Ontology meeting 2021-04-19

From GO Wiki
Jump to navigation Jump to search


  • Group members: Pascale, Karen, Harold, Raymond, Peter, Jim, Tanya, Kimberly, David, Chris, Paul, Eric
  • Present:
  • Regrets:


RHEA status and future plans

  • Purpose of using Rhea in GO:
    • Chris: Would rather say the purpose is to leverage the work of Rhea - annotations, mappings (EC, metacyc), reactions, work they have done to describe the reactions.
    • Peter: CHEBI is the expert resource for chemicals; Rhea is the expert resource for reactions.
  • 13,550 reactions in RHEA (= 50000 including quartets) as of 2021-04-12 (11,153 MF terms / 7,000 catalytic activities / xref: RHEA = 4279)
    • How many are mapped to UniProt entries? -> 9,000 approx
    • How many unique Rhea have UniProt annotations 9,125 https://www.rhea-db.org/rhea?query=uniprot%3A*
    • AI We need to identify those with NO GO mappings --->>> next ones to map to GO (or create reactions)
  • Discussion about specificity - for example https://www.uniprot.org/uniprot/O60733 has about 20 reactions; multispecificity transporters are another example
  • Paul: GO will create any specific reaction that we need for a model/annotation
  • Anne: RHEA has generic reactions for some reactions, for example RHA:15801 1,2-diacyl-sn-glycero-3-phosphocholine + H2O = a 1-acyl-sn-glycero-3-phosphocholine + a fatty acid + H+1
  • Concern Pascale - too many reactions for curators to choose from?
    • Anne: at first, used EC. Now annotation is to reactions from papers. Curators search by chemicals, synonyms, 2D structure of chemicals
  • Overall status: number of mapped reactions, how many more to go
  • strategy to complete the GO-RHEA mapping
    • Boomer? Jim/Chris tool to map EC/RHEA/GO/others?
    • Chris: plan: curate GO-> RHEA and get EC from RHEA + maybe use MetaCyc. Maybe treat the RHEA 'quadruplet' as a single term for simplicity?
  • strategy to maintain the GO-RHEA mapping
    • get list of obsolete RHEA reactions at every release so we can update GO
    • Note that Alan Bridge says that about 1000 ECs do not have Rhea reactions & no entry (~12% = unmappable in RHEA)
    • AI: fix incorrect mappings reported by Anne - July 21, 2020
    • AI: fix errors reported on Oct 1, 2019
  • Practical considerations:
    • directionality of reactions: (Pascale) both Alan and Anne confirmed that the directionality of reaction is often useful. If a GO term represents one of the directions (for eg for a kinase), they recommend we map to the directional reaction. For example alcohol dehydrogenase: in yeast it ferments alcohol, in human it is used in the other direction to degrade alcohol in the liver.
    • 1:1 mapping rule - does not always work, examples (see https://github.com/geneontology/go-ontology/issues/20824)
      • ATPases: GO wants to represent DNA helicases, motor activity, etc, RHEA classifies all of them simply as ATPases. Harold: reactions involving macromolecules are a problem for RHEA, because they cannot specific the position at which the reaction occurs
      • transporters: P-type versus ABC types: not all have a RHEA reaction, should we align or create all reactions shown to exist?

  • Rhea2GO annotations:
    • See https://www.ebi.ac.uk/QuickGO/annotations?reference=GO_REF:0000116
    • right now generated by GOA:
      • entries with Rhea mappings (applied during UniProt entry curation - or by RHEA?) automatically get the GO term using a script by GOA/Alex (similar to EC2GO or InterPro2GO)
      • can groups generate their own mappings? how do they get the Rhea mappings to their entries?
      • (probably a question for Alex, although we can provide specs): if an entry is mapped to a directional GO, and there is no direct GO mapping to that reaction but there is one to the non-directional parent, is the Rhea2GO mapping applied to that entry?
    • redundancy concerns (Helen and Steven, Flybase):
      1. What is the source of the Rhea annotations? I assume it is UniProtKB, and then I’d guess it's a combination of manually assigned Rhea IDs to Swiss-Prot entries and computationally -assigned Rhea IDs to TrEMBL entries. Right? If so, how are the computational Rheas assigned? If they are inferred based on EC numbers, we haven’t gained much!
      2. This leads to: what is the degree of overlap of Rhea2GO annotations compared to the current EC2GO? Presumably we can compute and look at that for a given species/set of proteins. One would expect/hope for a large degree of overlap if we’ve got the EC/Rhea xrefs aligned well in the GO. Any discrepancies would either reveal missing xrefs in the GO, or catalytic activities that are included in EC or Rhea but not the other.

Other items


  • Filter out Rhea xrefs that are not valid (Jim)

Guidelines for URLs/PURLs

Always use standard URLs.. maybe should use PURLs - see https://github.com/geneontology/go-site/pull/1671

What kinds of MF terms would be appropriate for the multi-drug resistance transporters.

Karen I’ve followed the line of thought that you and Paul suggested at the Jamboree, and would like to see what you guys think before I go any further.

The two relevant tickets are:

  • Decision (March 22): 'broad specificity transporter'

Can protein complexes be is_a cellular anatomical entity?

See ribosome question, https://github.com/geneontology/go-ontology/issues/21143

Also, if we allow that, how does that impact ShEx? Do we already have cases like this? What default relationship gets added?

  • Decision (March 22):
    • Risobsome is_an organelle
    • polyribosome ?
    • how to make the distinction?

Other Business