-
Notifications
You must be signed in to change notification settings - Fork 1
Technical Teleconference October 9, 2012
Attending: Luke, Wes, Xiaoxiao, Curt, Charlie
Date: 10/9/2012
-
Generally trees in a single file will share tips but may have different branch lengths and different topology. We should be able to recognize the case when the tips are not the same, because the analyses that can be performed differ depending on if the tips match or not.
-
Sometimes trees will have a corresponding character matrix. The rows in the matrix should match the tips in the tree, as the matrix corresponds to characteristics exhibited by the organisms at the tips. Sometimes there is a mismatch between the taxa (the tree tips) and the rows in the matrix. Pruning is performed to cut out any mismatches. There is a function in Geiger called treedata which implements pruning by a simple algorithm. We should be aware of this and have some similar capability in Arbor.
Kitware has implemented a phylogram rendering view (using the vtk charting technology) that draws a tree and superimposes a heatmap of character data on the tips.
Kitware will be implementing an extension to the current phylogram rendering to allow the user to expand or collapse nodes of a phylogenetic tree interactively. This will help the rendering of very large trees because intermediate nodes can be collapsed. No deadline date is set. Kitware will let us know when it is ready for demonstration.
UCF students, Charlie, and Curt are working on a few different topics:
-
We are looking at storing geographic observation locations in mongoDB and supporting queries against the points. The points could be queried directly (e.g. "which observations are inside a geographic area?"), or used to develop probability maps, by fitting Gaussian distributions at the observation points and summing the contributions of all points into a single probability distribution.
-
Just beginning to look at parallelization technologies to be applied to algorithm(s) in Geiger to increase scalability for very large trees. This will be a slowly-developing area, but one student is studying parallel implementations, starting with GPGPU (general purpose computing on the GPU) techniques.
A note was taken to discuss parallel implementations with Luke and how they might apply to the Bayesian algorithms to be added to Geiger later.
OneZoom (a tree layout algorithm that handles a vary large number of nodes) will be available soon from Luke's colleague, James Rosindell. OneZoom, due for release around October 16th, has an existing implementation in JavaScript and may be suitable for eventual integration into the Arbor prototype, since there is an existing javascript web context available in the application.
PIs to meet next week on 10/16.