In contrast, TRAPI 1.0 currently relaxes this constraint, giving data type of 'string' which doesn't need to be a CURIE but can be any ad hoc string value that the TRAPI implementation deems an adequate naming of the source 'relationship' of the edge.
This semantic "impedance mismatch" between the Biolink Model and TRAPI may potentially be problematic for downstream processing, such as for Edge Normalization of knowledge graphs.
We would like to provide documented guidance on possible options for Knowledge Providers compliance with the stricter Biolink Model semantic requirements of a CURIE being the value of the relation property of the Biolink Model Association class.
The current Biolink Model definition for the
relationproperty of the Biolink ModelAssociationclass (which maps onto the correspondingrelationproperty of the Translator Reasoner API (TRAPI)Edgeschema model) is that the value of the property should be a CURIE.In contrast, TRAPI 1.0 currently relaxes this constraint, giving data type of 'string' which doesn't need to be a CURIE but can be any
ad hocstring value that the TRAPI implementation deems an adequate naming of the source 'relationship' of the edge.This semantic "impedance mismatch" between the Biolink Model and TRAPI may potentially be problematic for downstream processing, such as for Edge Normalization of knowledge graphs.
We would like to provide documented guidance on possible options for Knowledge Providers compliance with the stricter Biolink Model semantic requirements of a CURIE being the value of the
relationproperty of the Biolink ModelAssociationclass.