ielis
commented
1 month ago The top-level JSON Schema document phenopacket-schema.json used by phenopacket-tools is in this folder under v2 subfolder. It is indeed a little hard to find.
I think validation of Phenopacket Schema elements such as phenopacket, family or cohort needs to have multiple tiers. The lowest tier can be implemented with JSON Schema to check types, absence of random fields, correct cardinalities, etc.
The upper level tiers should e.g. check presence of a Metadata | Resource element for all used OntologyClasses, using current term IDs for an ontology version (easiest to implement for HPO for me), or more exotic requirements, such as not using term (e.g. Clonic seizure) and its ancestor (e.g. Seizure) in a single phenopacket.
I think it may be hard to implement all these just using JSON Schema, Pydantic, or LinkML. So, we should have an API somewhere and a bunch of out-of-the-shelf validators. I would favor writing these in Python.
Nevertheless, right now we check the validity mostly manually in the code and we manually run phenopacket-tools afterwards. It would probably be good to automatize this.
I was confused as to why https://github.com/monarch-initiative/phenopacket-store/issues/97 was necessary
As I understand the following java repo does validation using json-schema https://github.com/phenopackets/phenopacket-tools
(I wasn't able to find the json-schema)
I had assumed that pyphetools used the same schema, but it looks like the only validation is procedural, not complete schema validation.
Shouldn't pyphetools do schema validation (using any of jsonschema, pydantic, linkml, ...)