OE IRC 21Sept06

From GO Wiki
Jump to navigation Jump to search

gwg has joined the channel
5:00 PM
hjd has joined the channel
MelissaH has joined the channel
MelissaH: Hi there.
jclark has joined the channel
jrichter has joined the channel
midori: Hi
jrichter: Hello, all.
midori: It's good to be back.
Alex_MGI has joined the channel
5:05 PM
jclark: what's on the agenda for today?
midori: OK, there's the 55 timestamp, so let's roll!
midori: John sent that email about 1.1 and 1.2 ...
jrichter: Show us how it's done, Midori! Moderate your heart out!
midori: Well, a lot of it will be verbatim quotes from your email 
jrichter: Yeah. I just wanted to get everyone thinking about what features they'd like in OBO-Edit 1.2.
midori: Topics from our fearless programmer: "OBO-Edit 1.2 features and OBO-Edit 1.1 completeness"
jrichter: Here's what I had to say about OBO-Edit 1.2 features:
jrichter: I think we need to focus on reasoner improvements in the next version, and on the graphical editor, because...
jrichter: our reasoner doesn't do everything it should, or even everything you'd expect it to do given the things you can say in OBO-Edit (for example, the reasoner doesn't do anything with symmetry right now).
midori: OK so far ...
jrichter: The weakness of our reasoner is one of the most important ways in which OBO-Edit lags behind Protege. If OBO-Edit's reasoner had more features and...
jrichter: was able to incorporate edits quickly (instead of having to do the reasoning all over again), OBO-Edit would be a lot more useful, and stack up quite a bit better against protege.
5:10 PM
jrichter: <pause for breath>
Alex_MGI: What does Protege do so well?
midori: Interesting that you bring up Protege ... apparently one of the criticisms of the GO grant was that some of 'em still don't get the difference between OBO-Edit and Protege ....
midori: even said we should just use Protege.
jrichter: I sent out that email after I talked to Judy about that criticism.
jrichter: Protege has a really first rate reasoner. That's really the main difference.
jrichter: (besides some very different interface choices)
MelissaH: Can you explain in biologist terms what this means for us?
jrichter: Let's say we developed the Gene Ontology in Protege instead...
jrichter: The advantages: The reasoner would point out thousands of logical problems in the GO, and we never would have been allowed to do some of the things that have ballsed us up, like creating a non-isa-complete ontology. Protege has a lot more focus on ontological correctness.
MelissaH: But then, many of our ontologies wouldn't have gotten developed at all if they had to be is_a complete.
jrichter: Disadvantages: (see advantages) and Protege is streamlined for editing OWL, so all the complex OWL features are right out front for you to tweak and nitpick to your heart's content.
midori: Nail. Head. Hit.
5:15 PM
jrichter: This makes it really hard for biologists to use, because a lot of times you need to throw together a basically correct framework with as few niggling details as possible, and work backwards.
jrichter: You can say almost anything you can say in Protege using OBO-Edit. But OBO-Edit hides more of the frreaky details.
jrichter: (At the cost of a little bit of expressive power)
jclark: Several people have told me it's really hard to use. Now I see why
jrichter: That's why ontologies developed in Protege tend to be small (with the glaring exception of FMA). In a small ontology, it's easier to really dig into the tiny details.
MelissaH: I tried briefly and it seemed to require a 2K$$ training lesson...
midori: I've only dabbled, but it seems the differences between the interfaces are huge. Protege's learning curve daunts me.
jrichter: Well, Protege is always getting easier to use. And it would immediately become clearer if you loaded an ontology you were familiar with.
MelissaH: BTW, just got FMA in obo this morning.
jrichter: Sweet.
Anon739 has joined the channel
jrichter: It's always been my opinion that Protege and OBO-Edit are complementary tools. OBO-Edit is for the big picture, and Protege is for tweaking the details.
MelissaH: For the FMA, protege requires a MySQL database installation.  Would this also be true for GO?
jrichter: Not at first, but...
jrichter: The MySQL database is there to hold the giant amount of data that the reasoner generates. To use OBO-Edit's reasoner on the FMA, I may have to rebuild the reasoner with a MySQL storage option.
jrichter: Oh, you mean GO in protege.
