search

NCATS-Tangerine / rhea-beacon

Knowledge Beacon wrapper for the Rhea Annotated Reactions database (https://www.rhea-db.org)
MIT License
0 stars 0 forks source link

The conceptual design of the Rhea beacon may need some refinement based on its nature #7

Open RichardBruskiewich opened 5 years ago

RichardBruskiewich commented 5 years ago

Rhea is an unusual resource in that each entry is, in fact, seems to be a reaction. The underlying concepts behind each reaction are chemical compounds, primarily indexed in CHEBI. It could perhaps be argued that the CHEBI entries are the real beacon concepts lurking in this knowledge source, and that the Rhea entries themselves are the "statements". In fact, one wonders if one could somehow return meaningful subgraphs (rather than single statements) that are tagged by Rhea 'statement' identifiers(?). In this spirit, classical beacon 'statements' would only be a degenerate, trivial form of statement (two concepts plus a single edge)?

Does this do violence to the concept of a beacon?

cmungall commented 5 years ago

I think shortcut edges are fine, but I think the reactions should be nodes too

E.g.

   R
 /   \
C-----C

Also it looks like you are modeling ECs as gene products, GO treats EC, RHEA and GO-MF as biolink molecular activities, in equivalence cliques

RichardBruskiewich commented 5 years ago

Roger that @cmungall.

I think the Biolink model (of concepts) accepts multiple categories now, right. So tagging EC's both as gene products (perhaps even gene?) as well as molecular activities, would likely be both feasible and advantageous?

I think after this particular issue was posted, we concluded that yes, include R's in the concept lists exported. Just need to properly tag them(?). Do they have type aligned with EC's, but only "molecular activity"?

cmungall commented 5 years ago

There is no explicit disjointness axiom in biolink, but this could still be confusing

On 9 Nov 2018, at 8:23, Richard Bruskiewich wrote:

Roger that @cmungall.

I think the Biolink model (of concepts) accepts multiple categories now, right. So tagging EC's both as gene products (perhaps even gene?) as well as molecular activities, would likely be both feasible and advantageous?

I think after this particular issue was posted, we concluded that yes, include R's in the concept lists exported. Just need to properly tag them(?). Do they have type aligned with EC's, but only "molecular activity"?