a new SIF rule for MIs, using "spoke" model (bite-prey)
IgorRodchenkov opened this issue · 1 comments
Original email message:
"I'm (Wei) a PhD student at Dr. Hua Xu's lab at University of Texas Health Science Center at Houston. One of the ongoing project in our lab involves data from PathwayCommons 5.
I used Paxtools to extract binary interaction information from the owl file of PathwayCommons 5 with the following command:
java -Xmx50g -jar paxtools-4.3.0.jar toSifnx Pathway\ Commons.5.All.BIOPAX.owl PathwayCommons.5.All.BIOPAX.Interactors_2.tsv PathwayCommons.5.All.BIOPAX.Interactions2.tsv "Entity/xref:UnificationXref,Entity/xref:RelationshipXref" "Interaction/dataSource/name,Interaction/xref:PublicationXref,Interaction/evidence:Evidence/xref:PublicationXref"
One problem with this command is that, it forms a binary interaction for all the participants in an interaction of the owl file. It's not always correct because some participants may not have interactions between each other. Take "MolecularInteraction_1414541474571" as an example. It is an interaction with 39 participants from IntAct. One of them is a bait and the other 38 are its preys, so there should be 38 interactions in total. However, Paxtools convert it to 39*38/2 = 741 interactions by considering an interaction between any two participants.
Is there any way to convert this type of interaction (with baits and preys) correctly using Paxtools?
Thanks!"
We need an additional SIF rule to use the "spoke" model (bait-prey edges, but no prey-prey edges) to biopax MIs if the information about which participant is "bait" and which are "prey" is present (in interaction's Evidence/ExperimentalForm).
PS: The user used PC2 v5 (perhaps the old SIF output). Current PC2 v7 does not have ExperimentalForms (current 4.3.1-SNAPSHOT paxtools' psimi-converter skips them).
PC2 v8 and future paxtools versions will get ExperimentalForms for MIs back.
Updated psimi-converter and paxtools-pattern to make this rule possible.