Are pocket restraints with affinity prediction working?

[KirillShmilovich]

I've tried a number of examples defining lists of residue indexes for the purposes of defining pocket constraints:

e.g., in the contacts field.

constraints:
  - pocket:
      binder: B
      contacts: [[A,1]]
      max_distance: 8

Even when defining example residue indexes (like 1 or the last possible index) appears not to impact the ligand placement.

Starting from a template ligand-bound structure, I'm concerned that the tools i'm using are not able to accurately return the correctly expected residue indexing for this field. I'm wondering if there is any guidance on how given a template structure I can get a list of residue indexes from the ligand contacts I could use for this purpose.

Thanks in advance and happy to clarify thing further!

[jwohlwend]

Yep sorry we are fixing that. I ll update here when it’s done!

[rajiv03]

Same here. It's not using the constraints. Please update.

[xiaoweilin0813]

I have also encountered this issue. Regardless of where I specify the interacting amino acid residues, the predicted binding sites are not in that location. For example, some transmembrane protein targets are in the transmembrane region, but even if I specify the target location, Boltz-2 predicts the binding site to be inside, which is clearly incorrect. I hope the authors can resolve this issue.

[mosayebi]

@jwohlwend Do you have any updates or possible workarounds for this issue?

[sherryliu987]

Also seeing this issue

[SSchott]

Contact constraints also do not seem to be working? token2 does not even seem to be parsed, but maybe I am missing something.

[JenkeScheen]

see https://github.com/jwohlwend/boltz/issues/435

[gcorso]

Hi all, pocket and contact conditioning support is now merged in the public repo (new release coming soon). On top of it, you can now force the constraint to be satisfied using inference-time potentials. Let us know if you find any other issues!