Validate citation metadata

[stellaprins]

The citation metadata does not have the right format for the nadkarni_mri_mouselemur_91um atlas.

from brainglobe_atlasapi.bg_atlas import BrainGlobeAtlas
atlas = BrainGlobeAtlas("nadkarni_mri_mouselemur_91um")
print(atlas)

Results in a ValueError because citation metadata of this particular atlas doesn't have expected format (str instead of str, str).

The citation metadata for this atlas is "https://doi.org/10.1016/j.dib.2018.10.067", while I understood from @alessandrofelder it should follow a strict format and be something like "Natkarni et al., 2019", "https://doi.org/10.1016/j.dib.2018.10.067".

@alessandrofelder and myself discussed taking the following steps:

• add metadata format validation to brainglobe_atlasapi\atlas_generation\validate_atlases.py
• identify atlasses with missing metadata
• go through the atlas generation scripts in brainglobe_atlasapi\atlas_generation\atlas_scripts and add missing metadata
• regenerate atlasses with the correct metadata

[stellaprins]

Problem with metadata['citation']

A single string "{citation}, {doi}" (eg. "Brainglobe et al. 2025, https://doi.org/BrainGlobe/999" is currently the correct way to define the citation metadata according to `METADATA_TEMPLATE``.

_rich_atlas_metadata does not process the citation metadata correctly in case there is no comma seperating {citation} and {doi}, and / or when there are commas situation in locations other than between {citation} and {doi}.

So for nadkarni_mri_mouselemur_91um where the citation metadata is a single string {doi} and for allen_mouse_bluebrain_barrels_10um where there are multiple commas in {citation}, the citation metadata is not correctly processed.

Metadata template

I want to use the METADATA_TEMPLATE frombrainglobe_atlasapi\descriptors.py for validation of the metadata value types.

    for key, value in METADATA_TEMPLATE.items():
        assert key in atlas.metadata, f"Missing key: {key}"
        assert isinstance(atlas.metadata[key], type(value)), (
            f"Key '{key}' should be of type {type(value).__name__}, "
            f"but got {type(atlas.metadata[key]).__name__}."
        )

However there are several isseus with the METADATA_TEMPLATE. The name value contains charachters that are not allowed (see #524), and the resolution and shape values are tuples while they are lists for the atlasses I've checked.

METADATA_TEMPLATE = {
    "name": "name/author/institute_species_[optionalspecs]",
    "citation": "Someone et al 2020, https://doi.org/somedoi",
    "atlas_link": "http://www.example.com",
    "species": "Gen species",
    "symmetric": False,
    "resolution": (1.0, 1.0, 1.0),
    "orientation": "asr",
    "shape": (100, 50, 100),
    "version": "0.0",
    "additional_references": [],
}

I suggest changing it to something like this:

METADATA_TEMPLATE = {
    "name": "source_species_additional-info",
    "citation": {"Someone et. al., 2020": "https://doi.org/somedoi"},
    "atlas_link": "http://www.template_brain_atlas_link.com",
    "species": "animal",
    "symmetric": False,
    "resolution": [1.0, 1.0, 1.0],
    "orientation": "asr",
    "shape": [100, 50, 100],
    "version": "0.0",
    ~~"additional_references": {"ref1":"doi1","ref2":"doi2"},~~
     "additional_references": [],
}

Here all references / citations are dictionaries with {"ref":"doi_url"} pairs, the name value matches the stricter format, and the tuples are made into lists to match what I've seen in the actual atlasses. Any comments or suggestions are welcome (@alessandrofelder and @adamltyson).

note: maybe I do not understand additional_references very well, they seem more like keywords.

Another, maybe easier (?), option is to keep the citation a single string and assume the doi always starts with "https://doi" (which is the currently accepted standard).