OE IRC 16Nov06

From GO Wiki
Jump to: navigation, search

Main chat:

5:25 PM
hjd has joined the channel
hjd: Hey
kchris has joined the channel
midori: hi there
jrichter: Hi.
kchris: hi 
jrichter: I'm uploading beta11 right now. It should be available by the end of this chat.
jrichter: I haven't gotten much done, since several of Midori's bugs have been very difficult to diagnose.
jrichter: Congratulations.
midori: yay! I win!
Alex_MGI has joined the channel
kchris: John, have you worked on Mike C's obo2obo problem?
jclark has joined the channel
jrichter: Yes I have. I thought I sent out a test script for him to try. Did I forget to do that?
5:30 PM
kchris: I don't know, did you send it to the list or directly to Mike?
jrichter: If I sent it, I sent it directly to Mike. I also asked him for an account on his Solaris machine, since that's the only sure way for me to debug this problem.
jrichter: I haven't heard back on the Solaris account.
jrichter: The problem seems to be specific to his unix installation, and I can't reproduce it on my equipment.
kchris: However, being the production environment, we really do need it to work there...
kchris: Thanks for the update on the obo2obo issue.
jrichter: I know. That's why I'd like an account on that machine so I can make sure it works properly. Can you check in with Mike about setting up an account for me?
kchris: He was out of town yesterday, but I will ask when he's back.
5:35 PM
jrichter: Cool. The next version of OBO-Edit may address his problem anyway, but I'm not optimistic.
midori: Tanya emailed to say she can't make the chat today, but she sent a testing question:
midori: Are we expected to retest everything everytime there is a new beta release?
jrichter: Hmm...
midori: (She sent John a report on testing done in beta 8.)
jrichter: Here's what I think...
j-lo has joined the channel
jrichter: There's no need to retest everything on each beta release. But beta testers should look at the release notes for each beta release. If I've tampered in an area in which you're testing, at least do some cursory checks to make sure things still work correctly.
midori: Thanks. Can you add that to the testing wiki too?
jrichter: Yeah.
jrichter: Have we heard anything else on hiring a technical writer?
midori: Not since the call last wednesday.
kchris: Following on fromTanya's question, should everything get tested on the final version before we agree to release it?
j-lo: I haven't heard back from go-top yet
j-lo: wouldn't hurt to remind them
5:40 PM
jrichter: Karen - I think that there should be one last pass of retesting once we think we've got a releasable version.
jrichter: But that seems quite a way off.
jrichter: I think we need to call a strict feature freeze if we ever want to get this out the door.
kchris: Good, I agree with having a last pass of retesting before release, just wanted to confirm we're all on the same page.
kchris: Also completely agree on the strict feature freeze, so that we have a defined set of things to get working.
midori: A feature freeze is fine with me.
j-lo: yes - no more new features until this release is official
Alex_MGI: John, do you want to propose the features to be included then?
jrichter: No more features besides what's currently implemented. Once all the documentation is done and all the bugs on the bug page are fixed, we're finished (in theory).
jrichter: (In practice, sometimes I need to create a new "feature" in order to get an existing feature working)
midori: I can put those in my mental bug fix box 
jrichter: Does anyone want to plead a case for something not in the current bug fix list?
midori: Does "the current bug fix list" mean all of the open bugs in the tracker?
jrichter: Yes.
j-lo: We were experiencing some odd behaviour with the reasoner yesterday
jrichter: ?
j-lo: With the implied links that Chris added for us
5:45 PM
jclark: I put screenshots on the wiki
jrichter: Is this the thing that I couldn't reproduce, and I was asking for a copy of the file?
j-lo: I think it might be best if we invite you to one of our vnc discussions to show you
j-lo: It's hard to describe the problem when we don;t quite understand how it works
jrichter: When's the next VNC?
j-lo: erm...Midori?
j-lo: or Jen? I'm not sure - it's thanksgiving next week isn't it?
jclark: I don't know when the next one is
jclark: I was on holiday for the last one
j-lo: we'll let you know John
midori: It's not scheduled yet, but we know it won't be next week because David can't make it, and I think Tanya can't either. But we (the GO Eds involved) can have a quick session just to show John what's going on.
midori: (Background for others: this has to do with using OBOL, cross-products, and the reasoner to make it easier to put in accurate and complete relationships for regulation terms. We're testing it in the context of the is_a-complete process ontology work.)
jrichter: I'd like to see this as soon as possible, so if anyone could spare some time today or tomorrow, I'd like to do a private VNC thing.
j-lo: could do after this meeting if you have vnc John
midori: Do we have a way to set up VNC (e.g. do you have access to a server)? If so, I could stick around today.
