Extended SIF PubMed IDs column: Source(s)

[jvwong]

The data in EXTENDED_BINARY_SIF format, served from (beta) SIFGraph service contains a column for PubMed IDs (section 1, column 5).

For example, the neighbourhood of TLN1 outputs:

SIRT7   interacts-with  TLN1    BioGRID 22586326        http://pathwaycommons.org/pc13/MolecularInteraction_37c9957e-9ead-4e5f-96d0-670284d6d247___null__86182_
...
ELF1    controls-expression-of  TLN1    MSigDB          http://pathwaycommons.org/pc13/TemplateReactionRegulation_d7aa58d38362a28444d8151192413cee;http://pathwaycommons.org/pc13/TemplateReaction_b4d560bcfc2e59fef106f19a7dada50c
KANK1   interacts-with  TLN1    Biofactoid      Talin-KANK1 interaction controls the recruitment of cortical microtubule stabilizing complexes to focal adhesions.  http://pathwaycommons.org/pc13/MolecularInteraction_635a5bb6bc9090144a4d3fd44ce76095
 ...

Looks like in a lot of cases the PubMed IDs are drawn from BioPAX bp:Evidence attached to the interaction. I wonder if:

1. Would a bp:PublicationXref directly attached to an interaction end up in the extended SIF? Or is it being exclusively drawn from bp:Evidence?
2. If #1 is true generally(not only Evidence): Is it justified to attach a Biofactoid entry's PubMed ID (the one already attached to the bp:Pathway) to the constituent interactions as a bp:PublicationXref in BioPAX? (I guess we won't really have anything approaching bp:Evidence from Biofactoid)

The real reason I ask is because the app-ui interaction app will retrieve the PMID-associated paper info, but this currently won't happen for Biofactoid-derived data:

[IgorRodchenkov]

@jvwong , I dived into our code and found the following:

• cpath2/pc2 web controllers use BiopaxConverter class, calls ExtendedSIFWriter.write in case 'TXT' (extended SIF) format was requested;
• with ExtendedSIFWriter.write (in Paxtools/pattern module) it all comes to calling private method: SIFInteraction.harvestPublicationXrefs(mediators); see also OutputColumn.getColumnValue; "mediators" are interactions (perhaps can be complexes too), their URIs are listed at the end of each extended SIF line (in the first, interactions section);
• harvestPublicationXrefs collects all the PublicationXref objects - BOTH from each mediator interaction's and its evidence's bp:xref properties;
• finally, we only use/print the PUBMED IDs to the output, because ExtendedSIFWriter.write also uses CustomFormat object (is SIFToText interface, stt var.) that explicitely specifies that xref.db must be "pubmed".

Hope it helps. Ask @ozgunbabur too.