AlexanderPico
commented
3 years ago From Andra:
Here we need to be a bit cautious and might need to resort to a designated wikibase. When we build the pathway bot we tried modelling interactions and reactions, but this turned out to be quite difficult to do. Labels for these interactions and relations can be very ambiguous. e.g. inhibitions/inhibits does have different meanings in a different context and wikidata being a multipurpose knowledge graph the semantics of these interactions will certainly be misinterpreted if we even can get the properties through.
Currently, pathways are modeled as gene sets in Wikidata. Obviously, the interactions between molecules is crucial information that lends itself to semantic graph representation. We could enhance the pathway bot (or make a separate dedicated bot) to import the individual interactions and reactions already contained in the source models. In collaboration with Reactome and Pathway Commons we will ensure the model works for all major sources of pathway information (i.e., GPML, SBML and BioPAX).
Further, even in cases where a completed pathway model is not available, there are evidence-based sources for interactions, including a growing number specific to isoforms and PTM states that can be imported directly.