Some non-blackbox pathways result in empty SBGN model

[IgorRodchenkov]

For examle, http://identifiers.org/reactome/R-HSA-69278,

http://beta.pathwaycommons.org/pc2/get?uri=http://identifiers.org/reactome/R-HSA-69278&format=sbgn

This pathway ('Cell Cycle, Mitotic') does not have any pathwayComponent values but does have processes in its pathwayOrder property.

Looks, this is Paxtools' sbgn-converter bug/issue.

[IgorRodchenkov]

@gbader @d2fong @jvwong @emekdemir @ozgunbabur @ugurdogrusoz

Wow,

http://identifiers.org/reactome/R-HSA-69278 http://reactome.org/content/detail/R-HSA-69278

• this pathway is a very interesting example! Not sure how we'd want to convert and display that, but Reactome does cool!

Well, this turns out to be a pathway that originally contained only (7) sub-pathways, and there're no interactions directly associated with it. When this (BioPAX sub-model) is detached/extracted from the large PC model, all the sub-pathways disappear, and so - nothing to show in the SBGN...

This might be ok?

I am not sure whether we want to get the complete sub-graph of a top/parent pathway and convert to SBGN (which can be very large, impossible to visualize anyway) or not...

There is a way to tell whether a pathway directly contains any interactions: Pathway/pathwayOrder/stepProcess:Interaction Pathway/pathwayComponent:Interaction

There is also way to tell if a pathway contains sub-pathways: Pathway/pathwayOrder/stepProcess:Pathway Pathway/pathwayComponent:Pathway

• the above queries could be used by PC web apps (pathways-search) to suggest a list of sub-pathways to view...

PS: (off-topic) Also, I am thinking to change Paxtools Completer logic - e.g., to include first-level sub-pathways with their names and comments but without member BioPAX objects (it won't help this here issue though)...

[IgorRodchenkov]

See the last comment and commit at BioPAX/Paxtools#25 - direct sub-pathways are now included in a pathway's BioPAX model returned by a query by default, but those sub-pathways come without their component and step processes (which are still there in the large model, and one would see it there if get the sub-pathways by URI).

So, the SBGN result will still be empty, but BioPAX now makes more sense.

I suggest to make Paxtool's sbgn-converter generate nodes/edges for "empty" and blackbox sub-pathways; see BioPAX/Paxtools#30.

[IgorRodchenkov]

Something's been done about this, including - improved the Paxtools sbgn-converter; no more ideas here. Closed (let's open a new, more specific ticket, if needed.)