5:50 PM
jrichter: (Jane - well done on knowing that Thanksgiving was next week. I have no clue about the date of British Empire Day [or whatever it is you do])
midori: aw, David clued 'er in yesterday
j-lo: Yes - I take no credit 
jrichter: Midori - It's probably best if you ran the VNC server, since you've got the files we need. If you have a public IP address, I can talk you through setting up your own server after the chat.
jclark: isn't it a bit late in the evening for that?
j-lo: I don't mind - I'll hang around too
jclark: D'y mind if I head off? I have stuff planned for tonight.
midori: OK, happy to try ...
j-lo: (that's fine Jen)
midori: Jen - fine with me 
jclark: ta
jrichter: Okay, let's figure this out after the VNC chat.
jclark: ta again
jrichter: Any other issues?
j-lo: obomerge
jrichter steels himself
j-lo: still isn't working
midori takes a step back
jrichter: Perhaps we can look at this during the VNC session as well.
j-lo: okay
jrichter: OBO-Edit beta11 just finished uploading.
5:55 PM
jrichter: If there are no other issues besides the ones from Jane & Midori, perhaps we should break early and get on with the bug show and tell.
Alex_MGI: Fine with me.
midori: I assume you got the reviews from the paper ...
jrichter: I did not!
j-lo: what were they like?
midori: Wha' ?!?
jrichter: I haven't heard a peep about the paper since I submitted it. Perhaps you're the only one receiving correspondence.
midori: I'll forward to John ... should I send to the list too?
jrichter: As long as they aren't full of curse words.
midori: Not as such ... 
j-lo: yes please
jrichter: What
jrichter: 's the short version of the reviews?
midori: aha -- it says it went to your bbop address. on the way to the list now
midori: "minor revisions" ... but I'm not sure how they're defining "minor"
midori: anyway, they're not too long, so watch this space ...
6:00 PM
midori: Reviewer: 1
midori: Comments to the Author
midori: "There is absolutely no doubt that OBO-Edit has a long standing with biologists in the ontology- and thesaurus business. In contrast, your application note submission leaves a rather premature impression. It focuses on implementation details but ignores interesting aspects in ontology creation and editing. Sure enough this is not a theoretic article, but the OBO-specific limitations to ontology modeling are necessarily those of the editor tool and should be mentioned. The additional limitations of OBO-Edit like the non-use of class instances are also missing."
midori: "Since you mention the planned extension to OWL im-/export, it would be highly desirable to read something about the mapability between OBO and OWL and the required OWL level or pitfalls."
midori: Reviewer: 2
midori: Comments to the Author
midori: General comments
midori: "This paper presents an editor for the OBO language, which has been developed by the Gene Ontology Consortium, and which is a relevant ontology language . The paper is clearly written and organized taking into account the limitated length of Application Notes. The editor itself is not very original and contains the typical features ontology editors do."
midori: "Due to the length of the paper is not possible to compare this ontology editor with others such as Protege, which is much more general than OBO-Edit. OBO-Edit is basically an adaptation of DAG-Edit."
midori: "Maybe due to my computer science- ontology background, I wonder why the OBO editor has not been developed as a plug-in for OBO. In fact, there is an import Tab for OBO files for Protege (http://faculty.washington.edu/gennari/Protege-plugins/OBO-import/) developed by the University of Washington in 2005. "
midori: "One of my concerns is related to the real need for the OBO language as an independent language instead of being built on top of OWL, which is a recommendation of the W3C for representing, sharing and exchanging semantic contents."
midori: that's the lot.
jrichter: Huh.
jrichter: What are we supposed to do with this information?
j-lo: mention the OWL OBO mapping?
midori: um ... The usual drill is that we're supposed to revise the manuscript, addressing the reviewers' comments one way or another. But I'm not sure what exactly we do to address some of these!
jrichter: I feel like we have no choice but to ignore the comments of the second reviewer. That reviewer wants to see a full bodied justification of why the OBO format exists, and this is not the right venue for that.
jrichter: We need another paper to address the second reviewer.
