questionable complex grouping term

[dosumis]

From @ValWood on October 21, 2015 20:33

This seems an odd grouping term, It seems to be mirroring the function ontology to group Any complex with transferase activity which would be a lot of different complexes!

transferase complex, transferring phosphorus-containing groups

we already have 110 annotation to this term from completely unrelated complexes....

Copied from original issue: geneontology/go-ontology#12116

[dosumis]

From @ValWood on October 21, 2015 20:38

similarly 'membrane protein complex' has 230 annotations (this one is probably Ok as at least it is classified as a component.....)

[dosumis]

From @bmeldal on October 22, 2015 7:22

It mirrors the activity GO:0016772 transferase activity, transferring phosphorus-containing groups and has capable_of relationship to it. It has been created as a grouping term to potentially strip all leaf complex nodes. I have done this a lot in preparation of the pruning exercise we are now discussing. @dosumis, is this going to far/unspecific? I used the template for the most specific activity I could find for the complex.

[dosumis]

Nothing logically wrong with it. Whether it is too abstract is down to whether it will be too confusing to users - i.e. a judgement call, rather than something we can write rigorous rules for.

I think 'membrane protein complex' is clear and useful. This term is less so and feels to me to be on the cusp of what is acceptable: A general term for complexes with transferase activity would be too far; kinase complex is probably OK.

[dosumis]

From @ValWood on October 22, 2015 10:3

OK. I agree membrane protein complex is useful

transferase complex, transferring phosphorus-containing groups feels a bit too abstract to me .....It feels it would perhaps be better to have some CC-MF link at the leaf node at the level of the individual complex rather than making CC grouping terms for every MF?

[dosumis]

It feels it would perhaps be better to have some CC-MF link at the leaf node at the level of the individual complex rather than making CC grouping terms for every MF?

We have these too. They are used to automatically classify under general classes like this.

If we go ahead and strip out lead nodes defined by composition, then we will use a similar system of inference to automatically add annotations to protein complex terms to IntAct complexes based on their manual annotations to MF terms.

[dosumis]

From @bmeldal on October 22, 2015 10:51

It feels it would perhaps be better to have some CC-MF link at the leaf node at the level of the individual complex rather than making CC grouping terms for every MF?

But the leaf notes might be going... :(

[dosumis]

From @ValWood on October 22, 2015 10:56

Maybe my problem here is the general nature of the grouping term.

Perhaps the functional grouping terms should be as specfic as possible "protein kinase complex" makes more sense to me as a grouping term than "transferase complex, transferring phosphorus-containing groups"

I'm guessing that this would make more sense to the average user.

[dosumis]

From @ValWood on October 23, 2015 10:31

Another issue with functional grouping is that it needs to be applied consistently For instance, there are many , many protein kinase complexes, and more transferase complexes that are not currently housed under this term In 2 minutes I spotted: chromosome passenger complex nuclear pre-replicative complex holo TFIIH complex bub1-bub3 complex (all protein complexes with protein kinase activity)

The complications from implement might outweigh the benefits...... its not good to add grouping terms without collecting all existing child terms as descendants when they are added.....

[dosumis]

Another issue with functional grouping is that it needs to be applied consistently

I agree. But we're a long way down this route to easily turn back. In the current model, all of this is based on auto-classification via capable_of relationships to function terms. In the proposed new model, a GAF inference job will add inferred annotations to terms like this based on manual MF annotations to complexes. Either way, curation effort is needed.

[dosumis]

Hi Val,

I think the way forward might be to collect a list of cases like this. The protein complex WG can discuss appropriate levels of generalisation for functional/location based complex grouping terms. Need some general guidance and perhaps a mechanism for flagging 'too general' classes. Otherwise this will be an ongoing issue and a time-sink for everyone involved.

We could start a list on the PC-WG gihtub site.

• David

[bmeldal]

Val, the lack of autoclassification for old terms without capable_of relationships bothers me too. It looks like David can see a way by using the manual annotations. A solution is definitively needed and your input very useful!

I'm teaching IntAct in Cambridge on Friday morning. I'm free from 1 pm (I think) if you want to catch up in person. Otherwise, I know you are at EBI regularly so we could catch up. Sometimes it's easier face-to-face over a coffee :)

Sorry for the few days silence, I had some time off with the kids and there was no opportunity to catch up with work til now! Back in the office tomorrow morning.

Birgit

[ValWood]

Hi Birgit. I do plan to come to EBI this week as I haven't been for a while. Are you around tomorrow or Thurs? If not I can probably do Friday p.m. Not sure yet as I have a visitor arriving at some point...

val