Supplementary material for Virtaneva et al. (2001) Proc. Natl. Acad. Sci. USA 98 (3), 1124–1129. (10.1073/pnas.031566698)

Supplemental Figure 5

Fig. 5.

Two-way hierarchical cluster analysis of AML and CD34+ samples. Log-intensity values of 1,959 genes, passing a variation filter, from 10 AML+8, 10 AML-CN, and 7 CD34+ samples were analyzed by using the programs Cluster and TreeView (2). The dendrogram represents similarities in the expression patterns between experimental samples (see Fig. 1B for sample labels). (A) Scaled-down representation of the 1,959 gene cluster diagram. Each column represents an expression profile of a single experimental sample; each row represents a single gene. Blue squares, transcript levels below the median; black squares, transcript levels equal to the median; red squares, transcript levels greater than the median (see scale at bottom). The colored bars in the center (BD) identify the location of three clusters displayed to the right. (B and C) Two representative clusters with genes of interest (highlighted in red and discussed later in the text; also see Fig. 4 A and B) and known function in apoptosis (B) or cell adhesion (C) are indicated specifically. Pairs of these genes (red highlight; e.g., BCL2 and TNFRS1FA) define boundaries of expanded subsets shown with the sample clustering to illustrate other genes in each cluster with a similar expression profile. (D) This is a large, intense cluster dominated by ribosomal genes (60 of the 96 genes in this cluster are filtered ribosomal genes from a total of 64 such genes on the HuGeneFL array). This cluster was highlighted, because it strongly influences sample clustering; however, it does not serve to differentiate AML+8 from AML-CN. Interestingly, intense ribosomal expression in cancer has been noted previously, but its significance and the conditions that produce it are currently poorly understood (5, 6).