MelissaH: oh dear.  
jclark has quit the server saying: Ping timeout
5:20 PM
jrichter: I don't know if a mysql server would be necessary for GO in protege.
MelissaH: yes, I guess I am wondering if the difference between oboedit and protege is the amount of reasoning, then large ontologies like the FMA require an extra database.
jrichter: Anyway, if our reasoner was better, we'd stack up better against protege, AND more of our groups could start seriously working on reasoner-intensive activities, like cross-product development.
midori: So can we go back to the OBO-Edit reasoner? What are the specific changes you want to make (if you can do a brief, biologist-friendly summary), and how will it make our GO- (or CL-  or fish anatomy-) building lives better?
Anon739 has quit the server saying: Quit: Anon739
jclark has joined the channel
jrichter: Improvements I'd like to make:
jrichter: 1) Correctly implement symmetry
jrichter: 2) Correctly implement domain & range
jrichter: 3) Correctly implement inverse_of
jrichter: 4) Correctly implement necessity & inverse necessity
jrichter: 5) Think about correctly implementing cardinality
jrichter: 6) Make the reasoner accept incremental changes without re-reasoning the whole ontology
jrichter: And maybe
jrichter: 7) Allow OBO-Edit to talk to 3rd-party reasoners
jrichter: That's it.
jrichter: And I can take or leave 7.
Alex_MGI: Perhaps it's too much to explain all these things here, but maybe you could write out a longer explanation to send people of these features.
MelissaH: Can we take them one at a time?  1. what do you mean by symmetry, new symmetric relations?
MelissaH: good idea alex.
jrichter: II'll send out an email of the meanings of each of the early things.
Alex_MGI: And do they all have to happen at the same time.
midori: Will these changes (e.g. #6) affect performance?
jrichter: Absolutely.
5:25 PM
jrichter: Right now, you can't really edit while the reasoner is on. If 6 were correctly implemented, you could edit without even realizing the reasoner was on (or with only a minor dip in speed).
gwg: I think 7. would be quite useful, especially if people have their own fave reasoner
midori: "you can't really edit while the reasoner is on" is something of an understatement ...
jrichter: Trouble with 7 is that its a lot of work, because it's tricky to build the bridge between OBO-Edit and the external software.
gwg: yes, I thought that might be the case...
jclark: are we likely to want to use other reasoners?
jrichter: Jen - If we could get the bridges to work, almost certainly.
jclark: would it be possible to build the bridges instead of the good reasoner and just use 3rd party reasoners?
jrichter: BUT the GO would need to be more formally correct before we got any useful info out of those reasoners. So I think we can afford to wait.
Alex_MGI: So there are other reasoners out there?
jrichter: Alex: there are lots of other reasoners.
jrichter: The OBO-Edit reasoner has a few advantages though:
jrichter: 1) It's fast
jrichter: 2) It can (in theory) accept incremental changes
jrichter: 3) It can explain itself
jrichter: It can be hard to understand those explanations, but they really are useful with some training.
jrichter: Note that option 6 is almost done, but it needs to be seriously debugged. I just cut the incremental changes filter to get the early OBO-Edit betas out the door, and then I never had time to put the feature back in.
jclark: are the externaladvisory board keen that we should do this?
jrichter: (change "filter" to "feature" above)
5:30 PM
jrichter: Jen - which "this" are we talking about?
jclark: the 1-7 plan above?
jclark: the features I mean
jclark: It seems to me that if the PIs and the external advisory board people said this is the way to go then that's useful to know.
tberardi has joined the channel
midori: I don't remember them saying anything that nitty-gritty ...
jrichter: I don't know if they're keen for us to do it now, but we will started getting pointed questions about it if we don't get moving. AND a reasoner is going to be very useful to the growing number ontologies with cross products.
jrichter: The second reason is really the more important. We need it.
midori: ... but my sense is that they'd approve.
jclark: sounds like we should go ahead then.
