User talk:Watkinson

From Informatics

Jump to: navigation, search

Transition Items

Item Transition To Progress/Status
caGrid Branch Merge Kiran
  • Plan to merge branch back in to HEAD and fix conflicting jar issues.
  • Merge was attempted on Wednesday, but with no success. It is proving extremely difficult, but another attempt will be made shortly.
MatrixREDUCE Kiran
  • Component is functional, but has complex Cygwin dependencies. Kiran is aware of these.
  • Goal is to eventually switch to a web service version of MatrixREDUCE, to be provided by Bussemaker's team.
Intragen Matt
  • Matt is familiar with the structure and design of the web application.
  • Will transition the fine details of the site.
Cutenet Kaushal
  • Kaushal is already an expert on Cutenet. Some of the Bison-related details of the integration need to be explained.
  • We now have Cutenet-to-Cytoscape working, so Kaushal can now be exposed to more details on the geWorkbench side.
ICTVdb Kiran/Matt/Ken
  • Ken has more domain knowledge than anybody else on the team, with the possible exception of Sun-Ah.
  • Matt is very familiar with the web application architecture.
  • Transition of some developoment details will need to occur between Sun-Ah and a full-time developer once she has completed the project.
Genepattern Kiran
  • Integration of two classifiers, K-Nearest Neighbors and Weighted Voting, are underway by the Genepattern developers.
  • Kiran already has a relationship, so he can take over this task.
  • Matt is familiar with the architecture of the classifier data structures in geWorkbench, so he can field questions in this area.

Development Log

TaskStart Date Projected End Date Actual End Date Progress/Status
geWorkbench Perpetual Perpetual
  • 11/14/06: Replacement clustering algorithm 'QuickClustering' has been implemented and is testing. This is very similar to the existing clustering implementation, but achieves better performance and slightly less RAM. Will be used for caGRID hierarchical clustering.
  • 11/15/06: Adding new distance metric ChebyshevDistance, which has interesting properties for hierarchical clustering.
  • 11/17/06: Fixed Mantis #749, and also resolved an issue where renaming a panel to the name of an existing panel causes unusual behavior. Also fixed #684, which caused the selector panel to hold its selections between data sets. This is not stable behavior, so it was removed.
  • 11/21/06: Working on caScript/caGrid issues.
  • 11/22/06: Made some changes on the caGrid branch to support caScript with Kiran. We made a new @Script method for hierarchical clustering (SOM and others to follow, possibly with some generalization). We added another @Script method to the project panel that allows a DSDataSet to be added to the project. Then we wrote up a script that loads a data set, clusters it, and adds the result to the project panel. A bug-fix was required to the caScript interpreter to support some variable assignments that we needed.
  • 11/22/06-11/30/06: Working on 1.05 mantis bugs.
MatrixREDUCE 10/26/2006 11/10/2006
  • 10/26/06: Eileen and Xiaoqing's computers are unable to run MatrixREDUCE due to a cygwin conflict. Will investigate this with Kiran.
  • 10/26/06: Multi-selection and export functionality underway.
  • 11/06/06: Multi-selection and export functionality complete, moving on to image export and print functionality.
  • 11/20/06: Bug #773 fixed-- issue where certain sequence files could not be found by the MatrixREDUCE DLL due to some unusual path manipulation in the DLL. Files are now always served to the program as relative paths, which don't appear to have the issue.
Intragen 10/26/2006 Unknown
  • 10/26/06: SSL functionality discussions with Aris and Hans-Erik.
  • 11/14/06: Tracking down file preparation bug-- can't find any logging data yet that indicates what the errors have been.
  • 11/15/06: A full-data request from yesterday appears to have finished successfully overnight. I just started a new request for all data and it is moving quite slowly. The apollo filesystem seems to be heavily stressed. This could be because it is also the filesystem used by the gaia cluster.
Geneways 9/15/2006 Unknown
  • 11/08/06: Discovered some issues with Cytoscape and the Adjacency Matrix data structure. These structures are dependent on microarray sets, which may not be appropriate for Geneways. Working on some refactors to excise the microarray dependencies.
  • 11/13/06: Investigated with Kaushal the ID system used by Geneways as well as some other DB structures. SwissProt seems to be the most universal way to access proteins in Geneways.
  • 11/14/06: Devised an API for Geneways integration. This is to be implemented by Kaushal and myself. The methods include:
    • getCutenetProtein(String) - Given a SwissProt ID, get a Cutenet Protein object (which includes the PK of the protein).
    • getInteractions(String,float) - Given a SwissProt ID and a threshold interaction strength, get all interactions involving the protein that are above a given threshold.
    • getInteractionStrength(String, String) - Given two SwissProt IDs, get the strength of interaction between them, or 0 if there is no interaction in Cutenet.
    • getAdjacencyMatrix(CutenetInteraction[], DSMicroarraySet) - Given Cutenet interactions and a microarray set, create an adjacency matrix for the proteins corresponding to probes in the microarray set.
  • 11/16/06: Implement AdjacencyMatrix creation aspect of the Cutenet API. However, this is not yet tested. The database side must be completed. Involved is a reverse mapping between SwissProt IDs and the probeset IDs of the array. This is currently generated on the fly as needed, but a more semi-permanent solution may be required.
caGrid 9/15/2006 Unknown
  • 11/29/06: Worked with Kiran to get the new Java implementation of Aracne running on the grid. Conversion was required between the stripped-down microarray and adjacency matrix classes of caGrid to/from the Aracne versions. This was fairly straightforward. Some Aracne arguments don't apply in the grid setting (such as file inputs/outputs) so some adjustments may be required, but it seems to be working fairly well.
GenePattern Integration Phase 2 12/1/2006 Unknown
  • 12/01/06: Phonecall with Aris, Marc-Denis and Josh Gould about writing classifiers (machine learning) in Workbench.
  • 12/07/06: Refactored the machine learning API to be in a more sensible package structure. Added a few minor features to the classifier, although more is needed there.
  • 12/08/06
Retrieved from "http://wiki.c2b2.columbia.edu/informatics/index.php/User_talk:Watkinson"
Personal tools