6:05 PM
jrichter: The first reviewer really wants to see a paper about the OWL-to-OBO mapping. That paper won't be published until well after this one. I know it won't, because I haven't had time to write my section.
midori: I agree ... we can explain that to the editors (and reviewers, who will see whatever we send). It's normal to include a set of responses to review comments in an accompanying document.
jrichter: These responses do raise an interesting question for the working group:
jrichter: Every time I discuss OBO-Edit or OBO, I'm asked some version of the question "why are you bothering with this"? Why not use OWL and Protege?
jrichter: What's our answer to these questions?
jclark: I hear protege is really hard to work
jrichter: I mean, what are individual working group members answers to these questions.
jclark: I don't know much about owl
kchris: I thought it was pretty much because our ontology isn't yet  ontologically correct yet, so we can really use either of those tools at present
jrichter: Once OBO-Edit is is_a complete, we'll have no problem using Protege.
kchris: and that our primary concern is that we produce an ontology that is usable for annotation, though we are now working with an ontological formalist to address those issues
jrichter: In fact, I think we could use Protege now, it's just that some reasoners would have a big problem with the lack of is_a completeness.
midori: I've done very, very little with Protege, but that little has left me with the impression that the learning curve is steep.
Alex_MGI: My impression of Protege is that it is rather complex with many irrelevent features that give it a steep learning curve.  But, I only looked at it once.
6:10 PM
jrichter: I hate to say this...
midori: Our reasons for using OBO-Edit are partly historical: when we started working on GO, Protege was not yet open source, and Suzi was very reluctant to commit us to using something we couldn't modify to meet our needs.
jrichter: But I think I'm going to propose that we form a working group to test Protege vs OBO-Edit and see which suits our needs better.
midori: ouch.
jrichter: (At the next consortium meeting)
jclark: would you be able to develop protege if it was better?
jrichter: Protege is really pluggable, so yes.
jclark: that's not so bad then
jrichter: We have nothing to lose except time.
jclark: that's good to know
jclark: best ask teh question before someone asks it forceably
kchris: and we have so much of that (time) to lose 
jrichter: Either we'll find a better program, we'll find some ways to improve OBO-Edit, or we'll finally have an answer to why we don't use the damned thing.
jclark: quite
Alex_MGI: We'll need to be thinking about relevant criteria for comparison, though.
jrichter: No doubt. I think that'll be the first task of this working group.
jclark: we'd have to learn to use it which sounds quite a big thing
jrichter: Perhaps this is better proposed at a managers call than at a full consortium meeting...
j-lo: I don't think Protege is intutitive for biologists
jclark: that's what I've heard
Alex_MGI: Is there anyone in the GOC, though perhaps not in this group, with experience in both programs?
6:15 PM
j-lo: It's true - Robert Steven's gave us a tutorial in it once
jrichter: What is it about OBO-Edit that makes it *more* intuitive?
jclark: after the cns meeting a protege user tol me she was going to switch to bo-edit
jclark: not sure I've never seen protege
j-lo: Protege sort of forces you to think backwards - it probably suits programmers better
midori: I'm downloading it now ...
jclark: backwards?
kchris: what little I saw of Protege (a grad student in Russ Altmann's lab demoed putting GO into Protege several years ago) involved lots of "frames" and other jargony words that meant nothing to me
midori: bottom-up is another way I've heard it put
midori: I also seem to recall a bunch of cryptic icons.
Alex_MGI: I think it's telling that Protege training workshops take several days.
j-lo: (unlike OBO-Edit 
jrichter: I feel like the main difference is that Protege treats relationships as aspects of a term. It's as if you ran OBO-Edit without being able to edit in the ontology editor panel. You have to edit using an equivalent of the parent plugin.
midori: hey, no fair - only half of our icons are cryptic! 
kchris: To be fair, I think OBO-Edit has a steep learning curve too, not having used Protege, I can't compare
jclark: that does sound less intuitive
jrichter: (If anyone can think of a better icon for "root selection algorithm", for God's sake send it to me.)
jrichter: Anyway, I'd like one of our GO managers to bring this up at the next call.
6:20 PM
jrichter: I think it's a really important question, and it affects our funding.
j-lo: Okay, we will - the next call isn't until the end of the month though
midori: 29th
Alex_MGI: Isn't the new GO grant already covering continued development of OBO-Edit?
