Manager Call 2019-04-24
- Agenda: Kimberly
- Minutes: Huaiyu
Calls after the GO meeting
- We propose to cancel all GO calls the week after the GO meeting
- Also, if there are 5 weeks in a month, we could also cancel calls on that last week
GO meeting recap
- Review action items: https://docs.google.com/document/d/1pphkeawGyfQUgS7psrrA1H3HDocSXhDLm4reHbegn50/edit#
- Release : happened on March 19th; and April 18 - are there issues we should be addressing ?
- noctua-dev is being used for testing imports, but it is often really slow for loading or editing/creating models. Is there anything that can be done to improve the performance?
- Once all of WB and MGI models are loaded onto noctua, will we have the same issue?
- GO-CAM versus an annotation: we need to clarify what GO-CAMs are: (connection of at least 2 activities), versus lists of annotations (like the XenBase phenotypes, for example). We need to allow both usecases.
- One possible way to deal with this is to have MODs annotate 'normal' annotations, and GO Central have curators doing models (based on MOD annotations if appropriate)
Seth and Dustin trip to BHB
Discussion points from GOC meeting
- Helpdesk email: Seth + Managers: work out a long term solution for the helpdesk
- Authoritative groups: Managers: do we need to change the policy ? For example SynGO annotated mouse and rat; MGI submits the SynGO annotations but not RGD (and anyway GOC would prefer to do the integration itself).
- Documentation is here Authoritative_Database_Groups (date unknown; moved from the old GO website)
- Can we produce human-readable (Error) reports if something goes wrong with the pipeline - for example missing files (for example right now uniprot-all is missing)
- Annotation and Ontology stats: it would be nice to track changes in the ontology and annotations by looking at the differences between releases. Can resources be assigned to this ? See also https://s3.amazonaws.com/geneontology-public/go-stats.tsv and https://s3.amazonaws.com/geneontology-public/go-last-changes.tsv (Laurent-Philippe) (info about which terms have been obsoleted, which are the new terms, etc).
- The QC group can provide specs if there are resources to implement it.
Priorities until next GOC meeting
- GO rules: priority - do we have resources now ?
- Alex would like to know how rules also apply to other resources such as Protein2GO
Discussion points carried over from previous meetings
Meeting with MOD developers to go over GO pipeline
Plan to durably solve those issues (last week we had timeout, this morning internal server error) ?
- Reference: https://agr-jira.atlassian.net/browse/AGR-1454
- Current strategy: server api.geneontology.org up and running for more than 2 weeks. Health checks every 30s, if fails, sends sms + email notification + auto relaunch the server after 1m30. BioLink timeouts of 4s are increased to 15s as it was causing issues and the id conversion is now handled by mygene (Deepak fix). GOLr would gain to have a docker or VM that could also auto relaunch in case of failures and scale on demand.
- We need an SOP for how to handle annotation (or other?) files that have pipeline-blocking errors that can be fixed by GO
- Is it okay to fix these errors?
- If yes, then is it okay to point to a GO version of the file in the source URL in the datasets.yaml file until the source file is fixed?
GO-CAM models GPAD outputs
- Should Kimberly and David plan to go to visit Jim some time this spring to focus on the GPAD outputs generated from GO-CAM models?
- Annotation extensions
- removing regulates_o_partof chain
- Spatial relations
- chemicals at pH7.3
Fate of "unmanaged" external2go files
- Ontology group would like to archived them since we don't have the resources to maintain them and no one is maintaining these (e.g. mips is deprecated).
Must list which projects are not maintained anymore, and at least simplify the list. Waiting from Paul review. It seems tigr will be maintained by NCBI.
external2go svn link: http://viewvc.geneontology.org/viewvc/GO-SVN/trunk/external2go/
Also looking at "Genome Properties" for GO mappings. Chris will put some slides on this for microbial annotations, possibly another grant to have sufficient resources.
Note that there has been a discussion of why the Reactome file hasn't been updated. https://github.com/geneontology/go-site/issues/1031
New IEAs (not InterPro2GO)
We decided earlier to not accept IEAs from outside groups, as all sequences would be run through the dependable InterPro2GO after uploading- but this group/method *seems* like they have more accurate annotations and higher coverage.
Long discussion last week - What's the next step ?
- I created a github project stub for this - Kimberly
- We were going to set up a call - has it be scheduled?
- Chris was going to review current specs, particularly wrt SO/MSO
- Discuss expanding the gpi file to include other entities used in annotation, particularly the With/From field. e.g. variation IDs, RNAi experiment IDs, etc.
- This has come up in discussions with several curation groups, including MGI, WB, and Zfin
- How does the GO gpi file work fit in with Alliance data files?
- Not all GO contributors are part of the Alliance, but it would be nice to have standardized file formats for other entities.
Judy, Pascale, Huaiyu, Suzi, Kimberly, David, Chris, Seth
GO Conference Calls
- Only the calls in the week after the GOC meeting will be canceled.
- Applies to all the GO calls, but not alliance calls.
- This will be announced at the October GO meeting. Suzi will put a note in the agenda.
GO meeting recap
- There is a google doc with action items compiled by Pascale. See the link above.
- They are assigned to individuals.
- Will discuss with each group. Some discussing points are in the agenda for this call.
- David: The March release was delayed by MGI.
- Seth: Probably needs more resources to fix issues. There are key pieces of the software that not enough people know; we need to train more people to be able to help.
- Chris asked whether it is possible to add Dustin to help on the release pipeline. Huaiyu explained that Dustin is currently working on three parallel projects, and it would be helpful to know exactly the tasks (either long-term or short-term). Will continue the discussion offline.
- Landing page is slow loading - Seth: this is a know issue.
- Seth: If it is slow using the editing tool, it is an issue. Should report if this happens.
- The slowness is in comparison with the production. Kimberly experienced the problem of taking a minute or so to add an evidence. Will report if this happens again, and Seth agrees to look into the problem.
GO-CAM vs annotation
- Kimberly: At the recent GO-CAM call, curators expressed disatisfaction pretty strongly about using Noctua. The bottom line is that they feel they are forced to create GO-CAM models while all they need is to make simple annotations.
- Huaiyu: We should do a better job communicating with the curators, and separate creating simple annotations vs. creating models. For curators who just want to use Noctua instead of Protein2GO, they can create simple annotations. For power users, they can use the tool to create models by adding causal relationships.
- Kimberly: Maybe to first to introduce how to create simple annotations, and then the causal models.
- Pascale: The repository now contains a set of annotations without causal relations (e.g., in XenBase). These are not models.
- Chris: Maybe to create separate repo: GO-CAM vs annotation sets?
- Kimberly: The annotation review tool being developed by Tremayne should be the priority and it will address the curators' concerns.
- Action points:
- Create documentation: different between GO-CAM models and annotations (or annotation sets).
- This will inform how to change the UI to make the distinction clearer.
- David and Judy: The goal is for Seth to have a face-to-face meeting with Lori and Harold, with Dustin as a backup, to discuss the GO-CAM round tripping in MGI.
- Chris suggested to start this with zoom calls ahead of the time
- Action points: David will set up the meeting - let's see how much of this can be done by phone conference.
- Time will be early July?
- Pascale: At the PI call, it was suggested to have it every 2 months, with GO meetings.
- The discussion was about whether to have one in June or combine this together with the BHB visit.
- Seth: No one is using the helpdesk anymore (please refer to his presentation at the GOC meeting).
- Action point: Seth to talk to Todd Harris