search

OpenBEL / converter-reactome_to_bel

Convert Reactome to BEL
Apache License 2.0
0 stars 0 forks source link

treat hasmember reactome relationships as hasMember in BEL #1

Open tythomson opened 8 years ago

tythomson commented 8 years ago

hasmember is used in reactome to represent family or set relationships. These are currently represented with hasComponent relationships in BEL, but should instead be represented with hasMember relationships. e.g. http://www.reactome.org/PathwayBrowser/#/R-HSA-163685&SEL=R-HSA-429699: VAPA/B has members VAPA and VAPB

tythomson commented 8 years ago

Note that there are some cases where the same set/family is not used on both sides of the reaction, and the reaction represents the conversion of one member of a family into the corresponding member of another family. For example, in R-HSA-1655842.1 (http://www.reactome.org/PathwayBrowser/#/R-HSA-1655829&SEL=R-HSA-1655842&PATH=R-HSA-1430728,R-HSA-556833), SREBF family members are cleaved. The set of SREBF members is listed as inputs/reactants, and the set of cleaved SREBF members is listed as outputs/products. In most other cases with sets/families, the same set/family is used in both the products and reactants. In this case, it's a different set on either side of the reaction with different set members (the cleaved form of the SREBF family).

I don't think that this complicates the implementation of the BEL conversion, but is something to consider downstream when interpreting families/sets in reactions from reactome.

tythomson commented 8 years ago

Another thought: the majority (maybe even all?) of the terms in reactome that don't map to CHEBI or Uniprot name spaces are ad hoc protein/chemical families that were defined to enable a single reaction to represent what would happen to many different entities. These translate to bel as 'a(family_name)'. It would be helpful to edit the entity names to make it clear that they are ad hoc groups/families and not just molecular entities whose names did not map to a known namespace. Perhaps we could introduce a new namespace ReactomeFam (e.g.., 'a(REACTFAM:family_name)') or just modify the entity name (e.g., 'a(reactome family - family_name)'). Same thing for hascandidate relationships (issue #2), and if we really wanted to be helpful we could always label these as CANDIDATEREACTFAM or 'candidate reactome family'.