Alex_MGI: (not that we shouldn't do the comparison)
jrichter: Yes. But our advisors and grant reviewers are always asking pointed questions about why we insist on re-inventing the wheel. That makes me uncomfortable.
midori: The proposal requests O-E funding, but as far as I know we haven't heard what we'll actually get.
jrichter: Holy crap! I thought of a question I was supposed to ask 3 meetings ago...
jrichter: How do people feel about a set-based search in OBO-Edit 2.0? Instead of building up complex filters in a tree view, complex filters would be built by widening and narrowing existing search results.
midori: That sounds excellent.
kchris: that would be great, I've never really gotten the hang of building complex filters
jclark: I like the one we have but I don't really know what you mean
jrichter: (The good thing is that the set-based approach is actually equivalent to the current approach, so it's just a matter of adding a new skin on top of the old guts)
6:25 PM
jclark: so we could choose either?
jrichter: Yup.
jclark: funky!
j-lo: good idea - will make it much more intuitive i imagine
jrichter: Cool. I'll keep that in mind for OBO-Edit 2.0.
jrichter: Can we move on to the VNC stuff?
midori will need a 2-min break ...
jrichter will too
Alex_MGI: Can the rest of us go then?
jrichter: Yup. Bye everybody!
jrichter: (except Jane & Midori)
kchris: bye!
kchris has quit the server saying: Quit: I'll be back!
midori: I'll do transcripts as usual.
hjd has quit the server saying: Quit: Chatzilla 0.9.76 [Firefox 2.0/2006101022]

Follow-up:

midori: crap - I got disconnected from Conversation
jrichter: You can't remove an implied link. You have to remove the links that support it.
midori: anyway, screenshots linked to http://gocwiki.geneontology.org/index.php/Conference_calls#EBI.2FMGI.2FTAIR_Conference_call_11-15-06
jrichter: Jane just showed me.
midori: ... and if I'm serving up VNC, it'll be port 5900
midori: OK ... I must have missed some comments when I got disconnected; do you have a transcript?
j-lo: we were just talking about obomerge
j-lo: i can save the transcript if you like
midori: ta!
7:10 PM
jrichter: What you've described on the Wiki looks buggy. It shouldn't be showing different results when you save and reload (unless you saved with "realize implied links" enabled)
j-lo: we didn't
jrichter: The Wiki doesn't say what terms and relationships you were actually working with.
midori: We also were confused by what relationships did and didn't show up in the graph viewer.
j-lo: it was go/scratch/IsaComplete-xp.obo
midori: I suspect we didn't fully understand what we were supposed to see, but it didn't seem right.
j-lo: did you see those cycles it made?
jrichter: I actually have no idea how the graph viewer displays implied relationships. The graph viewer was created long before anyone started using any cross product features.
jrichter: Let me download the appropriate file and load it up.
midori: For example, in screenshot1, why does regulation of response to stimulus appear to have no parents?
midori: It does in the file!
jrichter: Midori - is this the latest version of IsaComplete-xp.obo?
7:15 PM
midori: It was as of yesterday, I think.
jrichter: I'm firing it up right now.
jrichter: Is this file the result of an OBOL run?
midori: Chris said something about a pared-down version of OBOL, iirc.
jrichter: I'm not seeing anything particularly strange. Can you point me to a problem area of the graph.
jrichter: ?
midori: (in other news, we're pretty sure VNC's not getting out ... you may now cuss out ntl)
j-lo: John - run the verification check
jrichter: In progress...
j-lo: ...just try saving or something
midori: We picked some blue squiggles more or less at random; e.g. looked at regulation of vernalization response GO010219.
7:20 PM
jrichter: Is it true that "regulation of synapse structure and function" is a type of "regulation of synaptic transmission"
jrichter: ?
j-lo: yes
j-lo: it's a bit odd - we're going to change that
jrichter: Then at least one of these cycles is built into the logic of the ontology.
jrichter: Do you have the ontology loaded up so we can work through this?
j-lo: how so?
j-lo: I'll just do it now...
midori: No, it's part_of, not is_a
jrichter: "regulation of synaptic transmission" is part_of "regulation of synapse structure and function"
midori: p.s. I"ve got it loaded
jrichter: "regulation of synapse structure and function" is_a "regulation of synaptic transmission" because...
jrichter: regulation of synapse structure and function --part_of--> synaptic transmission
jrichter: regulation of synapse structure and function --OBO_REL:is_a--> regulation of biological process
7:25 PM
jrichter: Anything that is part_of "synaptic transmission" and is_a "regulation of biological process" must have an is_a relationship to regulation of synaptic transmission
jrichter: The easiest way to puzzle these things out is with the explanation plugin.
j-lo: what does the red arrow in the DAG viewer mean?
midori: you know, we didn't even think to look at the explanation plugin...
jrichter: If you'd like to go through an example, select "regulation of synapse structure and function -is_a-> regulation of synaptic transmission" and open the explanation plugin.
jrichter: (red arrows mean "redundant")
jrichter: Does everyone have the explanation plugin open?
j-lo: yep
midori: ok, od now (took a while to get to the relationship in the ont ed panel)
7:30 PM
midori: er, do now ...
