I was recently running some small molecule dockings (template-based approach) and encountered a problem and I was wondering if someone could provide some help.
I am running Haddock 2.4 on a local machine.
At first, I generated the ligand (all-atom) parameter files with the web server of Prodrg. Here, I submitted two conformations of a compound, i.e. cis and trans conformation. However, I noticed that the web sever was only able to generate correct parameter files for one conformation (trans-conformation). For the cis conformation, the PRODRG server changes the ligand structure to trans. Unfortunately, I was not able to solve this problem (despite testing different pdb and mol2 files of the ligands). Thus, I decided to switch to ATB (all-atom) parameter files, as suggested in the HADDOCK website (Small molecule docking – Bonvin Lab). This parameterization scheme was able to provide correct parameters for both cis and trans conformations of the ligand and therefore I continued with those files. However, when Haddock generated the initial 200 poses I noticed that a different part of the molecule had a “weird” stereochemistry (one of the H atoms on a six-membered ring is axial instead of equatorial). I checked the initial ligand pdb file and the angles/dihedrals involving this H atom in the toppar files and they are correct. I also tested different initial placements of the ligand, in case there was a protein-ligand clash, and the problem with the H placement persists. Moreover, dockings for similar ligands (containing the same six-membered ring) did not show this problem.
If something is unclear or if I should provide some files please let me know.
PS: Note that if you have the topology for the trans form, it should be rather simple to edit it to create a cis conformation topology (a question of changing one angle)
sorry for the late response.
Indeed changing this parameter to “unknown” works. The ligand has a valid conformation.
However, once turning this option to “unknown” the protein seems a bit unfolded.
I ran several tests and noticed that if I change “delenph=false” (which I was using in my docking runs) back to “delenph=true” everything works fine again.
Might there be a solution leaving “delenph=false” since I wanted to follow the protocol mention in: Protein–ligand pose and affinity prediction: Lessons from D3R Grand Challenge 3 | SpringerLink
Indeed changing this parameter to “unknown” works. The ligand has a valid conformation.
Does it keep that conformation after the flexible refinement?
However, once turning this option to “unknown” the protein seems a bit unfolded.
I would not think this should cause unfolding.
I ran several tests and noticed that if I change “delenph=false” (which I was using in my docking runs) back to “delenph=true” everything works fine again.
HADDOCK uses a united atom force field with by default only polar hydrogens kept. So if this works for you it should be fine.
In order to comprehend the unfolding of the protein structure I compared two pdb files of the protein, i.e. protein structure files in the begin folder of a docking that worked against a one that did not work.
Here I noticed differences within the occupancy column of the respective pdb files.
For instance:
Docking options: delenph=false; hydrogen_build=unknown → unfolded protein structure
ATOM 2 HN THR A 200 -16.937 -0.345 20.526 0.00 15.00 A H
and
Docking options: delenph=false; hydrogen_build=all → expected protein structure
ATOM 2 HN THR A 200 -14.057 0.469 19.968 1.00 15.00 A H
I made a few test runs and it turned out that if I provide a protonated protein structure everything works fine even with the option hydrogen_build=unknown.