an easy problem…