Aim 2.2.a. Evaluate consistency of relation and entity types used within BioPortal ontologies.

[caufieldjh]

Specifically:

• Get total counts of KGX validation error by type across all transforms (e.g., "X transforms have at least one error of type Y")
  • Which transforms did not work?
  • How many of those transformation issues are resolvable?
  • How many of those transformation issues are not worth resolving in this repository (e.g., something structurally wrong with ontology on Bioportal; entry may be duplicated, test, or otherwise broken)
• Enumerate usage of entity types by ontology (e.g., "X ontologies use biolink:OntologyClass)
• Enumerate usage of relation types by ontology (e.g., "X ontologies use biolink:subclass_of)

[caufieldjh]

Output of newly added script get_all_transform_stats.sh:

All processed ontologies:
910
All successful JSON transforms:
896
All successful KGX TSV transforms:
888
Transforms with at least one of the following errors:
MISSING_NODE_PROPERTY
0
MISSING_EDGE_PROPERTY
0
INVALID_NODE_PROPERTY
833
INVALID_EDGE_PROPERTY
796
INVALID_NODE_PROPERTY_VALUE_TYPE
28
INVALID_NODE_PROPERTY_VALUE
833
INVALID_EDGE_PROPERTY_VALUE_TYPE
0
INVALID_EDGE_PROPERTY_VALUE
796
MISSING_CATEGORY
0
INVALID_CATEGORY
888
Category 'OntologyClass' is a mixin in the Biolink Model
888
MISSING_EDGE_PREDICATE
0
INVALID_EDGE_PREDICATE
495
MISSING_NODE_CURIE_PREFIX
0
DUPLICATE_NODE
0
MISSING_NODE
0
INVALID_EDGE_TRIPLE
0
VALIDATION_SYSTEM_ERROR
0

The big take-home here is that entities in every transform gets assigned biolink:OntologyClass despite Biolink modeling OntologyClass as a class mixin rather than intending it to be a class type itself.

Do we know enough about each ontology to assign a mode specific class to nodes?

There are other metaclasses, like [biolink:TaxonomicRank](https://w3id.org/biolink/vocab/TaxonomicRank) - these may still make sense to use in some contexts.

[caufieldjh]

Finding appropriate mappings vs. Biolink is a goal for kgx - that will help to reduce the number of OntologyClass nodes.

[caufieldjh]

Completely failed transforms:

ID	Name	Issue
NIFSTD	Neuroscience Information Framework (NIF) Standard Ontology	#15
EXACT	An ontology for experimental actions	#15 ; Small, alpha status, last uploaded 2014
DOID	Human Disease Ontology	#15 ; Unknown - would really expect this to work
ECOCORE	An ontology of core ecological entities	Empty? Error in Bioportal? Last updated Mar 10 2022
ETHIOPIADISEASES	EthiopiaDiseaseList	Empty? Does not render on Bioportal
LC-CARRIERS	Library of Congress Carriers Scheme	Empty? In SKOS format; does not render on Bioportal
SCDO	Sickle Cell Disease Ontology	#15
FENICS	Functional Epilepsy Nomenclature for Ion Channels	#15 ; Using webprotege: prefix (unsure if related to transform fail)
FOVT	FuTRES Ontology of Vertebrate Traits	#15
PTRANS	Pathogen Transmission Ontology	#15 ; Does not render on Bioportal
TIMEBANK	Timebank Ontology	#15
GSSO	Gender, Sex, and Sexual Orientation Ontology	#15
CST	Cancer Staging Terms	Unknown ; Does not render on Bioportal
MARC-RELATORS	MARC Code List for Relators	Empty? Does not render on Bioportal

[caufieldjh]

Transforms translating to Obojson but not to KGX TSV:

ID	Name	Issue
PDRO	The Prescription of Drugs Ontology	Unknown CURIE prefix: file
VICO	Vaccination Informed Consent Ontology	Unknown CURIE prefix: file
IXNO	Interaction Ontology	Last updated in 2011; Unknown CURIE prefix: file
IDQA	Image and Data Quality Assessment Ontology	Unknown CURIE prefix: file
KTAO	Kidney Tissue Atlas Ontology	Unknown CURIE prefix: file
GAZ	Gazetteer	Unknown CURIE prefix: file; KG-OBO transforms GAZ w/o issue, see https://kg-hub.berkeleybop.io/kg-obo/gaz/no_version/
CANONT	Upper-Level Cancer Ontology	Last updated in 2012; Unknown CURIE prefix: file

These are generally issues with the OBONamespace set to a local file path, and in at least one case (VICO) it's because of references to another namespace beginning with file: (GAZ).

