I only have the tertiary structure of two proteins predicted by Alphafold. Can I make any docking with them?

Question

[Enrico]

I am working with functional genomics to validate the interaction between a protein of a fungus with Arabidopsis chitinase. I would like to make some docking predictions of that interaction because that can seriously affect my experiments depending on which binding site the fungus protein interacts with.

My main problems are:

• I only have the alphafold structure prediction of both molecules.
• The fungus protein has a disordered region that is highly probable to play a role in the interaction.
• I dont know which is the binding site of neither of the proteins

I've read about IDP-LZerD software, but it is too complex for me to set up. Any ideas on what software can I use?

Answer 1

[Mensur Dlakic]

There are online protein docking servers, and I will let you Google them because I would have to do the same. You have to keep in mind that using static protein structures - what you have - is not the best way if you expect the disordered region to be involved.

A better way would be to feed both proteins simultaneously into AlphaFold so it can model their structures in a context of a complex. If they are expected to bind in a 1:1 ratio, you provide both sequences in the same file and run AlphaFold in multimer mode. It works really well, and you can see an example of a modeled homodimer below.

[Enrico]

I've searched for docking softwares already and used alphafold multidimer. The issue is that as my protein has an IDR, the results from alphafold and other docking softwares I used (such as Haddock) are totally different.

[Mensur Dlakic]

The issue is that as my protein has an IDR, the results from alphafold and other docking softwares I used (such as Haddock) are totally different.

That is to be expected. AlphaFold simultaneously folds the proteins as it creates their interaction interface, so the whole process is dynamic. While Haddock and other "classic" docking software allow limited mobility of select regions of the backbone and side-chains, for practical purposes they work with static structures. If a large conformational change is expected in one or both partners upon docking, I would expect AlphaFold to work better than Haddock. Still, that is only a general recommendation, and in any particular case either one of them can be more appropriate.