conflate parameter of Node Normalizer

[erikyao]

Original Purpose: gene-gene equivalence detection

Previously we decided to leverage Node Normalizer to find equivalent NCBI Gene IDs for those gene-presenting CUIs. A predication whose subject or object is an equivalent NCBI Gene ID is considered redundant to the CUI and thus should be deleted by the parser script.

E.g. after the piped CUI C1418660|5361 are separated into 2 rows, there are two predications:

row-id	PREDICATION_ID	PMID	PREDICATE	SUBJECT_CUI	SUBJECT_NAME	SUBJECT_SEMTYPE	OBJECT_CUI	OBJECT_NAME	OBJECT_SEMTYPE
69865	14008146	16541019	INHIBITS	C1418660	PLXNA1 gene	gngm	C1418661	PLXNA2 gene	gngm
69866	14008146	16541019	INHIBITS	5361	PLXNA1	gngm	C1418661	PLXNA2 gene	gngm

5361 is equivalent to C1418660 so the second predication can be deleted.

Move Further: protein-gene equivalence detection when conflate is true

Passing {"conflate": true} to the Node Normalizer means "asking the endpoint to return conflated data" (currently only Gene-Protein conflation). See Babel output formats >> Conflation.

We do have such protein-gene data in the SemMedDB predications, e.g.:

row-id	PREDICATION_ID	PMID	PREDICATE	SUBJECT_CUI	SUBJECT_NAME	SUBJECT_SEMTYPE	OBJECT_CUI	OBJECT_NAME	OBJECT_SEMTYPE
64933	10603013	16530496	ASSOCIATED_WITH	C0020063	PTH protein, human	aapp	C0029463	osteosarcoma	neop
64934	10603013	16530496	ASSOCIATED_WITH	5741	PTH	aapp	C0029463	osteosarcoma	neop

Node Normalizer with {"conflate": true} is able to report the equivalence between C0020063 and 5741

QUESTION: Shall we enable conflate to delete such redundant predications (like the second row above)?

Outlaws: peptide-gene equivalence?

E.g.

row-id	PREDICATION_ID	PMID	PREDICATE	SUBJECT_CUI	SUBJECT_NAME	SUBJECT_SEMTYPE	OBJECT_CUI	OBJECT_NAME	OBJECT_SEMTYPE
64923	10597756	16530483	INTERACTS_WITH	C0027893	neuropeptide Y	gngm	C0039194	T-Lymphocyte	cell
64924	10597756	16530483	INTERACTS_WITH	4852	NPY	gngm	C0039194	T-Lymphocyte	cell

Node normalizer CANNOT report the equivalence between C0027893 and 4852.

QUESTION: Will this be a trouble for BTE?

[erikyao]

Jan 4th meeting, Colleen's input:

• protein-gene: let's do {"conflate": true}
• peptide-gene: we don't have to report their equivalence for now