swanss
commented
2 years ago Hi Victor,
In the paper we present peptide binder designs that target sites on TRAF6 that are not known to bind to peptides (sites B and C). That being said, more recently I've used AlphaFold-Multimer to predict the bound structure of the designed peptides, and it predicts that all designs interact with site A (the native site). It's a little difficult to interpret this result, but it could suggest that it's harder to bind the alternative sites with a peptide.
In general, I would say that the problem of de novo design (of any kind of a binder, peptide or mini-protein) to a site that is not known to natively bind a protein/peptide is unsolved. If you have any examples, please point me towards them. That being said, it should be possible to design a peptide that binds a new site as long as there are sufficient exposed hydrophobic residues. It's hard to draw a rule as to what is enough, but after briefly looking at 7BW1 it seems like there are enough hydrophobic patches to give it a shot. However, the pocket is relatively small, so it might be hard to fit the backbone + appropriate sidechains without disrupting the protein structure.... it will be interesting to see what designs are proposed by the method.
I'm working on new methods for scoring designs that may be able to better reflect whether a protein site can be bound by a peptide, I can keep you updated if you're interested!
victorconan
Hi Sebastian,
Very interesting work! Thanks for sharing!
I am not sure if I understood the paper properly, but it seems that only protein targets with known peptide binders in PDB are suitable for this method. Because it seems you have to know the binding sites of the protein targets to the peptide. Is that correct? For example, if I am interested in PDB id: 7BW1, it does have small molecule ligands, but it does not have peptide complexes. Can I still use this method for designing peptides for such targets?
Thanks!