Get phenio relations from BBOP-Sqlite, then filter

[kevinschaper]

Instead of generating our own outside the pipeline, we can use https://s3.amazonaws.com/bbop-sqlite/phenio-relation-graph.db.gz.

This is a full expansion, and we only want a subset. My initial guess is that we can filter on just rdfs:subClassOf and BFO:0000050

It's likely that we'll be doing it in a python way rather than a shell way, but if awk is handy, here it is:

awk '{ if ($2 == "rdfs:subClassOf" || $2 == "BFO:0000050") { print } }' phenio-relation-graph.tsv > phenio-relations.tsv

[kevinschaper]

Also include UPHENO predicate that will connect anatomy to phenotype

[kevinschaper]

We can get the relation graph output directly from the latest phenio release (phenio, not kg-phenio).

Making this high priority for this week, Chris noticed that the closure field wasn't working for MONDO:0000508

sqlite> select count(*) from edges where object= 'MONDO:0000508' and predicate='biolink:subclass_of';
409
sqlite> select count(*) from denormalized_edges where subject_closure like '%MONDO:0000508%';
0

Our relations file didn't have any entries for MONDO:0000508, but the newer file has plenty of rdfs:subClassOf triples.

The important thing to remember with the phenio relations file is that it needs to be filtered to include rdfs:subClassOf, BFO:0000050 and the upheno predicate that connects anatomy to phenotype

[caufieldjh]

Just so I can be sure I'm parsing this correctly and that all outputs are working as expected - does the phenio relation-graph file contain the necessary closures?

[kevinschaper]

@caufieldjh I think it does have the necessary closures. I don't know if I generated something wrong in the one that we're using right now, or if it's just out of date.

[kevinschaper]

I'm not sure whether it's UPHENO:0000003 (phenotype is associated with entity) or UPHENO:0000001 (phenotype affects entity) that I should include.

I think I may go forward on this PR with affects UPHENO:0000001?

Does that sound right @matentzn @sabrinatoro ?

[matentzn]

Sure, note that UPHENO:1 is a sub-property of UPHENO:3: https://github.com/obophenotype/upheno-dev/blob/master/src/ontology/components/upheno-relations.owl