Docking 3 large proteins on local haddock3

Hello there!

I was just trying to dock 3 proteins (1100 - 2300 aa in size, actually they are forming protein complex) and stuck editing cfg files. I have a local version of haddock3. I was hoping that someone will give a bit of advice or protocol for this kind of docking.
What I have:
Core i5 with 6 cores
3 pdb files from AlphaFold
WSL on windows

Also, I was interested in if pdbs need to be processed for example in FoldX repairPDB tools or something that will decrease the Gibbs energy. Will that better influence the docking procedure?

Look at the docking-protein-homotrimer example coming with haddock3, but in your case without symmetry and NCS restraints.

Do you have any information about possible binding sites?

Make sure the PDBs have chainID (and different ones). Remove all the “junk” (disordered regions) from the AlphaFold models and validate/clean the PDB with our pdb-tools (also installed with haddock3)

Thank you for pointing to this example. While reading, I felt that the pipeline is for symmetrical proteins, when in my case complex is asymmetrical.

Unfortunately, I only have wet biology truncated proteins experiment, where NBAS protein interacts with ZW10 and ZW10 with RINT1, but RINT1 and NBAS are not showing any signs of interaction. How do I apply this info to cfg file?

Thank you for your advice on AF models! Definitely going to use it!

Thank you for pointing to this example. While reading, I felt that the pipeline is for symmetrical proteins, when in my case complex is asymmetrical.

Which is why I explained to remove the symmetry-related parts…

Unfortunately, I only have wet biology truncated proteins experiment, where NBAS protein interacts with ZW10 and ZW10 with RINT1, but RINT1 and NBAS are not showing any signs of interaction. How do I apply this info to cfg file?

You can only do that in the definition of your restraints. If you have none, then this will be tricky.
In that case in principle you could model the two interactions separately and then combine the results, excluding obviously any models resulting in overlapping regions on ZW10