2005-08-03 Conference Call Minutes

Draft Minutes (so far just Mike Cary's notes)

Participants: Gary Bader, Michael Cary, Dan Corwin, Ken Fukuda, Christian Lemer, Joanne Luciano, Ranjani Ramakrishnan, Jonathan Rees, Alan Ruttenberg, Frank Shacherer, Andrea Splendiani, Imre Vastrik, Vincent Yau, Jeremy Zucker, any others?

The main purpose of this call was to decide a course of action for the release of Level 2. We began by discussing the process for a) deciding whether states will be in the release and b) making the necessary updates and putting out the release candidate.

The decision on whether to include states in Level 2 or not was fairly obvious given recent email list discussions and informal polling by Gary: we haven't been able to agree on a state proposal, so, in accordance with our decision in Detroit, we must revert to the 0.9 version of Level 2 for the release candidate. There were no objections to this course of action.

We next discussed how to move forward from here - v0.9 can't be released as it is because a few bugs were discovered in it since the Detroit meeting.

Fixes include making sure all cardinality constraints make sense and changing the types (unification, publication, etc.) of a number of xrefs (see above page, e.g. making confidence xref a publication xref). Another fix includes removing the subclasses of PEP (but keeping one called sequence participant) - this was agreed to at the Detroit meeting.

The documentation currently covers version 0.9, so we don't need to update it for the release candidate. There will be a few small changes before the documentation is final, however.

Alan pointed out that there were a couple open issues still with version 0.9 - e.g. cardinality of the CONTROLLER property. Mike gave a hypothetical situation where one might want to leave the CONTROLLER property empty. At any rate, we need to tie up these loose ends in the near future so we can release the release candidate.

Some of the fixes we need to make in Level 2 are also present in Level 1. Fixing them there would make the resulting new version of Level 1 (1.6) backward incompatible with the previous version (1.4), but not fixing them in Level 1 would make Level 2 backward incompatible with Level 1.

Christian Lemer suggested that we need to better define the relationship between Level 1 & 2. Must Level 1 be a subset of Level 2? Must Level 2 be backward compatible? Should a minor revision in Level 1 be propagated to Level 2? Note that if we imported Level 1 in Level 2, backward compatibility would be ensured and minor revisions to Level 1 would propagate forward.

There are many ways to define and enforce backward compatibility. We agreed that we should create a wiki page on the issue.

Jeremy looked over the proposed changes to Level 1 and said there probably isn't too much data out there that would be negatively affected by them. We agreed that, before making a backward-incompatible change in a new version of Level 1, that we should obtain the consent of (at least) BioCyc and WIT.

Alan said that since PSI-MI has state information, a translation of PSI-MI into BioPAX Level 2 will be lossy. He warned, because of this, that we should not ask data providers to make a huge commitment to translate their PSI-MI data into Level 2, as they would then be less likely to create a converter to BioPAX Level 3 (which presumably will have a better representation for states). We agreed that we should take on the burden of creating a translator for PSI-MI into BioPAX Level 2, instead of asking individual data providers to do so.

We next talked about the Level 2 release roadmap. The release candidate will contain an updated v0.9 OWL file and the pdf documentation. We will also try to include some examples, perhaps from Reactome (Imre was ok with this idea). The final release of Level 2 will have polished examples. We will talk about what examples should be included in the final release on the discuss list. The final release will also contain a statement on our backward compatibility policy (to be worked out on the wiki).

We decided not to make a 1.6 Level 1 update a requirement of the Level 2 release. The Level 1 update will be an independent track and Level 1 data providers will be involved in that process.

Gary and Mike will try to put out the release candidate by Friday, August 5th. They will follow the normal process for releases (see Release Checklist).

Christian Lemer's team has been working on this - they are finalizing documentation and waiting for feedback from KEGG.

Jeremy spoke to Mustafa. He only has weekends to work on it; Jeremy will touch base with him (hopefully) this weekend.

Reactome data in BioPAX format (Level 2, v0.9) is available on their website, thanks to Guanming Wu's efforts.

The organizing committee sent out invitations to speakers yesterday. The wiki page is up to date on details of the meeting. Alan will probably give a talk there, as will Imre, Christain and ~7 others.

Speakers' travel expenses will be covered. The details of travel funding for other participants hasn't been worked out yet. The slot for an OWL expert hasn't been filled yet, Alan will suggest some names to the organizing committee via email.

We next discussed if/how we should endorse wiki pages. Gary said that pages that present new content should be discussed before being endorsed, but pages that simply record decisions made elsewhere (e.g. via an email list or at meetings) should simply be endorsed (yes/no) on the monthly conference call.

Christian Lemer asked how we should discuss / comment on endorsed pages. We could divide the page into two parts (endorsed section, comments section), or maybe use style sheets to differentiate endorsed content from non-endorsed content. We agreed to review and discuss the endoresment process, then create a wiki page on it and then endorse that page, then consider other pages to endorse.

2:09 PM - Ending the call proper.

Jonathan Rees said it seems funny to call them best practices when we are really talking about things in the spec that can't be specified by OWL. They are really more than best practices. Dan Corwin suggested we change the name from best practices since we are not using the term in the normal way. Alan said we need to determine the extent to which we represent things in OWL and why, in cases where it is possible, we don't move further toward "all-out OWL". What other design goals interfere with going all-out OWL?

The reason Alan started the page was to get an idea on what was being talked about when people said something was a "best practice". We decided to rename the page something like "On the issue of when we use OWL semantics and when we don't", to focus the page on the question of when and why we use these things we're calling best practices instead of OWL axioms, constraints, etc. Gary said that these decisions have been made on a case-by-case basis so far. We agreed that the issue was a "standards issue" (i.e. relating to exactly in what ways and how strictly BioPAX is a "standard"). Gary suggested we should add a FAQ item for this issue, but we agreed it should be resolved on its own wiki page before doing so.

The last comment during the call was by Jonathan, who said that he wished more people were spending time on the WIKI.

-End of Call-

We did not get to discuss:

Addendum - after the meeting.

Jonathan added a page To OWL or not to OWL in response to the best practices discussion.

last edited 2005-09-29 06:30:04 by AlanRuttenberg