Regulation Implementation Plan

Implementation plan:

• Enable reasoner to create implied links. (Chris Mungall and John Day-Richter) DONE
  • Sort regulation cross-products:
    • Molecular function
    • Biological process
    • Biological quality
    • Unassignable
      • Chris has run the reasoner and there is a report in the go/scratch directory that list all of the 'regulation of...' terms where the target of the regulation is not a valid biological process. ftp://ftp.geneontology.org/pub/go/scratch/regulation-of-non-process.txt
• Biologists categorize unassignable (Tanya Berardini, David Hill) DONE
  • David has gone through this list and noted the following.
    • Terms where the 'regulation of...' is a regulation of a molecular function. RoMF
    • Terms where the 'regulation of...' is a regulation of a biological quality.RoBQ
    • Terms where the 'regulation of...' is a valid subtype of another 'regulation of...' term and we probably do not want to make subtypes of the original BP term. Terms like 'regulation of transcription involved in....' are examples of this. The transcription is basically the same, but it is the regulation that differs.
    • Terms where the 'regulation of...' and the BP parent need to be made univocal or there is a missing BP type.
    • Terms that are still questionable. I have noted suggestions about what to do with these.
  • The file with David and Tanya's notes and questions can be found here. Regulation Worksheet
• Rerun reasoner DONE
  • Requirements for reanalysis

1. Do not report any terms that are is_a children of 'regulation of molecular function' or 'regulation of biological quality'.
2. Do not report any terms in Worksheets 2 or 3 where it says 'ok subchild/subtype of xxxx'. This wasn't done, but we have a new parse that has additional lines. Tanya and David will go through this and comment out the lines that are ok.
3. For terms that are is_a children of RoMF, check to be sure that the corresponding MF exists.DONE
4. For terms that are is_a children of RoBQ, check if they correspond to PATO terms. Tanya and David looked into this. It is not straightforward, but we do not think it should hinder the process of getting the regulates relationship into GO.

• Clean up the current GO DONE
  • Export to instantiate implied is_a links.(Chris, Tanya, David)
  • QC check these links and modify as necessary (Chris, Tanya, David)
  • Remove redundant is_a links and QC check (Chris, Tanya, David)
  • Convert regulates relationships back to part_of relationships to be compatible with current GO format and QC. (Chris, Tanya, David)DONE
  • Commit new internally consistent regulation and BP graph back to the live ontology. (Chris, Tanya, David) DONE 1/15/08
• Rerun QC reports to check the new file (Chris) DONE
• Resolve QC issues (Chris, Tanya, David)
• Send out announcement that this change is coming down the pipe (Tanya)
• Final implementation of both new relationships and logical definitions (Chris, Tanya, David)

• Outstanding questions/thoughts
  • For biological qualities, it makes sense to have positive and negative regulation be a part of the parent regulation term, but does this make sense with biological processes? DOES NOT AFFECT IMPLEMENTATION
  • Should we make relationships between the RoMF terms and the processes in which the execution of the functions take part? For example, should 'regulation of epidermal growth factor receptor activity' relate to the epidermal growth factor receptor signaling pathway? If so, how? DOES NOT AFFECT IMPLEMENTATION
  • In some cases, a regulation process can be a bona-fide part of another process, or be a bona-fide type of another process. We need to be careful about which part-ofs are replaced by regulates. Chris has it worked out.
  • Do we want to instantiate all of the implied regulates relationships that are generated by legitimate 'regulation of ...' subtypes? DOES NOT AFFECT IMPLEMENTATION