v0.5.20

Highlights

Optional extra: LLM for Named Entity Recognition (Annotation)

Optional extra, https://llm.datasette.io/

pip install 'oaklib[llm]'

This provides a lightweight version of OntoGPT. Currently only the annotate endpoint is implemented.

Create a file conf.yaml

categories: [PhenotypicFeature, Disease]
sources: [human-phenotype-ontology]

The conf file is optional - this affects the prompt

Note the conf file format for LLM is experimental and liable to change:

Run:

runoak -i llm:sqlite:obo:hp annotate -C conf.yaml "the patient has bowed and shortened tibia, as well as absent ears and no lungs"

The output is in the standard OAK text annotation format

TODO:

• experiment with different prompt settings
• stabilize format
• further documentation

Attaching metadata to relationships

As requested by @joeflack4

See:

• #656

Example code:

        >>> from oaklib import get_adapter
        >>> adapter = get_adapter("sqlite:obo:mondo")
        >>> rels = list(adapter.relationships(["MONDO:0009831"]))
        >>> for rel, metadatas in adapter.relationships_metadata(rels):
        ...     for p, v in metadatas:
        ...         print(rel, p, v)
        <BLANKLINE>
        ...
        ('MONDO:0009831', 'rdfs:subClassOf', 'MONDO:0002516') oio:source NCIT:C9005
        ...

Or on the command line:

runoak -i sqlite:obo:mondo relationships MONDO:0009831 --include-metadata -p i

What's Changed

• Adding llm as an extra, adding llm implementation for annotation by @cmungall in #658
• Additional documentation by @cmungall in #655
• Adding ability to retrieve axiom annotations on relationships (axiom annotations) by @cmungall in #656 @joeflack4
• minor obographviz config change: new color disambiguating chebi by @cmungall in #657

Full Changelog: v0.5.19...v0.5.20