Do not lookup bonds for hetero residues

[Emrys-Merlin]

The PR aims at resolving issue 818. The idea is to disable the bond lookup in the compound dictionary for hetero atoms. I tried to keep the changes minimal, but I'm not sure if that introduces too much special casing. I'm looking forward to your feedback.

My first, naive approach to globally disable the bond lookup did not work, because the bond parsing from .pdb files and .cif files works a bit differently, as far as I understand. While for pdbs CONECT entries are parsed as bonds directly and then extended by bonds found in the dictionary lookup (see here), for cifs, the whole _chem_comp_bond table is parsed as the compound dictionary and all the bonds (for both hetero and regular atoms) are looked up in it (see here). So, I introduced the ignore_hetero flag, which is False by default (preserving the current behavior) and I set to True for pdb parsing. Lastly, I introduced water special casing, because in the test file 1crr, the pdb does not contain CONECT entries for the water atoms, while the CIF _chem_comp_bond does. I am not sure how common that scenario really is or if it would be possible to adapt the test case.

[padix-key]

Thanks for tackling this 👍. I think this can be solely fixed by changes in biotite.structure.io.pdb as custom_bond_dict already supports the required flexibility. This way we would automatically respect the differences in PDB and PDBx bond definitions.

We could use the approach from here (see the Notes section). We could supply the custom_bond_dict only with hetero=False residues plus water. So in PDBFile instead of calling connect_via_residue_names(array) we could do something like this:

custom_bond_dict = {
    custom_bond_dict[res_name]: bonds_in_residue(res_name)
    for res_name in itertools.chain(np.unique(array.res_name), ["HOH"])
}
connect_via_residue_names(array, custom_bond_dict=custom_bond_dict)

[Emrys-Merlin]

Thank you for your feedback. Unfortunately, I run into a 404 when I click your link. I think, I understand your idea though and will start working on the refactor.

[padix-key]

I pasted the wrong text 😅, now the URL in the link is working.

[codspeed-hq]

CodSpeed Performance Report

Merging #820 will not alter performance

_{Comparing Emrys-Merlin:disable_hetatm_bond_lookup (81dab9a) with main (a541322)}

Summary

✅ 59 untouched benchmarks

[padix-key]

Looks good to me, can I merge this PR?

[Emrys-Merlin]

Yes, please feel free :) I would have done it, but, unsurprisingly, I do not have the permissions :D