Annotation Conf. Call 2017-01-10

From GO Wiki
Jump to: navigation, search

Bluejeans

  • https://bluejeans.com/993661940
  • to join via phone: enter Meeting ID 993661940 after dialing in
    • from US: +1.408.740.7256 or +1.888.240.2560
    • from UK: +44.203.608.5256
    • from Switzerland: +41.81.588.0256

Agenda

Next Consortium Meeting

  • June 1-3, 2017
  • The regular meeting will be followed by two optional workshop days on LEGO and Reactome curation on June 4 and 5
  • Pankaj Jaiswal has graciously offered to host the meeting at Oregon State University in Corvallis, Oregon
  • He has set up a really nice meeting webpage at https://sites.google.com/view/goc2017
  • Please register!

Review of Jenkins Pipeline

View GAF builds in Jenkins

Working Groups

LEGO Modeling Discussions

  • We would like to reserve the second Tuesday annotation call of each month for discussion of LEGO modeling.
  • The proposal is to have two presenters each month for an ~30 minute discussion of their model.
  • The goal is to get more people comfortable with LEGO modeling and increase the number of processes represented by LEGO models in GO.
  • Please sign-up for a time slot here: LEGO Discussion Rota
  • We would like to do this on a volunteer basis, so please consider taking a turn.
  • We all benefit from discussing modeling issues from across a broad spectrum of organisms, and there is likely commonality amongst the pathways and processes that we are trying to represent.

Minutes

On call: Alice, Chris, David H., David OS, George, Giulia, Harold, Helen, Jim, Karen, Kimberly, Li, Melanie, Midori, Moni, Pascale, Penelope, Petra, Rachael, Ruth, Sabrina, Stacia, Stan, Tanya, Terry, Val

Regrets: Paola

QC Pipelines

  • Two pipelines - Jenkins and Mike Cherry's checking scripts
  • Focus of this call - Jenkins pipeline
    • GAF checks have been failing - red balls
      • Due to an upstream failure in ontology build, was not being caught, but affected GAF check pipeline
      • This was an edge case, but has now been addressed such that the inconsistency check will happen upstream of the GAF check
  • Annotation rules
    • See github link
      • Rules include those from Mike's original checking script, tool-based rules (e.g., Protein2GO), Jenkins rules
      • Rules were in XML, and a web page documented each rule in a human-readable format
      • Replacing system with a markdown and yaml file
      • Still need a page for human readable review (script will pick up the content of the md file for display)
      • These files are to be used for software developers to implement
      • Working on universal implementation of all of these rules on top of the graphstore that Jim is implemeting
  • Jenkins annotation inference pipelines
    • Inter-ontology links and annotation deepening
    • Evidence codes in general
      • First implementation used IC evidence code
      • But, now preserving original evidence code and using GOC as contributed by entry
      • However, these annotations may really be the result of a combinatorial evidence chain
    • Annotations from deepening
      • Two main issues:
        • When IC was used, primary GO ID was in the With/From but new annotation was from extension
        • Also some relations, such as regulates_o_occurs_in, if not used correctly will result in incorrect annotations
        • These annotations will need to be fixed to get the correct inferred annotation
        • How many annotations have actually used this relation?
        • Need examples of each of these relations in LEGO models - DOCUMENTATION
          • Need distinction between has_regulation_target and the chained relations that also use regulates_o
          • Protein2GO now has the options to use the chained relations, but there isn't sufficient documentation about how and when to use them!
          • Annotations may be correct, but the inference is wrong
          • See the MAPK example in the GOC-OWL inference github ticket
      • ACTION ITEM: Make sure that annotation extensions in tools use the more human readable relation name
      • Evidence code - no longer will use IC; will propagate original evidence code
        • Logically, the annotations are implied by the inference.
      • Qualifiers in deepened annotations
      • contributes_to, normally only used for MF, was carried over to a CC annotation
        • this is a straightforward bug that can be fixed - no inference over contributes_to from MF to CC
      • Annotation redundancy - will implement check to remove redundant annotations from annotation deepening
      • Regulates_o_occurs_in
        • ACTION ITEM: Look at annotations that use a regulation term and occurs_in in the AE to check on these annotations to make sure annotation deepening is correct