back to table of contents

B. CLUSTERING OF PATIENTS BASED ON GENE EXPRESSION PATTERNS

Fig. 2: "Unsupervised" hierarchical clustering and principal component analysis (PCA). (a) We used 8222 probe sets (fold-change, referred to as "relative change" in the published manuscript) and 8002 (post-treatment) for "unsupervised" hierarchical clustering (all genes after filtering). Post-treatment expression tended to cluster samples by lineage, ploidy and molecular subtypes, whereas this was not the case for the fold-change in gene expression. (b) The top three principal components using all genes on the array, did not discriminate by treatment, based on either post-treatment gene expression or fold-change in gene expression, reflected as distances in three-dimensional-space.