Open ValWood opened 7 years ago
I don't follow. GO annotation should be only for the normal function of a gene product in normal process - whether the gene product involved in a process, regulating it or causally upstream of it.
You could choose to use GO term as part of your phenotype annotation (as PomBase does indirectly via phenotype terms) and casual chains might possibly be useful there - but this should be kept separate from normal GO annotation.
Well at PomBase we do, but lots of other groups who do not have a phenotype ontology want to annotate "indirect causally upstream" events to processes when they are non-regulatory. I am encountering quite a lot of examples from Term Matrix Project.....
So take "DNA replication" as an example of an upstream process. If DNA replication occurs normally, chromosome separation is unaffected. DNA replication does not normally regulate when chromosome segregation occurs, DNA replication occurs in S-phase and the subsequent transitions are regulated. Chromosome separation is tightly regulated during M-phase based on spindle checkpoint being satisfied. However DNA replication is "causally upstream" of "chromosome separation".
Would we expect "DNA replication" gene products to be annotated to chromosome segregation with a "causally upstream" relation or qualifier, even though chromosome segregation is not "regulating" chromosome separation? I would argue not, but if some resources want to keep these annotations we need a way to separate them with a specific relation which means "causally upstream, not regulating and not involved in".
The DNA replication/chromosome separation is a somewhat obvious example, but it clearly illustrate the issues I am encountering from the annotations of indirect effects from pleiotropic mutant phenotypes. Sometimes it is not clear if an effect is indirect or regulatory (although based on biological knowledge it is often possible to say that this is causally upstream but indirect).
This is a recurring issue of annotating processes based on "general" phenotypes where a mutant can affect numerous downstream processes indirectly. In a normal cell without a compromised gene product many homeostatic mechanisms and feedback loops prevent certain processes which supply substrates for onward processes from exerting a direct regulatory effect. However, their manipulation/mutation can be associated with many abnormal consequences (pathologies).
If these annotations are valid, distinguishing them from involved_in/and regulates is critical to ensure that these annotations are not used in mappings or phylogenetic inferences.
I still don't think we have clarity on what we mean by the concepts of indirect/direct/causally upstream. This is the ultimate purpose of these tickets.....
Pulling out of https://github.com/geneontology/annotation_extensions/issues/75 so it does not get lost
Which relationship can be used to annotate things which are "indirect effects" and not part of a process or its regulation. It seems critical that we have the ability to keep normal processes, and annotations which describe when things go wrong but would not regulate a process in a non-pathological situation completely separate. This does not seem to be the case at the moment because "causally upstream" sits above some normal processes.
I can provide illustrative examples if the question is not clear
@ukemi @vanaukenk @cmungall @dosumis