Knowledge Graph Standard API

[stevencox]

Is it feasible to create a Sharpener Translator Knowledge Graph Standard (KGS) API?

In general the Sharpener takes a list of genes, parameters for the expansion method, and produces a new list of genes. I think. Is that correct?

If a new endpoint - eg, /graph/query - were implemented to accept a KGS Message object:

• The query could be required to be gene_list-[expansion_type]->gene_list where
  • gene_list would need to be added to the biolink_model
  • expansion_type would also be added to the biolink_model
  • subtypes of each could also be added to allow users to make granular selections
• The knowledge graph would be populated with the list of genes
• The options element of the Message object would include expander specific options.

Assuming there's important semantic significance to preserving the notion of a list of genes, the input and output graphs will probably want to reify an object to represent the list with edges to each list member. Then, in principle, the output would include an edge associating the list objects with properties characterizing the transformation.

[stevencox]

Looks like "gene group" is already a thing in the biolink-model:

  gene grouping:
    abstract: true
    mixin: true
    description: >-
      any grouping of multiple genes or gene products

There's a gene to gene association but not (yet) a gene group to gene group association.

[kennethbruskiewicz]

There is an alternate conversation for standardization that ties into the provenance discussion. We want our Transformers to report meta-data about themselves, including their computation and the kind of scoring data they return. Unsure if it's Biolink's problem and I heard EDAM codes cover this sort of thing?

[stevencox]

@vdancik - any further thoughts on this?