MelissaH: I was going to suggest more cross-product testing, and requirements gathering.  Where are we at with this?  How many of us use the cross-products editor frequently?
tberardi: I don't.
jclark: not me yet
hjd: Nor I
jclark: I would use a good reasoner more
MelissaH: I don't but really need to.
Alex_MGI: no
gwg: I tried, but 'twas in vain
jclark: John, can you give us an idea of how long all this would take?
jrichter: Melissa: Chris and I are in the early phases of putting together an OBO meeting, where the cross-product users will come together in one place. This will be the ultimate place for bug testing and requirements gathering. We'll keep you posted.
midori: The GO people haven't used it much yet, but we're poised to do the GO-CL cross-products soonish (timescale of a few weeks to a few months? Chris might have a better idea ...).
MelissaH: Seems like these go hand in hand.
MelissaH: ok, would love another meeting.
MelissaH: hand in hand I mean reasoner improvements and cross products features....
5:35 PM
jrichter: Jen - 1-5 would take about a week and a half, I think. Number 6 is one of those things that is truly impossible to say, because I don't know how long it would take to weed out bugs. If it went more than a month I'd probably give up.
jrichter: It would probably take between 3 days and 3 weeks.
jclark: sounds like 1-6 could be a good thing to get on with then.
jclark: 1-5 I mean
jclark: How long would the cross product thing take?
jrichter: What's the cross product thing?
gwg: the cross-product meeting? A day? A couple of days?
jclark: Melissa said that she'd like cross products imrpved and I'm wondering how long it would take to do what she wants
jrichter: The OBO meeting would be between 2 and 4 days.
jclark: and then you would need to do some work on the cross product plugin would you?
jrichter: Melissa wants to do testing of the cross product feature to determine what features might be added, so there's no immediate programming work to do.
MelissaH: I think its difficult to say how the cross-product features should be added or improved until we have some better requirements gathered.  All I know is that none of seem to be using it very well yet.
MelissaH: of us
Alex_MGI: Would there be a way to load two ontologies and identify exsiting cross-products of one or the other that already exist in one of the ontologies?
jrichter: I'm inclined to cut the Cross Product Plugin altogether, frankly. The tab in the text editor works a lot better.
jrichter: Alex - how would that work? By comparing term names?
MelissaH: I think Chris has a script that can check for these types of matches.
Alex_MGI: Yes, I suppose.  I wrote many of the cell-specific immunology terms based on the revised Cell Ontology.
jclark: yes
5:40 PM
jrichter: Term name comparison... that sounds like a job for OBOL, I think.
jclark: or structured synonym comparison more like. 
jclark: that did seem to be chris's plan
midori: Yes, OBOL could identify most of the TONS of implicit cross-product terms we have in GO; that's basically how he started the GO-CL work.
jclark: wasn't obol going to be part of obo-edit eventually?
Alex_MGI: Can OBOL then insert the requires information to indicate a x-product into the OBO file?
jclark: that would be mighty handy
Alex_MGI: (After user review of course)
tberardi: I can see that being useful for x-products to the Plant Structure ontology as well.
jclark: also obol's automatic generation of terms would be a useful new feature
tberardi: aka Plant anatomy terms
jrichter: On OBOL integration into OBO-Edit: It's possible to integrate OBOL into OBO-Edit. However, I think we've got about 3 weeks to a month of work to make that happen, and the time just grows and grows depending on how much OBOL functionality we want to give to OBO-Edit.
Alex_MGI: I think it's worth the effort, personally.
jrichter: The OBOL plugin would require external software, as GraphViz does, because OBOL requires another language interpreter.
jrichter: (prolog)
jrichter: Note that OBOL might not be able to do something as granular as Alex wanted...
midori: Maybe the thing to do is for us ontology editors to figure out what we want to do with the ontologies, and let you & Chris hash how how to make it happen. 
