Figure 10

Analysis of the variable sets separately using PCA. Representation of the samples from the validation set by the projection onto the first two principal components from the copy number data (left panel) and gene expression data (right panel), extracted from the tuning set. Samples with different leukemia subtypes are shown with different colors. The E2A/PBX1 subgroup, well separated from the rest using the correlation-maximizing methods (Figure 5), is not emerging as a deviating group when the data sets are analyzed individually with PCA. The T-ALL subgroup (light green) has a specific gene expression pattern, and is deviating from the rest using PCA on only the gene expression data (right panel). However, since these subjects do not have a specific copy number pattern, they are less discernible using CCA.