load_sdf imports first reference id only

[laurenreid1]

Tested using develop branch but bug appears to also be present in main.

The following snippet calls animal.load_sdf and includes the reference_col keyword argument to signal that the reference_id SDF column should be used to register the HIPPO references:

animal = hippo.HIPPO(f"<TARGET>.sqlite", update_legacy=True)

animal.load_sdf(
                target=target_name, 
                path="<PATH_TO_SDF>", 
                inspiration_col="inspiration_ids", 
                reference_col="reference_id",
                compound_tags = ["<TAG>"],
                pose_tags = ["<TAG>"],
                name_col = "ID",
            )

However animal.py lines 837-845 (in develop branch):

takes the first rows reference_id and assigns it to the reference variable. The following rows then skip these lines of code because the reference variable is not None. The result is that every pose is assigned to the first row's reference in the DB.

Suggestion:

Change animal.py lines 837-845 (develop branch) to something like this:

            if not reference:
                ref_str = row.get(reference_col)
                if ref_str:
                    try:
                        row_reference = int(ref_str)
                    except ValueError:
                        row_reference = inspiration_map[ref_str]
                else:
                    row_reference = None

            elif isinstance(reference, Pose):
                row_reference = reference.id

and lines 891-906 to:

            data.append(
                dict(
                    alias=alias,
                    compound_id=compound_id,
                    target_id=target.id,
                    path=pose_path,
                    metadata=metadata,
                    inspiration_ids=inspiration_list,
                    reference_id=row_reference,
                    mol=mol,
                    inchikey=inchikey,
                    smiles=smiles,
                    energy_score=energy_score,
                    distance_score=distance_score,
                )
            )

Note - introduction of row_reference

[mwinokan]

Thanks @laurenreid1 this is a key bug to squash. Please make the suggested changes and request a PR to the dev branch. It would be great if you could run the unit tests before you make the PR. Coverage is currently limited but it should be a minimal sanity check.

From the README.md:

Some tests are provided in the tests directory, which can be run with pytest:

cd tests
pytest

N.B. the numbered tests, e.g. test_00_cleanup.py need to run in sequential order to set up the database. Other tests can run in arbitrary order thereafter. The tests will fail if https://fragalysis.diamond.ac.uk/ can not provide the protein target's data, as specified in tests/config.py.