MelissaH: Is OBOL ready for this, anyway?
jrichter: Still answering questions, hold on...
jclark: I think that hving obol generate XP type terms for me would be very handy
jclark: I have to mail chris to get them done right now
jclark: and that takes too long
5:45 PM
jclark: If I could just plug in a list of anatomical parts and get obo-edit to make all the development terms that would save hours. 
jrichter: More to Alex - OBOL often has to really tear apart an ontology to properly suggest cross products, so it might not be able to use your existing terms to come up with suggested cross products. It might have to suggest a LOT of new terms first.
jrichter: And those new terms wouldn't necessarily fit well with your existing ontology. For example, OBOL always wants to create a ton of chemical name terms when it works on GO.
jrichter: Also, OBOL is trained for a specific ontology. The OBOL plugin wouldn't work at all for an anatomy ontology.
midori: Personally, I'd love to have O-E suggest new terms (under controlled circumstances, of course)
Alex_MGI: Mostly, I want something that looks at exisiting terms and analyzes them against a second ontology.
jrichter: To answer Midori's point - I think it would be a really good idea for curators to come up with a wish list of activities to automate, so Chris and I could work on software solutions. At the same time, the software should be developed so that it can help curators imagine new things to try based on what the software offers.
gwg: John, could you also write a plugin that analyzes SourceForge entries, extracts terms, checks if they're OK, and then adds them if so? That would really save us a lot of time in the editorial office...
jrichter: Hey! You're cute when you're being a jerk!
jrichter: I can also make OBO-Edit play solitaire for you...
jrichter: Anyway-
5:50 PM
hjd: klondike or spider?
jrichter: There were some really specific software requests here (like Alex's). Let's get those up on the feature requests tracker.
tberardi: getting back to the topic at hand...
midori: On a more serious note, SF item #1551858 illustrates the kind of term suggestions I'd find useful .... (link: https://sourceforge.net/tracker/index.php?func=detail&aid=1551858&group_id=36855&atid=440764)
Alex_MGI: A regulation term generator would be great.
hjd: I hadn't thought of that, but yes: ability to autogenerate the "regulation" terms needed for a new process
jrichter: I'm inclined to hold off on OBOL integration until the next round. OBOL only works for the GO (and other ontologies for which it is trained), and I'd like to focus on more broadly useful things at first.
Alex_MGI: I actully wrote one a an Excel spreadsheet at one point.
midori: (adding to Alex & Harold) ... and follow back up the tree to suggest terms to fill in gaps 
jrichter: Midori - can you make this into a sourceforge request?
Alex_MGI: Absolutely
midori: <gulp> ok
Alex_MGI: (Harold hates those terms, you know)
jrichter: Is there more reasoner/OBOL discussion, or should we move on?
hjd: not as much as" biosynthesis of..."
Alex_MGI: I'm satisfied with your arguments against OBOL for the moment.
MelissaH: I'd like to say a word of special thanks to John for making it to the anatomy and GO users meeting.
midori: seconded!
5:55 PM
jrichter: (Note: I really, REALLY want to do OBOL integration; but after some agonizing I think that it needs to wait one more round)
MelissaH: AND for implementing multiple ontology viewing features ASAP!
jrichter: Yeah, I'm real quick when I'm being physically threatened.
kchris has joined the channel
MelissaH: Maybe John, you could explain to everyone what these features are?
jrichter: Real quick: OBO-Edit now makes much better guesses about which path to a term to select after you've searched...
jrichter: And OBO-Edit no longer shows non-transitive relationships in the DAG Viewer by default.
jrichter: You will find these features indespensible once "regulates" is added to the GO.
jrichter: We're almost out of time, but I wanted to briefly discuss why I want to put graphical editing in the next version:
jrichter: 1) Lots of people are using the Graph Viewer all the time. Might as well make it possible to edit with it.
jrichter: 2) It is visually impressive, which helps draw users, grant funding, etc.
jrichter: 3) A feature like that emphasizes OBO-Edit as a graph oriented ontology tool, further setting it apart from Protege.
6:00 PM
midori: sounds good
jrichter: 4) The infrastructure improvements that the Graph Viewer would require would make OBO-Edit's code more robust. That has no effect on how OBO-Edit looks to the user, but it will make future changes easier to implement.
jrichter: (okay, I'm done)
jrichter: Counterarguments? Is there a better use of time for me?
midori: still sounds good.
