Figure 3

SOM Clustering of data combined using CROPPER. Four different Parkinson's disease datasets were combined by aligning the metagenes with CROPPER. The data consisted of a total of 9 conditions originating from the datasets. The conditions are shown in the ordinate axis and their z-transformed values are shown in the y-axis. For normalization, the differences in the data value distributions were z-transformed. This was followed by calculation of the z-ratio, in which the differences of the z-values of the treated samples were subtracted from z-values of the controls (for method details, see Cheadle et al. 2002 [7]). The limit for significant alteration was z-ratio ± 1 defined as more than one standard deviation in the z-values of control and treatment data points. The gene expression data from human represent 4055 metagenes. These were clustered into 16 clusters using a self-organizing map (SOM). The expression profiles of the metagenes with altered expression in both human and animal data sets were coloured in green. These 247 "green" genes were considered to be candidates for human neurodegenerative diseases. The genes with altered expression only in the animal experiment datasets, but not in the human datasets were coloured in red. The 225 "red" genes may suggest mechanisms in animal neurodegeneration models, but not in human Parkinson's disease. The separation of red and green genes to peripheral clusters indicates good clustering resolution. The lists of "red" and "green" metagenes were further analyzed for the enriched human KEGG and GO terms based on the human gene identifiers corresponding to the assigned metagenes.