Saturday, March 27, 2010

Coming this summer plus Phenoscape in Chicago last week

I've registered for the Evolution and iEvoBio meetings this summer. I agree with the desire of the iEvoBio organizers to make a place for informatics approaches at the Evolution meetings - I've certainly given both talks and posters where I got the strong sense that I wasn't speaking to the right audience. There are several presentation plans I could have followed. What I have chosen to do is to present Phylontal as a poster at Evolution, explaining the need and approach in detail, something like the brown bag I gave at NESCent in December. Although it may not be the best venue, a poster seems the best format for Phylontal at this time and it helps justify my going to Evolution as well as iEvoBio. At iEvoBio, I'll give a lighting talk on pipelining OwlWatcher and Phylontal. The third player in this chain - a character constructor - I'll leave for later.

I have made an 'in-progress' release of Phylontal at Google code, but there's no rush. It loads up NeXML and OWL files, allows the user to assign ontologies to tips, then allows the user to lexically match terms from two sister taxa. It's as much a proof of concept for the frontend, operating as OWL + per taxon ontologies. There is another potential use case for phylontal - starting with annotated matrices, extract taxon+anatomy term pairs from annotated NeXML matrices with OBO ontology support and a separate NeXML file for the tree. This would approximate the usage in Phenoscape, and the desired output would be Phenoscape's multi-column homology table.

Chris Mungall gave two interesting presentations on reasoning with homologies at a Phenoscape meeting at the Field Museum in Chicago last week. He brought up some interesting cases involving '2 taxon' versus '3 taxon' homology relations, and an important discussion of the interaction of homology statements and is_a hierarchies. People were most impressed with Chris' approach being extensible to serial homologies, the most important point being that serial homologies strictly exist only within an individual. I think that is an important insight and it lead to a simple taxonomy of homology relationships that actually make sense for anatomical reasoning. The corresponding treatment for behavior classes is worth looking into (e.g., distinguishing similar actions by an individual animal vs. homologous behavior patterns).

There was also some discussion of individuals and inferring classes by abduction or generalization over dinner at some point. I'm beginning to think this may be an important new growth area for biological ontologies, and it was good to hear that several people were thinking in this direction.

Finally, after several experiments, I think the OwlWatcher player is settling down towards a usable configuration using only two threads, rather than the initial four. Still need to deal with the packaging issue, and I fear the first post-Quicktime release won't be an easy install.

I didn't submit a proposal for a phyloinformatics summer of code project this year; I'll help out if it's appropriate for someone else's project, but I didn't have any brainstorms this year. None the less, if you're a student reading this, you will probably find one or more projects of interest there.

No comments: