“Supervised” hierarchical clustering and principal components analysis (PCA) using purine pathway genes discriminating TEL-AML1 and non–TEL-AML1 B-lineage ALL. (A) Each patient's ALL cells are depicted as vertical columns, with green symbols indicating ALL with TEL-AML1 fusion (n = 14) and red indicating non–TEL-AML1 (n = 24) ALL. Eighteen gene probe sets (rows, with gene names shown) were used for the hierarchical clustering. The relative level of gene expression is depicted from lowest (green) to highest (red) according to the scale shown at the bottom. (B) PCA plot using the 18 selected purine pathway gene probe sets (U133A array) discriminating TEL-AML1 versus non–TEL-AML1 in the training set. (C) PCA plot using purine pathway gene probe sets on the U95Av2 array that correspond to the probe sets discriminating TEL-AML1 versus non–TEL-AML1 identified in the training set.