Teleconference about metadata interoperability

The iPToL TreeViz working group recently met in Austin to discuss our prototype software for visualizing large trees. One of the key features we plan to implement next is the display of metadata in the tree viewer (text labels, images, colour / shading on branches, icons on nodes, etc). In addition to identifying the types of metdata that we need to display, we need to design an API for attaching metadata to trees.

Agenda

1. What solutions can we use in the development and early release phases of the project? 
2. What is the optimal solution, and how do we move towards that goal (i.e. so that the work we do now is compatible with the long term goal).Developing syntax for NeXML might be part of this goal. 
3. Which groups are responsible for which pieces? We have been imagining a model where Core Software / Data Interop writes translation layers that allow output from Trait Evolution to be input into the viz tool. But, perhaps there is a better model.

Executive summary:

Attendees: Karen Cranston, Adam Kuback, Brian O'Meara, Bernice Rogowitz, Rutger Vos

1. Agreement on adopting NeXML as the file format for trees and metadata, although we will need to be flexible in the short term before NeXML ready.
2. TreeViz will work with Trait Evolution (Brian O'Meara, etc) to develop initial use cases and sample metadata.
3. Development of phyloreferencing syntax will simplify our ability to connect metadata with trees.
4. Identified need to balance performance with flexibility when rendering metadata. The viewer should have default behavior for dealing with metadata while panning and zooming (i.e. choosing a color for a collapsed clade). There should be options controllable by the user, although the options will be defined by scalability issues. User should always have fine-grained control over the final image for purposes of communication.
5. We didn't address point #3 on the agenda due to Core Software being busy this week with the Discovery Environment release.

Transcipt:

RV: are we interested in language for events on tree or just for visualization?
KC: initially viz (first) but want interface to allow for mapping to biological concepts
RV: interop with other apps; Mesquite, FigTree, etc. those have no idea about events but do about colours, viz. elements
BR: carry semantics of viz. elements across applications; viz. elements agnostic about biology but want to ultimately link
KC: Trait evolution as initial use cases from BO group, see wiki
BO: Felsenstein threshold model can't be currently visualized
BO: Probably some time before Trait Evol. will have output data
BR: can we get a simple data set to start with?
BO: Use Mesquite sample data sets; simulate larger data sets with Mesquite or R
BR: Sample data for multiple linked representations
KC: Can we use NeXML as default file format? Initially, we probably want to import simple formats like csv while we develop NeXML viz namespace.
RV: NeXML uses namespaces which would be useful for viz elements
BO: pointing to nodes: internal nodes can be referenced in NeXML, what about MRCAs?
RV: NeXML does not have syntax for phyloreferences
KC: couple of proposals for phyloreferences (VoCamp, Thompson); Interop group (Piel) very interested in phylorefs
BR: don't we still need specific use cases? How many different items do we want people to be able to map?
BO: What is the development model: we provide use case & developers figure out method? Or we say "I want to paint branches red"?
BO: Simmap XML has a clunky hack for the types of data in the use cases
KC: there will be scale issues when we display metadata on very large trees
RV: semantics could be endless, e.g. for collapsed clades. Want fine control over viz. elements, user can say I Want Color XXX Here. Separate reason for color from ability to provide specific color.
BO: The existence of collapsed clades comes from viz. tool, not specified by user.
KC: Balance between scalability of default behaviors in exploratory mode
RV: How do we keep momentum? wiki, email, other calls
KC: Will post minutes to wiki, create new pages for viz. vocabulary and use cases; Probably second call in a couple of weeks.
BO: suggest IRC meetings

Action items:

• Karen will create new Developer pages for viz element namespace, use cases
• Brian and Karen will work on large-scale data sets for the use cases