linkml-aop

A LinkML schema for Adverse Outcome Pathways, with EMOD (Evidence Model) expansions.

Purpose

The schema serves two purposes at once.

Validating content already in the AOP-Wiki. AOP-Wiki content is curated by people, in free text as much as in structured fields. Expressing the data model as a LinkML schema makes it possible to check that content against stated expectations - that a referenced Key Event exists, that a controlled vocabulary is used where one applies, that an assertion carries the evidence it claims.

Providing a basis for AOPs derived by automated approaches. Adverse Outcome Pathways can be assembled from analysis of multimodal data and from text mining. Such a pathway is only as useful as the record of where each of its parts came from, so the schema has to be able to carry that record: which observation was made, by what method, and what was inferred from it.

Both purposes rest on the same requirement - what evidence supports must be distinguishable from what has been inferred - which is why Assay, Observation, and Evidence are separate classes rather than fields on an Event, and why the definition of Event is explicit that a Key Event is an abstraction rather than something that happened.

Source of the schema

The schema is generated from the MySQL database behind AOP-Wiki EMOD 3.0 and then curated. See the repository README for the generation and curation workflow, and src/docs/dev/definition_rationale.md for the reasoning behind individual class definitions.