The HADDOCK category is meant to discuss any HADDOCK-related issue. For general information about HADDOCK refer to HADDOCK – Bonvin Lab
Dear community, I am seeking to determine/model a protein-small molecule complex for a very flexible protein. Most structural approaches have not been successful, but I have experimental thermodynamics for 11 alanine mutants of the expected binding site, as well as for 6 chemical analogues of the small molecule and HDX-MS data. It turns out, the ab initio alphafold 3 model from the protein sequence and the ligand SMILES is completely consistent with the biochemical data.
Dr. Bonvin generated a model based on the alphafold3 structure I sent him and results from a virtual alanine scanning. The results from the ala scan very closely recapitulate the experimental data. However, I would like to generate a model of the complex by directly inputting my experimental data to generate restraints. Looking at the tutorial for protein-small molecule docking, it seems this is somewhat complex. What I am wondering is, given that I already have the ligand in the proper binding site from the AF3 model, in the right orientation, can I bypass the initial rigid body steps and run a restraint-based energy minimization on the AF3 model?
I would appreciate if anyone can point me in the right direction for this, thanks!
Yes, you can bypass the rigid-body docking and perform an energy minimization.
On the HADDOCK2.4 web-portal, you can click on “Refinement” in the menu, and supply your input file. Unfortunately, this does not allow you to add the restraints.
But, if you use HADDOCK3 (local version), you will be able to perform exactly what you need, by running this configuration file:
Alternatively, you can also use the flexref module to perform a flexible refinement, which would allow for greater conformational changes of your ligand
run_dir = “minimize_af3”
molecules = [“your_af3_model.pdb”]
[topoaa]ligand_param_fname = “parameters_for_your_ligand.par“
ligand_top_fname = “toplogy_for_your_ligand.top“
[flexref]
# here the parameter sampling_factor set to 10 allows
# to perform 10 flexible refinements per input molecule
sampling_factor = 10
ligand_param_fname = “parameters_for_your_ligand.par“
ligand_top_fname = “toplogy_for_your_ligand.top“
ambig_fname = “restraints.tbl“
unfortunately, using HADDOCK3 requires you to:
have it installed locally (not too difficult)
have both parameter and topology files for your ligand, which is not easy. But I can do it for you if you send me your ligand (or multiple ligands if you want to try this approach of the 6 analogues too)
You can send me by email your ligands as PDB files.
Maybe you can give me your thoughts on the most straightforward way to tackle this modeling project. My PI prefers to perform all docking steps in one algorithm, if feasible, meaning it would be preferable to find a way to do initial docking and refinement steps in HADDOCK rather than starting with AF3 (Protenix) modeling, followed by HADDOCK refinement. However, the binding site of the molecule is difficult to access in rigid body due to a highly flexible “gating” element.
Goals: 1) Model protein-ligand complex using my experimental data as restraints, preferably at all stages of docking and refinement and using only HADDOCK, 2) Refine the model to find most likely protonation states and side chain/ligand conformations, 3) Generate ala scanning energetics for model validation against experimental results. I may be interested in doing this for other compounds as well, but for now just focusing on the one.
It seems my options are:
Dock complex with experimental restraints using Haddock 2.4. My colleague says I can enter advanced option to use intial flexible docking rather than rigid body, though he has only used haddock for protein-protein. It seems that I will need to obtain ligand topology and bring in HADDOCK 3.0 to generate energetics for ala mutations, but this should bypass the need to use special script to enter restraints the refinement stage. I would need guru permissions for this approach I believe.
Start with AF3-based Protenix server to generate initial model as described in OP (there is ability to specify interatomic distance constraints). Then move to HADDOCK 3.0 for all refinement steps, using your script to enter restraints at this stage and enter ligand topology.
Which do you think will be the simplest solution here?
Indeed within guru access in Haddock2.4, you can perform everything mentioned, with the exception of the Alanine scanning. But the parameters/topology are generated on the fly when you submit to the webserver, simplifying one step.
On the other hand, with haddock3, you can perform the alanine scanning within the exact same workflow.
Finally, entering a buried binding pocket is not an issue in HADDOCK, as it requires simply to reduce the VdW component in the force-field during the rigid-body docking to be able to “penetrate” through the protein and reach the binding site (parameter inter_rigid = 0.001 in [rigidbody] module in HADDOCK3). Please have a look at the tutorial we have about protien-ligand docking:
This approach is valid too, as obtaining a bound-like conformation of the protein, especially if it undergoes big conformational changes, is good to have.
Bur requires to have the topology and parameter files for your ligand
3. My suggestion
Perform the initial docking with AF3-based Protenix sever to obtain a bound-like protein structure
Use the conformation of the protein as input in HADDOCK3
Use your experimental restraints to perform the integrative-modelling of you complex of interest in HADDOCK3
Terminate the workflow with the [alascan] module, making sure you select the residues of interest.
obtaining the parameters and topology of the ligand (I can do it for you if you send your ligand by email v.g.p.reys[at]uu.nl)
Conclusion
I see that the tools/scripts you need here are not fully integrated into HADDOCK3 yet, but only sparse script that we have in-house, making it tedious.
This, once again, gives me the idea/will/energy of spending some time to integrate all the required tools, for you and any other user willing to perform protein-ligand docking), in HADDOCK3 directly.