multi-chain target hotspots

[PCRandChill]

Hi all,

Thank you for open-sourcing this interesting project.

While running the VHH prediction through the CLI interactive option, I specified target chains: A,B and hotspots A43,A44,B55
I get the error:

  File "/home/usr1/miniconda3/envs/mber/lib/python3.11/site-packages/colabdesign/shared/prep.py", line 42, in prep_pos
    assert len(idx) == 1, f'ERROR: positions {i} and chain {c} not found'
           ^^^^^^^^^^^^^
AssertionError: ERROR: positions 55 and chain B not found

I renumbered my PDB from 1 on both chains and the error persists. The workaround I found at the moment is merging the chains, and providing the target chain as one.

Thanks!

[wentingbian]

Has there been any update on this issue? I encountered the same error and also worked around it by merging the chains. However, this forces me to manually specify the hotspot, since I want the VHH to bind only to one chain. Ideally, I'd prefer the VHH to bind anywhere on that specific chain (except for the binding interface of the other chain) to allow for greater diversity.

Thanks!

[eswan01]

Hi,

We're working to resolve this issue. We plan to push out an update with some fixes within the next week or two. In the meantime, yes merging the chains should yield the desired behavior for multi-chain targets.

[aghanbar-lilly]

@PCRandChill @eswan01 I believe the error occurs because after template processing, the chain IDs have been changed. In src/mber/core/truncation.py, the write_truncated_pdb method changes all chains to 'A':

def write_truncated_pdb(self, output_path: str, inclusion_masks: Dict[str, List[bool]]):
    """Write a new PDB file containing only the truncated residues with chain A."""
    # ...
    for chain_id, mask in inclusion_masks.items():
        for include, residue in zip(mask, self.chain_residues[chain_id]):
            if include:
                res_atoms = residue['atoms'].copy()
                res_atoms.chain_id = np.full(len(res_atoms), 'A')  # <- Changes chain to A

it takes your input PDB with original chain IDs and creates a truncation focusing on the hotspot region. It renames all included residues to chain A and uses this truncated structure for design

The hotspot specification needs to match the chain ID in the processed structure, not the original input. So because of this issue, if you use any chain ID other than 'A' in your config file you will encounter this error.