annotation assertion axioms (including subsets/slims) should by default be treated as node properties

[cmungall]

I am seeing e.g. in the chebi kgx that subsets like 3_STAR are being treated as nodes with inSubset as edges (type related_to)

edges should connect domain entities

for all annotation assertions (triples where the property is an AP) should be treated as node propeties

[justaddcoffee]

Thanks Chris, a bit of context that might help the KGX dev team. Here is an example of the CHEBI KGX TSVs that Chris is referring to, containing these sorts of edges:

~ $ tar -xzvf chebi_kgx_tsv.tar.gz
x chebi_kgx_tsv_nodes.tsv
x chebi_kgx_tsv_edges.tsv
~ $ grep -i star chebi_kgx_tsv_edges.tsv | head
CHEBI:141943-biolink:related_to-OBO:chebi#2_STAR	CHEBI:141943	biolink:related_to	OBO:chebi#2_STAR		OIO:inSubset	tmpl2n2_ja9_chebi_relaxed.owl
CHEBI:65323-biolink:related_to-OBO:chebi#3_STAR	CHEBI:65323	biolink:related_to	OBO:chebi#3_STAR		OIO:inSubset	tmpl2n2_ja9_chebi_relaxed.owl
CHEBI:164715-biolink:related_to-OBO:chebi#2_STAR	CHEBI:164715	biolink:related_to	OBO:chebi#2_STAR		OIO:inSubset	tmpl2n2_ja9_chebi_relaxed.owl

This was generated in kg-obo with this code applied to (I think) this file

So you should be able to reproduce this with something like:

$ kgx --version
kgx, version 1.5.1
$ wget http://purl.obolibrary.org/obo/chebi.owl
$ kgx transform -i owl -f tsv chebi.owl

I think we'd like the object in these triples (OBO:chebi#2_STAR, OBO:chebi#3_STAR, etc) to be node properties instead of nodes themselves.

[sierra-moxon]

@justaddcoffee - is this still an issue?