MelissaH: sounds good to me too.
Alex_MGI: So John, can you write up a list of proposed new features for 1.2 as you see it, and send it out?
jrichter: There are only 2:
jrichter: 1) Reasoner improvements
jrichter: 2) Graphical editing
Alex_MGI: And is 1.100 beta4 on the way?
hjd: yes, 
jrichter: I don't want to break "reasoner improvements" out into sub-tasks, because they're actually highly integrated.
midori: It would be good to have another look at the feature requests tracker, to make sure there's nothing either (a)  easy enough to do alongside John's two or (b) equally important. (Tho I can't think of any of the latter ...)
Alex_MGI: Yes, but explain them in more detail.
jrichter: Alex: beta4 is on the way. I've found a bug in obomerge that I'm trying to hunt down before the end of the day...
MelissaH: yes, we'd like to have a better idea of what we are getting, so we can make test cases for what we want it to do.
MelissaH: the reasoner, that is.
jrichter: As far as I know, I've completed all the feature requests for OBO-Edit 1.1 (except for a piece of the "interface customization improvements" request that would take a long time for comparatively small payoff).
jrichter: Okay, let's do action items:
jrichter: 1) John releases beta4
jrichter: 2) John writes an email explaining what the proposed reasoner improvements will do (and maybe explaining what the reasoner is doing now)
jrichter: 3) Curators post suggestions on the request tracker for automated tasks that OBO-Edit might be able to do
jrichter: Anything else?
6:05 PM
jclark: john can I talk to you about an unrelated thing at the end
jrichter: You bet.
Alex_MGI: When is the 1.1 release date?
Alex_MGI: And do we all have to sign off again?
jrichter: You want to say October 6? That's 2 weeks + 1 day. Let's us have an IRC chat before release.
tberardi: What about updating the user's guide?
jrichter: "Lets" not "Let's". I suck.
jrichter: We ALL need to sign off again, I think.
midori: No, "Let's" was right; it was the "us" ...
jrichter: I'll be updating the users guide starting with todays release, and updates will come out in the bug fix releases. Is that okay?
jrichter: Midori - shut up.
midori: Should we parcel out testing again? I've noly done a little so far ...
midori: (sorry for not shutting up  )
tberardi: Yes, that sounds great, John.
jrichter: In this case, don't we need everyone's input? I thought we only parceled out testing for a bug fix path to an existing release.
midori: noly --> only 
jrichter: (patch, not path)
hjd: Unfortunately, I haven't been able to get past the "load GO " stage to test anything (beyond the beta 1_
6:10 PM
jrichter: I know. But I think I fixed it. Try starting a terminal, using "cd" to change to the oboedit directory, and launching oboedit with the "oboedit" command. That should work.
midori: Harold: try copying the startup script from beta1 ... worked for me.
tberardi: These are mac specific issues, yes?  I haven't had trouble on windows.
hjd: evidently
jrichter: These are mac/platform independent issues.
hjd: ???
jrichter: The startup script is a little broken. But I think I've fixed it.
jrichter: We're 10 minutes over. Any other issues?
MelissaH: bye bye, have another meeting.....thanks everyone.
tberardi: Did you guys start at 9 am?
MelissaH has quit the server saying: Quit: MelissaH
midori: I'll send the transcript as usual. Just one last thing: next chat should be in 2 weeks, on OCt. 5th. Shall we do 90 or 9:30 AM Pacific time? (Either is fine for me.)
jclark: john can you skype me to talk about the web thing? it wont take long
tberardi: 9:30 please, I just can't make it by 90
kchris: 9 am is really too early for me, I can just make 9:30
midori: 9 AM or 9:30 (it keeps converting colon-zero to a smiley on me)
tberardi: 9:30
kchris: 9:30
midori: OK; if I space and send a mail that says 9: 00 again, reply and change it!
kchris: OK 
tberardi: OK.  Will catch up via the transcript.
kchris: me too
jrichter: Thanks all. See you in 2 weeks.
tberardi: bye