OBO-Edit Working Group, Thursday September 3rd, 2009 8.30 a.m. PST
Conference call details:
Outside US: 1-719-867-3417
Attendees: Amina, David, Karen, Midori, Jennifer, Jane, (Chris after 5.15pm).
Action items carried over
- Amina needs to change default selection model and asks for clarification.
- Should the OTE be in local or global mode during scroll lock?
- Should locked global OTE respond to selection of search result?
- Midori: it's okay if it moves when I actually change the selection. The problem is if I add a parent to the term and the graph moves a lot even when scroll locked.
Midori demos bug.
- A problem appears where Midori is unable to lock OTE using keyboard shortcut. She has a mac and a one button mouse.
- Amina tried to use right click with her mouse on Midori's computer, but the lock view feature does not work for her either.
- Sometimes Midori sees the background go grey indicating that lock view works, but then the view jumps around anyway.
We discuss what will happen at the meeting. Amina will present by remote. She would appreciate if someone with a webcam would let her see the audience so she has some idea of what kind of reaction she is getting. We wonder if she would like one designated person to give her feedback.
Talking about display of has_part and regulates in AmiGO. Should we have them displayed the same way in OBO-Edit. Should we try to have a breakout session during the meeting? Will there be time? The AmiGO people have no concrete proposal as yet. Will this discussion take too long? Jane is eager not to have the two groups waste time by discussing separately. Karen will send some examples of how OBO-Edit displays.
David: Some relationships are used for grouping annotations, but has_part is not used for grouping. This is a different kind of relationship. Karen: There is a problem with direction of arrows.
Should we discuss this at the meeting? Perhaps if we had options to show at the meeting?
David: Can we put somewhere in the relation information about why it is displayed in the opposite direction? Can we mark granularity somehow? Karen would like the smaller term to be below the bigger term. Maybe it is hard to label granularity.
Karen: Regulates doesn't really imply granularity, but with has_part it is intrinsic.
David likes the idea of the relationship indicating which direction it is shown so the programmers don't have to hard code information about direction for each relationship.
Could it just go in the Typedef? has_part is asymmetric so can we use the asymmetry on the instance level in some way to get a logical indicator into the Typedef? This might be better than an adhoc marker in the Typedef.
Chris mailed the web presence list about this. He has a good image showing how this could work and Karen liked it. The diagram had the spliceosome with the has_part arrows inverted. It's not clear whether there is formal encoding in the ontology about direction.
Action: David will mail Chris about this.
David will try to come over to the Consortium meeting for a couple of hours.
Amina explains what has gone before about has_part.
Chris points out that OBO-Edit is for more advance users, so it could be different if appropriate.
Karen thinks it would be nice to have OBO-Edit configurable to make sense to biologists for making figures and showing to audiences.
David explains his idea about Typedef to Chris...
Chris says yes we could add metadata in the relationship declaration, though it is still not clear how we will display the relationships.
e.g. the type level inverse of has_part is only_part_of.
David: what if we just switched the direction of the arrow?
Chris: We'd have to change the label as well as the arrow. This would also lead to more mistakes because of the direction of implication.
David: Yes this is worrying. We may get things back to front until we are used to the new arrangement. Would it be too much to ask for it to be customisable?
Chris: Perhaps? It depends how much work it would be and how high a priority. Would need new metadata on relationships, and also config on UI. Maybe an experiment. I am very leery of trying to hide the truth of the logic. Perhaps if we did the experiment then we could let people try both ways and see what they prefer?
Karen: Customisable may be very helpful. Advanced users can use the straight logic, while biologists will use the more intuitive version.
Chris: When talking to fellow curators instead of making a diagram it is different. For curators it is important that they should understand the relationship or they will at least make errors of ommission in annotation.
David: It depends how you look at it. If you are used to looking at things as sentences then you would never put things back to front. e.g. myelin sheath has_part myelin then that can never be confusing.
Karen: But the graph view is confusing. Suddenly smaller parts are above. In English you can invert sentence order, so that is not such a big thing.
David: Hiding the quantifier is the problem. If you are used to reading the sentence it is easy but this is hidden in OBO-Edit. We really need to understand the logic.
David likes the idea of browsing sentences.
Karen thinks it's really hard to find terms with the new relationship.
David is concerned about assumptions about grouping. People work down the graph and choose a level that works for grouping. With has_part there will not be grouping and it could be confusing.
Karen thinks that fixing spliceosome so it groups properly is a different argument than making the display correct.
David wants us to note that grouping is very important, and especially assumptions about grouping.
Chris says why don't we do the experiment. Make the demo and see whether we find it helpful to have it one way or the other. Test to see how people annotate using the two arrangements.
Karen says if we can't get grouping with has_part them maybe we need to say part_of_some to make it clear what is going on, and so we can group.
David says there is no general logic solution for Karen's problem. Perhaps individual solutions in individual situations, but there is no general logic solution.
Maybe we could define spliceosomal complexes as including all the parts of the spliceosome even when the are free floating and not part of a spliceosome.
Discussion breaks up due to lack of time.