Figure 4

Unsupervised hierarchical clustering analysis. The unsupervised analysis is based on 2821 interquartile range (IQR) filtered probe sets of the HG-U133 Plus 2.0 microarray of the 99 experiments included in the study. The signal used is PQN. The three major clusters that were identified by the algorithm represent B lineage ALL (orange), T lineage ALL (blue). and AML (green) leukemia types. Then the dendrogram splits and samples are subdivided according to leukemia subtype characteristics: 1. Pro-B-ALL with t(4;11); 2. c-ALL with t(9;22); 3. T-ALL; 4. c-ALL with t(12;21); 5. Pre-B-ALL with t(1;19); 6. ALL with hyperdiploid karyotype; 7. c-ALL-Pre-B-ALL with DNA-Index DI = 1 and negative for recurrent translocations; 8. AML with t(11q23)/MLL; 9. AML with normal karyotype or other abnormalities. The graph on the left shows the correlation between distances for clustering validation (0–1-vector where 0 means same cluster, 1 means different clusters). Samples are labeled by patient numbers (#1 – #27) and total RNA extraction methods (method A, method B, or method C). For patient samples #25, #26, and #27, three individual technical replicates were performed.