Open ukemi opened 5 years ago
What we are really looking at is two separate molecular functions that are enabled by related but distinguishable forms of a gene product. Function 1: plain gene product enables ligand binding. Function 2: gene product:ligand complex enables kinase activity. And sometimes there's a third distinguishable function. Function 3: Phosphorylated gene product:ligand complex enables binding of effector molecule.
A really important biological point is that the plain gene product has no kinase activity. Even though it is a domain of the plain gene product that acts as a kinase, it only has the ability to do that when it is associated with ligand.
But the key here is not really the gene product/s or complexes per se. It's the definition of the molecular function as it stands alone in the ontology. The nucleotide receptor activity does have an input of a nucleotide. However, defining it that way is insufficient because there are other receptor activities that will also have nucleotides as inputs. The example in the Reactome GO-CAM model was a receptor tyrosine kinase activity which has an input of ATP for it's kinase function. This is distinguished from which gene products enable the various receptor activities. That's a whole other issue.
OK. I'm trying to assert that a single gene product can have two or more entirely separate functions, and that annotations of the gene product to one of the functions should not spill over into annotations of the other functions. Bacterial flagellar proteins are annotated to locomotion in a bacterial context and to innate immune function in a human symbiotic interaction context.
Same thing here even though the difference isn't as big the distinction is still real. There is also the practical question of whether the different details of the various functions such as whether or not the gene product needs to be part_of a complex could be used as differentia to prevent this spillover.
SIMILAR ISSUE #14823
We decided to explore the idea of using the 'has primary input' relation for these.
I just scanned through the receptors with logical definitions and they aren't doing much work for us at all. The only inferences are for GPC-x receptor being inferred as an X receptor. For the ones that can be, they are asserted as has_part x-binding. This would automatically make x an input, so the logical def ( receptor activity has_input x) is redundant with inheritance from an asserted subclass.
There is a property 'has positive regulatory component activity' with the defintion:A relationship that holds between a GO molecular function and a component of that molecular function that positively regulates the activity of the whole. More formally, A 'has regulatory component activity' B iff :A and B are GO molecular functions (GO_0003674), A has_component B and A is positively regulated by B. It seems like this would make a N&S definition for the receptor terms. Is there any reason why we didn't use these?
It would mean that the acetylcholine binding would both be a part of acetylcholine receptor activity and regulate acetylcholine receptor activity.
Would it make more sense to have something like this? A relationship that holds between a GO molecular function and a component of that molecular function that positively regulates the activity of the other components. More formally, A 'has regulatory component activity' B iff :A and B are GO molecular functions (GO_0003674), A has component activity B, A has component activity C and B positively regulates C.
This would be represented by a property chain: 'has positive regulatory component activity'-o-'has component activity'->directly_positively_regulates?
This way in something like a kinase receptor, the ligand binding would activate the kinase activity, not the receptor activity itself. The kinase activity would then be directly upstream of other MFs in the causal model.
@balhoff, @cmungall and @vanaukenk what do you think?
Hi @ukemi This touches on the compound functions. This is a pattern that @vanaukenk and I cam up with, I wanted to discuss with @thomaspd . The idea was to use the 'regulatory_component _activity' relation that David OS had started to look at:
signaling receptor activity:
transmembrane transporter activity -transports_or_maintains_localization of some chemical -results_in_transport across some membrane
Would that help ?
Thanks, Pascale
Hi @pgaudet. This looks very similar.
@ukemi you have this issue placed in the pathways2go project - https://github.com/orgs/geneontology/projects/24 Is this a requirement for completing that project? I'd like to move it elsewhere if not.
Unless the examples discussed above are breaking otherwise OK version 1 GO-CAM models, I too would like to move it elsewhere. This sounds like something that, once resolved, could be brought back to Reactome and GO-CAM to improve annotations and reasoning, but not needed for now.
Hi @ukemi
If you agree with the pattern
signaling receptor activity:
is a molecular_transducer_activity
-has regulatory_component _activity ligand binding
-directly regulates some molecular_function
I am happy to implement. I discussed with @thomaspd and he agreed. This is one of the outstanding issues for the MF refactoring, I am eager to get that out the door !
Thanks, Pascale
I certainly haven't had time to test it or think about it in great detail. If you are sure that it will work and that the relations you are using for the logical definitions are well-fleshed out with respect to what they mean and how they chain/behave, then it should be ok. ight want to run it by @cmungall and @balhoff too.
The small molecule receptor activity pattern would be:
is_a molecular_transducer_activity
-'has small molecule regulator' some 'small molecule'
-'directly regulates' some molecular_function
Right @thomaspd @vanaukenk @ukemi ?
A nucleotide receptor is defined as a signaling receptor that has input a nucleotide. Kinase receptors have an input of ATP, but they are not nucleotide receptors. We need to modify these logical definitions. Should we convert them all to has_ligand, even though the definition of that property seems to include both receptor agonists and anatagonists?
This was identified as part of the Reactome import project.