jrichter: When you double-click the link at the top of the explanation panel, you'll see a line that looks like:
jrichter: COMPLETENESS: matches=[regulation of synapse structure and function --part_of--> synaptic transmission matches regulation of synaptic transmission ~~part_of~~> synaptic transmission, regulation of synapse structure and function --OBO_REL:is_a--> regulation of biological process matches regulation of synaptic transmission ~~OBO_REL:is_a~~> regulation of biological process]
jrichter: (Select that line to see the full text)
midori: erm, I'm not feeling enlightened ...
jrichter: Every time you expand a link in the explanation plugin, it shows you the reasons why that link exists.
jrichter: In this case, the link exists for one reason: the definition of COMPLETENESS (aka intersections, aka cross products)
midori: yes, but what does "matches" mean?
jrichter: An intersection is used to create a new link when a term is found that "matches" the intersection definition.
jrichter: So if we say that anything that is part_of synaptic transmission and is_a regulation of biological process is a regulation of synapse structure and function...
midori: ah ... I get it now, but it's awfully cryptic in the plugin
midori looks whether the user guide makes all clear
jrichter: anything that "matches" those definitions gets an is_a relationship to regulation of synapse structure and function
jrichter: The user's guide does not.
j-lo: okay - I see
jrichter: These messages need to be improved, but let me get through the discussion of how it works now.
jrichter: If you expand the line that begins with "COMPLETENESS", you'll get a list of links...
midori: indeed. (re user's guide)
jrichter: These are the supporting evidence.
jrichter: These links can themselves be expanded.
7:35 PM
jrichter: That way, you can see why THOSE links exist. There are often many steps to deciphering this info.
jrichter: Although this turns out to be an easy case.
jrichter: If you expand "regulation of synapse structure and function --part_of--> synaptic transmission", you'll see several explanations.
jrichter: That's okay - a link can have many reasons for existing.
jrichter: If a link has the reason "GIVEN", it means that it was created by a human curator.
jrichter: BTW: OBO-Edit decides that a link is redundant if it has the explanation "GIVEN" and some other explanation, because it means that the link doesn't need to be explicitly defined in the ontology.
jrichter: Whew!
jrichter: Any questions?
j-lo: when you select a term (e.g. regulation of angiogenesis) it appears twice in the DAG viewer - once as redundant and once not
j-lo: why is that?
jrichter: Sorry! I didn't realize what red arrow you were talking about!
jrichter: A red *triangle* doesn't mean redundant.
jrichter: A red *triangle* means that this is an intersection link - that is, a link that will appear in the cross products tab.
j-lo: Ah - okay
midori: Is it redundant to have both an intersection link and an ordinary link saying the same thing?
j-lo: I think that explains the problems I was having!
jrichter: OBO-Edit thinks so, but only because OBO-Edit thinks that anything that can be inferred from the reasoner is redundant when explicitly stated.
jrichter: I need to step out, I've got a call on the baby monitor.
7:40 PM
j-lo: ok
jrichter: I may be gone for 10 minutes or so. If you need to leave, let me know.
midori: It's getting late here anyway -- shall we wlap up?
midori: .. wrap up ... my typin bis going
j-lo: hah hah!
j-lo: Lets finish then
midori: ok 
midori: might as well send the transcript out .. cc david too
j-lo: i can't seem to copy it on this client...
midori: crud. I'll save what I got ... there's a bit missing, tho, from when I got booted.
j-lo: there wasn't much u missed
midori: ok ...
j-lo: see you tomorrow
midori: ok (unless I work from home ... might do, I'm sniffling more than before)
7:45 PM
j-lo: ok - bye!
j-lo has left the channel
jrichter: Still there, Midori?
midori: yup...
jrichter: I was hoping for some opinions on a better message from the "COMPLETES" rule...
midori: oughta go soon
jrichter: (I ought to use a different icon for rules and evidence... that would help too)
midori: one thing I can tell you quickly: the user guide has to explain the distinction between -- and ~~ around relations.
midori: for completeness ... maybe label genus and differentia terms as such?
jrichter: good idea
jrichter: Did we address all the reasoner-related problems, then?
midori: The main ones I recall ... it woudn't hurt for you to join the next webconf anyway, in case David or Jen can think of something else.
7:50 PM
jrichter: yeah, i'll be there
jrichter: see you in 2 weeks, then.
midori: ok. see you. happy thanksgiving to you, heather, and baby-lo
jrichter: happy irish beating day to you and yours!
midori: thanks, i think. bye!