Error when using USC classification

Help
sebkg
2009-10-01
2013-05-20
  • sebkg
    sebkg
    2009-10-01

    I am getting the following error when using the USC classification:

    java.lang.ArrayIndexOutOfBoundsException: 0
    at org.tigr.util.FloatMatrix.<init>(FloatMatrix.java:127)
    at org.tigr.microarray.mev.cluster.gui.impl.usc.USCGUI.createNewExperiment(USCGUI.java:432)
    at org.tigr.microarray.mev.cluster.gui.impl.usc.USCGUI.trainThenClassify(USCGUI.java:228)
    at org.tigr.microarray.mev.cluster.gui.impl.usc.USCGUI.execute(USCGUI.java:144)
    at org.tigr.microarray.mev.MultipleArrayViewer$7.run(MultipleArrayViewer.java:2636)
    at java.lang.Thread.run(Unknown Source)

    I am using 2 classes, "Train & Classify" and default parameters. The actual data can be found here:
    http://sydneybioinformatics.org/download/data.txt
    http://sydneybioinformatics.org/download/assignments.txt

    I have tried running the exact same scenario on several different platforms (32bit XP & 64bit Vista).

    Can anyone offer any advice?
    thanks,
    Sebastien

     
  • Eleanor Howe
    Eleanor Howe
    2009-10-01

    Hi Sebastien,

    I've had a look at the USC code and tested the module out with your data (thanks for providing that, by the way - it was very helpful). It seems that the error appears when USC's result set is size zero. The stack trace is not a very friendly way of reporting results, so I've put this on our list of things to fix for v4.5, due out in November.

    In the meantime, that message is MeV's socially-inept way of telling you that the parameters you chose result in no significant genes being returned. The rest of the module should work fairly normally. Please let me know if that's not the case?

    Eleanor

     
  • sebkg
    sebkg
    2009-10-01

    Hi Eleanor,

    Thank you for your reply but unfortunately the problem doesn't seem to be that simple. For example,  I have just run a USC analysis using the data I provided you with. The first listed result has:

    8.7 #Mistakes, 11.7 Avg#Genes, 0 Delta, 0.5 Rho

    But when I try and save this result I get the error I reported…

    Thanks,

    Sebastien

     
  • Eleanor Howe
    Eleanor Howe
    2009-10-02

    Hi Sebastien,

    The first listed result (8.7 #Mistakes, 11.7 Avg#Genes, 0 Delta, 0.5 Rho) is not the final result of USC. It is a set of parameters that USC then uses to test your test dataset against the training set. The results of *that* test is the list of genes you would like to see. Unfortunately, for this set of parameters no significant genes are found.

    When I run USC with your data, the training set data file is written out fine, and the USC module proceeds to test the test dataset against the training set, using the rho and other parameters I select. If those parameters lead to a result set of genes that is size zero (the debugger tells me how many genes there are), the module chokes and throws the error you see.

    This error is definitely a problem; it's an inappropriate response to a result set of size zero, so MeV will be fixed to provide a more useful response for v4.5. Until v4.5 comes out, however, that error message is telling you that the parameters you selected result in no significant genes. I think you will find that the training set datafile is written out fine.

    Eleanor

     
  • sebkg
    sebkg
    2009-10-23

    Hi Eleanor,
    thanks again for your help. Now that I am satisfied with how the USC component is running may I ask whether it is possible to use USC in scripts. The USC button appears to be greyed out when I attempt to create a script.
    Thanks,
    Sebastien

     
  • Eleanor Howe
    Eleanor Howe
    2009-10-23

    Hi Sebastien,

    I'm afraid that right now the USC module isn't set up to be scriptable. I'll put that down as a feature request for a future release, though.

    Eleanor