Multi-body interface for structure refinement

I would like kindly to ask which parameters should be changed in the multi-body interface of HADDOCK2.2 in order to perform water refinement and energy calculations of a trimeric protein (but not the whole docking process). I want to use multi-body interface and not the refinement interface, since I have a trimer and want to get energetics taking in account all three chains simultaneously.

Best regards,

Why don’t you simply try our new 2.4 refinement interface? It should do the trick.

Note that it is still in development and only supports proteins at this time.
You will need to upload a single PDB file with all three chains.
It should however have a MODEL / ENDMLD statement in it (not per chain, but per structure).

And otherwise a recipe for this can be found at:

And similarly for the 2.4 server:

Thank you very much for the advice and information.


I have run a few tests for water refinements using HADDOCK2.4 refinement and docking (submit) interfaces and I have noticed a few things which I don’t understand and would be very grateful for an explanation (I have an ZN atom in some of my structures, therefore I can’t use only the refinement interface of HADDOCK2.4 since heteroatoms are not supported there):

  1. results after refinement and docking interfaces for the same structures differ from each other, sometimes even very drastically, although I though the values should be very similar or the same, since only water refinement step is performed in both cases for the same initial structure:
    e.g. HADDOCK score of ~4470 a.u. and ~2110 a.u. after refinement with refinement and submit interfaces respectively, but for example desolvation value has almost the same value of ~-320 kJ/mol I both cases.

  2. The final structures are sometimes clustered into two clusters after the docking (submit) refinement, whereas I have never seen that using refinement interface. Is it to expect or something goes wrong during the refinement performed using docking (submit) interface?

The structure refinement with submit interface was used like is described here: , putting the number of structures for rigid body docking, for semi-flexible refinement and for the final refinement to 20 and number of trials for rigid body minimisation to 5 and later to 1 (since as I understood the refinement interface has one trial for rigid body minimization).

Thank you very much,

Can you please share the links to your two runs?

You can in principle compare the json files to see if the settings are similar or not.

  1. Here is the refinement interface:
    Here is the submit interface
  2. refinement without ZN and no clustering
    Submit with ZN (but the same structure) and clustering

Comparing the json files of those two runs, the molecules are entered in a different order.
This might indeed affect the results since you are adding up numbers in a different way (the chaotic nature of such computations).

The main difference seems in the van der waals energy which remains positive. You must still have clashes.

For clash removal, you could try to coarse grain the molecules, still running the water refinement at the end. This is a protocol that works quite well.

We used it in our recent manuscript to refine docking models from LightDock:

Note that we should have very soon a fully functional refinement interface for all molecule types supported by HADDOCK.
And it will even allow refinement of single chains!

Thank you very much for your help and advises, I will try to coarse-grain the structures and submit chains in the same order as it is done in refiniment interface.

Dear Marija

We have updated the refinement interface of the 2.4 server. It can now handle all different types of molecules that HADDOCK supports including ions. In case you are dealing with a zinc finger, make sure to change the coordinating CYS names to CYF (i.e. without the H atom).

And as new feature it can also refine single molecules (i.e. does not have to be per se a complex).

1 Like