pgaudet
commented
4 years ago Thanks @cmungall
It would be nice to really be clear on how we handle phenotypes. This may be something useful but it certainly makes a lot of noise in GO annotations.
I also don't understand 'response to drug' - other than detoxification, other responses are phenotypes/side effects, unlikely to be 'evolved'.
But where should we house those annotations? It seems there is a need. One option would be to allow to capture phenotypes directly, using different relationships (for example, in PMID:18032670 we could annotate that cocaine impairs GO:0007268 chemical synaptic transmission, or something along those lines. Annotating with different relations would allow us to only display 'normal functions' while still capturing phenotypes.
(it's also very strange that response to cocaine is an 'adult response').
I'll remove the IBAs. (Generally we shouldn't have IBAs to 'responses' terms).
ValWood
We have a hierarchy:
(some MGI annotations highlighted)
I have Questions
First, I think the two subclasses give false precision. In What sense are the dopamine receptor genes like Drd2 involved in behavioral response, but not a cellular response? I think we are annotating the phenotype here. We already have an evidence code for IMP vs IDA, let's not make false precision in the ontology.
A general gripe: we have over-used the rule about annotating with as much specificity as you know and no more. Logically this is fine under an open-world assumption, but users don't make the OWA!!! If people come in with a query for "cellular response to cocaine" they will get a massively under-represented set. If we provide a term B that is a subclass of A, then it behooves us to be able to be able to do reasonable job of making annotations more specific than A.
Second, does this even belong in the ontology and is it correct?
Did Drd2 evolve to respond to cocaine (either 'at the cellular level' (what other level??) or at the 'behavioral' level)? Is the organism genetically programmed to respond to cocaine? I don't think so. Yet, the definition of BP is A biological process represents a specific objective that the organism is genetically programmed to achieve. And we have annotated a PAINT ancestral node to BRTC!
Regardless of whether it is programmed to or not, is Drd2 (or other dopamine receptors) actually involved? I think there is an interesting story here. cocaine is able to exert part of its behavioral and cellular effect by elevating dopamine levels in the striatum https://www.ncbi.nlm.nih.gov/pubmed/19138563/ . It achieves this by binding to and inhibiting the presynaptic dopamine transporter https://www.ncbi.nlm.nih.gov/pubmed/18568020/ - so it doesn't seem appropriate for GO to annotate how chemicals disrupt parts of biological pathways... unless the organism has evolved a defense
However, another mechanism is via binding a receptor heteromer D1 and the σ1-receptor https://www.ncbi.nlm.nih.gov/pubmed/20956312 and this may be true for D2 as well... this does seem like a valid GO annotation, but to a complex.
But none of the biological story comes through in GO, just some over-annotated gene sets at inconsistent levels of granularity.
My recommendations: