Pocket comparison of homologuos proteins suitable to inform mutagenesis

Question

[Schmoho]

I have a set of homologous bacterial proteins, where experimental evidence exists that a drug ligand binds to some of them, but not to others. The working hypothesis is that this is dependent on the protein structure.

Now I am looking for a suitable descriptor of the respective binding pockets that might inform how to mutate either of them to reproduce the binding behaviour of the other one.

Note that for the entire set of proteins only AlphaFold predictions of structure are available.

Is that a sensible endeavour to begin with? What would be a suitable descriptor, or better yet a tool that implements a comparison like that?

Answer 1

[dthorbur]

What do you mean by "suitable descriptor"? As in you want a scoring mechanism? ipTM and pDockQ could be a starting point for PPI investigations. EquiRank00438-0/fulltext) was recently published and looks interested, but I've not tested it. Using a docking tool like HADDOCK is also an option even if it's not considered as powerful as AF2/3.

In some cases AlphaFold works very effectively at distinguishing between real and false protein-protein interactions. But it really depends on how different the sequences and structures for the proteins are, or if there are enough sequences in the MSA databases. If there are only a small number of differences between sequences or poor MSAs it's unlikely the model will be able to distinguish a difference. But, with a small number of proteins, this isn't very much work so it's certainly worth a shot.