GO-CAM Working Group Call 2018-07-24

From GO Wiki
Jump to: navigation, search

Meeting URL





Modeling Transcription in GO-CAM

  • Sabrina - PMID:28687631 'Clock1a affects mesoderm development and primitive hematopoiesis by regulating Nodal-Smad3 signaling in the zebrafish embryo.'
  • Kimberly - PMID: 28578929 'Morphological Diversity of C. elegans Sensory Cilia Instructed by the Differential Expression of an Immunoglobulin Domain Protein.'
  • Kimberly - unc-86 regulates transcription of mec-3

Relations between MF and BP

Relations between MF and BP (from previous call)

  • We discussed models of transcription factors and their relation to transcription BP terms
  • Transcription factor activities are modeled as 'part of' a regulation of transcription process
    • Why? In the ontology, regulation is modeled is a biological process, and so functions need to be part of biological processes that regulate other processes.
    • This can be captured succinctly in models by stating that a transcription factor activity is part_of '(+/-) regulation of transcription'.
    • This will also produce the correct annotations in the GPAD output file.
    • If a transcription factor activity is connected to the process of transcription with a regulates relation, the reasoner cannot then use the logical definitions of the process terms to infer that the transcription factor activity is 'part_of' a regulatory process.
  • Transcription factor activities are 'part of' transcriptional regulation
Figure 1. Transcription Factor Relation to BP
  • Regulation of transcription is logically defined as a biological_process and (regulates some 'transcription by RNA polymerase II')
Figure 2. Equivalence axiom for regulation of transcription - Protege
  • Positive regulation of transcription is logically defined as a biological_process and ('positively regulates' some 'transcription by RNA polymerase II')
Figure 3. Equivalence axiom for positive regulation of transcription - Protege
  • Negative regulation of transcription is logically defined as a biological_process and ('negatively regulates' some 'transcription by RNA polymerase II')
Figure 4. Equivalence axiom for negative regulation of transcription - Protege

General Rule

  • MFs and BPs are linked by part_of (or other relations, e.g. acts upstream of, acts upstream of or within) but NOT by regulates relations

Relations between MF and Input(s)

  • has_input vs has_direct_input
  • Improve documentation on 'has input' and subproperties, deprecating unnecessary terms
  • obsolete 'has direct input' ?
  • has_direct_input has been used much more widely in conventional annotation for MF extensions
  • has_input has been used less widely, but how is it used?
    • select participants in catalytic activities
    • metal ions (e.g. calcium-dependent protein binding)
    • drug binding
  • Is there a meaningful distinction between these two relations for MF? What are we really trying to capture with MF inputs?
  • Proposal: review MF annotations using has_input
    • This would be good information to have in light of the Relations Ontology work that is currently being done

Relations between BP and input(s)

  • Duplicating has_input for MF and BP results in multiple entries in the AE field of the BP annotation in the GPAD

Relations between MF and Inputs (from previous call)

  • We discussed the relation to use when capturing the input of the transcription factor.
  • Even in conventional annotations, groups have differed on how they do this.
    • To illustrate, groups have used: has_input, has_direct_input, has_regulation_target
  • For MFs, is there any real difference in meaning between has_input and has_direct_input?
    • No, an input to an MF is, by necessity, a direct input to the activity.
  • For BP, the use of has_input may mean something different, though.
    • For BP, curators have used has_input to indicate the entity that is affected by a process but for which the precise MF that directly mediates that effect is not known.
    • As an example, Rob described how SGD has annotated IDA vs IMP for protein kinase activity and phosphorylation.
  • The curation task of indicating inputs for MFs vs BPs needs to be clear for GO-CAM models.
  • Rob was asked to create a GO-CAM model for the phosphorylation example.
  • Sabrina's paper from last month also touches on this; in one case, the evidence for direct regulation of transcription was supported by several different pieces of evidence, whereas in the other cases, there wasn't sufficient evidence to assert the direct regulation. We need to re-visit this model to make sure we have a satisfactory way of distinguishing between these two cases. This will likely be a common occurrence during curation of transcription papers.

Relations between BP and MF of transcriptional target


  • On call: Bob, Dave F, David H, Dmitry, Edith, Giulia, Harold, Kimberly, Laurent-Philippe, Pascale, Paul T, Petra, Sabrina, Seth, Stan, Suzi A, Suzi L (others -please add your name to the list if you joined later)

Transcription Models - Decisions Made

Relations between MFs and Regulatory BPs

    • Transcription factor activities are related to regulation of transcription BP terms using the 'part_of' relation
      • This is consistent with the use of relations in the ontology and produces the desired annotations in the GPAD output
    • Note that this is a general guideline that will apply to any MF that is 'part_of' a regulatory process
    • A consequence of this is that any regulation terms needed for annotation will have to be instantiated in the ontology

Capturing Direct Regulation

  • Direct regulation of a TF on the expression of a gene may be captured by adding the gene as input to the TF activity, e.g. has_input 'gene x'

Capturing Unknown Mechanism of Regulation

  • If it is not known if the TF directly regulates the expression of a gene, then the input for the TF activity is left blank.
    • In this case, however, it is okay to use evidence from another experiment that might have shown different context (i.e. a different gene was regulated) as supporting evidence for the TF activity.
  • The curator can model the unknown mechanism of regulation by saying that the TF is part_of regulation of transcription that is causally_upstream_of_or_within the positive or negative regulation of transcription that ultimately controls the expression of the gene. The gene is then added as 'has input' to the most distal transcriptional regulatory process.

Root Node vs Existing Molecular Functions

  • Curators should always try to construct models using the known MF of a gene product, even if that MF was not specifically demonstrated in the paper they are annotating.
  • Associated evidence for that MF will always point back to the paper in which the MF was interrogated.
  • Creating models in this way will allow us to build on existing knowledge to create the most comprehensive and up-to-date model for a given BP.
  • If a gene product has more than one MF, curators should use the biological context of the annotated process to select the most appropriate function(s) for that gene product.

Transcription Models - Decisions Outstanding

  • has_input vs has_direct_input
  • Most, but not all, conventional MF annotations have used 'has_direct_input' for contextual information
  • GO-CAM models, and the form editor, however, use has_input
  • We need to converge on a consistent way to capture MF context
  • Note that the ongoing RO work in which this topic is being discussed may have a significant impact on what we ultimately decide to do here
  • In the meantime, what should we do?
  • One thing we could do is review MF annotations that used 'has_input' to determine what curators were really trying to capture with 'has_input' for MFs

Screenshot of Model used for Discussion

Sabrina model 2017-07-24.png
  • The top part of the model shows alternative representations of the relation between the TF and transcription.
  • The part_of relation is the agreed upon correct way to model the relationship between the TF and regulation of transcription.
  • Note that this model is still under discussion and not the final version; this screenshot represented the model as of 2018-07-24.