PGAP2 mystery

[Richard Bruskiewich]

The display of SemMedDb (explicit) predications in the results table of the Knowledge.Bio desktop when PGAP2 is used as the current concept, reports ATAD5 concepts in relations rather than PGAP2.

The attached document shows the sequence of (mysql) queries giving this anomaly, which stems from the observation that the database "Predication_Argument" table, tracking the subject and object concepts associated with a given Predication, violates the strict cardinality expectation of having only one subject and one object per Predication. In fact, in many cases in the database, either/or more than one subject and/or object is recorded against a given predication.

It is unclear what causes this violation of cardinality, although one strongly suspects is is a data curation issue in the copy of SemMedDb used in the system. However, one would clearly like to filter out irrelevant concepts and only report relations aligned with the target "current concept".

[Richard Bruskiewich]

the code responsible for retrieving and using a given Predication's set of Predication_Arguments was modified to filter on the "current concept". This code should work for all instances of Predications in which the case where either subject or object concepts are in excess, but NOT both (i.e. the 'other' side of the relation is a singular concept) AND the current concept is in the multiple subject/object concept set. In this way, the relation is decidable. Otherwise, the relation returned is not guaranteed to be the sole or correct one (though it should correctly display the current concept)