[caufieldjh]

See #23 for Unknown CURIE prefix: file issue.

[caufieldjh]

With issues #15 and #23 resolved, the only remaining problematic transforms are:

• ECOCORE (use current BioPortal submission)
• ETHIOPIADISEASES (drop)
• LC-CARRIERS (drop)
• CST (drop)
• MARC-RELATORS (drop)

[caufieldjh]

ECOCORE has a new version on BioPortal - can just use this for now:
https://bioportal.bioontology.org/ontologies/ECOCORE/?p=summary

Can drop LC-CARRIERS and MARC-RELATORS.

[jvendetti]

Hi Harry.

ETHIOPIADISEASES

The latest submission in our system was corrupt. I recreated/reprocessed the submission so that the ontology is accessible again:

https://bioportal.bioontology.org/ontologies/ETHIOPIADISEASES?p=summary

CST

It looks like the end user uploaded an ontology source file for this entry, but we were never able to load the data into the triplestore, because our code errors out when we try to serialize to RDF/XML format with the following error:

org.semanticweb.owlapi.rdf.rdfxml.renderer.IllegalElementNameException: Illegal Element Name (Element Is Not A QName): http://www.w3.org/2000/01/rdf-schema#comment:

I think this one could probably be dropped for now.

[caufieldjh]

Great - thanks @jvendetti !

[jvendetti]

Hi @caufieldjh. It turns out that the maintainers of ETHIOPIADISEASE told John that they no longer need this entry in BioPortal. I had originally reprocessed it, but I've now deleted the entry.

[caufieldjh]

Great, thanks! One more off the list.

[caufieldjh]

Updated statistics, including for types:

*** General ontology counts:
All processed ontologies:       910
All successful JSON transforms: 906
All successful KGX TSV transforms:      903
All transforms with KGX validation logs:        902
All transforms with ROBOT measure reports:      883
All transforms with ROBOT validation reports:   904
Ontologies with failed transforms:      
./transformed/ontologies/ETHIOPIADISEASES
./transformed/ontologies/LC-CARRIERS
./transformed/ontologies/CST
*** Transforms with at least one of the following errors:
MISSING_NODE_PROPERTY   0
MISSING_EDGE_PROPERTY   0
INVALID_NODE_PROPERTY   844
INVALID_EDGE_PROPERTY   807
INVALID_NODE_PROPERTY_VALUE_TYPE        31
INVALID_NODE_PROPERTY_VALUE     844
INVALID_EDGE_PROPERTY_VALUE_TYPE        0
INVALID_EDGE_PROPERTY_VALUE     807
MISSING_CATEGORY        0
INVALID_CATEGORY        902
Category 'OntologyClass' is a mixin in the Biolink Model        902
MISSING_EDGE_PREDICATE  0
INVALID_EDGE_PREDICATE  502
MISSING_NODE_CURIE_PREFIX       0
DUPLICATE_NODE  0
MISSING_NODE    0
INVALID_EDGE_TRIPLE     0
VALIDATION_SYSTEM_ERROR 0
*** Node type counts:
biolink:NamedThing      731
biolink:OntologyClass   903
biolink:BiologicalProcess       76
biolink:Cell    110
biolink:CellularComponent       46
biolink:ChemicalSubstance       119
biolink:Disease 15
biolink:Event   2
biolink:ExposureEvent   3
biolink:Gene    9
biolink:MolecularActivity       49
biolink:NamedThing      731
biolink:OntologyClass   903
biolink:OrganismalEntity        128
biolink:Pathway 6
biolink:PhenotypicFeature       44
biolink:Protein 79
biolink:SequenceFeature 56
biolink:SexQualifier    1
biolink:Source  2
biolink:TaxonomicRank   3
biolink:Unit    2
biolink:AnatomicalEntity        112
*** Edge type counts (i.e., predicate types):
biolink:related_to      376
biolink:subclass_of     899
biolink:part_of 52
biolink:inverseOf       408
biolink:subPropertyOf   449
biolink:has_part        165
biolink:has_participant 99
biolink:has_unit        29
biolink:preceded_by     69
biolink:has_attribute   76
biolink:positively_regulates    35
biolink:negatively_regulates    37

This includes all node types across all ontologies, and a selection of the more common predicate types.
Note that these are largely the result of type assignment by KGX.
As expected, nodes with biolink:NamedThing or biolink:OntologyClass are ubiquitous, suggesting that many may be re-assigned to more informative types.
Though predicate types appear more consistent, there is a long tail of sparsely-used types (not shown) across all ontologies.

[caufieldjh]

Closing issue as complete - reopen as needed