[Ontology-editors] What does the MF ontology represent?

[David Hill]

Hi Chris,
>
> BFO makes the (I think) very useful distinction between a function and 
> a process/occurrent. Functions are realized as processes. The 
> canonical example being a sperm cell - it has a function even if that 
> function is never realized. bfo:Functions are subtypes of realizables. 
> Another word for them might be potentials.
>
> At first it seemed natural to treat the MF ontology as bfo:Functions.  
> However, it transpired that GO folks don't really think of MFs as 
> bfo:Functions. This lead to a lot of confusing talk between GO and 
> other OBO folks. At the last RO meeting, David, Tanya, Suzi and I 
> decided that the cleanest solution is to simply state that MF is_a 
> bfo:Process. Note that this is implicit in our decision to use part_of 
> / has_part to link MF/BP
I still think this is the best solution. Let the potential be captured 
in the annotation.
>
> Note that so far we are not talking about annotation, just the meaning 
> of the terms themselves. When we make an annotation we are perhaps 
> saying something about function or potential in the bfo sense - not 
> every gene product realizes the function evolution honed it for. That 
> goes for both MF and BP.

Absolutely. But, from a GO perspective, we care about the execution. 
This may differ from someone who cares about the role of a gene product 
in for example a disease. In the latter case, the information about the 
potential may be very valuable.
>
> If GO is to be part of the OBO Foundry it has to declare what the 
> appropriate bfo superclasses of its roots are such that relations can 
> be made across the foundry coherently. I would like it to be formally 
> stated that MF is_a bfo:Process. This would quell Alan's concerns 
> (below and see also gofriends).

Yup.
>
>
> Of course there are some terms in MF for which it makes little sense 
> to state that it is a process - protein tag, structural constituent 
> etc. I propose we acknowledge that there are some errors and we will 
> work to resolve these.
Jane, Midori and I talked about these types of terms in Cambridge, and 
Tanya and I have talked about them as well. There are not many of them. 
Are they really useful?
>
> I think the eventual solution will look pretty much exactly what 
> Amelia outlined in St Croix. It may not be as radical as moving most 
> of MF into BP (though I am not against). It could involve splitting MF 
> into a process sub-branch and a real-function sub-branch (logically 
> this is equivalent, but it feels less radical as terms stay in the 
> same namespace for the most part, though this is somehow less satisfying)
I would like to keep function and process separate for the practical 
reason why they were first made distinct. That is the same function can 
play a role in multiple processes. It's just easier to think about if we 
think about what molecules do and their overall objective. I realize 
that they are in fact not separate, but the artificial boundary makes it 
easier to think about. I'm not sure it should be our (GO's) role to work 
on potential. That type of info can be inferred from annotations. Note 
that when I say separate here, I just mean conceptually separate. I am 
all for creating the links between MF (molecular process) and BP. I 
guess what I'm saying is I don't really have a problem with a function 
being a BP at the molecular level of granularity. I think we have always 
thought of it that way.
>
> The function sub-branch would look like this:
>
> antioxidant
> energy transducer
> enzyme
> motor
> nutrient reservoir
> signal
>  chemoattractant
>  chemorepellant
>  hormone
>  ligand
>  morphogen
>  protein tag
> structural molecule
> receptor
> regulator
>  chaperone regulator
>  enzyme regulator
>  transcription regulator
>  translation regulator
> transporter
> toxin
> triplet codon-amino acid adaptor
Actually, I think we need to look closely at these. I think with the 
links between function and process, we might be able to dump some of 
them. The most obvious are transcription regulator activity. This just 
means 'executing some unknown function that is part of the process of 
regulation of transcription'. As Tanya and I began to look at the F-P 
regulates links we noticed ones like this that seemed to be just 
crutches because we didn't have links between F and P.
>
>>
>>
>>     • calcium channel regulator activity (MF) regulates calcium 
>> channel activity (MF)
>>
>>
>> The second one seems completely wrong. At best one can have the 
>> realization of a function regulating something.
I disagree. This means that the function that is executed by the gene 
product modulates some inherent quality of the execution of  the 
function of a calcium channel, such as the rate at which it moves 
calcium across the membrane. It's function could be to plug up the 
channel, or to open it.

Tanya and I just returned from the Society for Developmental Biology 
meeting and we heard talk after talk after talk about processes that 
regulate developmental processes. In my opinion, we can't ignore this 
view (Tanya feel free to pipe up). In fact during one talk someone 
specifically said we know that molecule X regulates process Y, but what 
don't know how. Of course what we want to know is what X is doing to 
regulate Y.


David
>