Our widgets for functional genomics use Orange, a data mining
and machine learning suite. Orange can be accessed through scripting in
Python,
or by visual programming in Orange Canvas.
In functional genomics, we have also designed a web-based tool for mutant data analysis called GenePath (also featured
in Science's NetWatch).
This widget provides means to present the results of the k-means
clustering. Details assigned to genes in the input data file can be
displayed. In our example, we display GO annotations. With the
"Merge" options we can set the level of image
granularity. In order to display all genes on screen and get a better
overall picture, we have merged groups of 21 genes (rows, option
Merge->Rows) and displayed their average expression as single
rows. Moving the mouse over a specific time point and row displays
details about all the genes in a row.
Clusters "1" and "5" are similar to the ones
discovered in (Van Driessche et al., 2002). There is a dramatic change
in expression between 6-8 hour in both groups. Genes in cluster
"5" have a lower-than-average expression during growth and
early development (0-6 hour) and a higher-than-average expression in
later times (8-24 hour). Genes in cluster "1" have a
higher-than-average expression during growth and development (0-6
hour) and a lower-than-average expression later in development (8-24
hour). Further analysis with the "GO Term Finder" widget will
show the presence of genes associated with cysteine protease and
vegetative ribosomal genes. This concords with the findings in (Van
Driessche et al., 2002).
We have selected all genes from cluster "1" for further analysis. These genes
are sent as a new data token to the four linked widgets: "Scatterplot", "Heat
Map (2)", "GO Term Finder" and "Genome Map."