Cell-type specific GO terms

From GO Wiki
Jump to navigation Jump to search

Call Details

  • Thursday 21st April
  • 5pm BST (UK)
  • 9am PDT
  • 11am CDT
  • noon EDT

GO conference line.


There has been an increase in the number of requests for cell-type specific GO process terms. E.g.

This has led to one main question:

  • Q. Should these type of terms be added into the ontology?

And a set of related questions:

  • Q. Would this information be better captured using column 16?
  • Q. If so, should we continue to add the terms in until all/most groups are using column 16 for anntoating?
  • Q. Once column 16 comes in, will the cell-type specific terms remain in GO, or be merged into the more generic parent?
  • Q. Should cell-type specific terms be added in, if they have parentage in another node (e.g. nurse cell apoptosis ; GO:0045476 has oogenesis parentage, also).

The call is for ontology editors to decide whether to add these terms in, and limits for doing so.

Reasons to add the cell-type specific terms into GO

  • Most groups are not using column 16, so not adding in the terms will result in a loss of annotation data.
  • Most tools can not search on column 16 information, so can only search on these terms if they are pre-composed.
  • The mechanisms of E.g. chemotaxis/apoptosis etc may be different in different cell types.

Reasons not to add the cell-type specific terms into GO

  • GO could expand hugely if we add in terms for every cell type.
  • GO would need to keep track of changes in the cell ontology so that we are aligned.

Chris says

  • terms such as "X differentiation" should continue to be pre-composed. Use judgment in deciding granularity.
  • we should forge ahead on getting bp_xp_cl into go_editors (needs some work on relations). Then we can set up a TG template to get these terms on demand.
  • don't describe the experiment in c16 - e.g. no "in vivo cell line cell".
  • using c16 to state redundant info is redundant but harmless. E.g. if you annotate to "chondroctye differentiation" then it's redundant to add "chondrocyte" to c16
  • we should come up with examples where we definitely don't want to pre-coordinate using CL. E.g. "A metabolism in cell type B"


Cross Product Reading


In particular:

  • Tables 2 and 3
  • Sections 2.2 and 2.3
  • Section 3.3. On pre- and post-composition

What's presented in this paper describes in fact a subset of the different roles that cells can play in biological processes. It doesn't address for example signaling. Where we have a process with some instance of cell types A sending a signal to cell type B, and we want to indicate both A and B post-compositionally, we need to specify the role/relation, otherwise we lose information.


  • 1/ Decide if these terms should be pre- or post-composed.
  • 2/ Decide on granularity wanted.
  • 3/ Will the rules change once column 16 is in use by all groups?
  • 4/ Is there a way to monitor changes in the cell ontology to ensure GO is in sync?

Meeting Minutes


  • Chris
  • Heiko
  • Rebecca
  • Midori
  • Amelia
  • Laurel Cooper
  • Alex
  • Terry
  • Tanya
  • Chris M

  • Chris gave a presentation (see above pdf) on the formal definitions that are required for column 16, and discussed how to retrofit the data and notes that MGI have been capturing in column 16 so far.

  • The take-home point of the presentation is that post-composition is good if it enhances pre-composed terms but post-composed terms wouldn't replace pre-composed terms.
  • Formal relations are essential to state what relationship the ID in column 16 has to the GO term.
  • GO has a precedence for pre-composed terms.
  • Chris wouldn't advocate obsoleting existing terms.
  • Post-composition is most useful for capturing information that shouldn't be put in GO terms: for example chondroocyte development that occurs during stage X of development.

  • We may require a union of the cell type ontology and plant ontology for cell terms.
  • Chris, David and co need to work through some of the existing MGI column 16 data to formalize the relationship to the corresponding GO term.
  • We should continue to add in GO terms for cell-type specific terms, using our judgment for how specific to go: if there are hundreds of types of neuron, then we probably don't want differentiation terms for every type.
  • Alex commented that pre-composed terms are also useful because the mechanisms of processes can be different between cell types.
  • Heiko will work on TermGenie, so that TermGenie can take care of many of these requests.