Hierarchical clustering structure and stability. (A) Hierarchical clustering of the organs based on mean expression levels across organs for all genes. The branch points are enumerated for reference in the text and Supplemental Table 2. (B) The heat map visualizes the variance of the distance (see Methods) in the HC hierarchy estimated from 1000 HC trees using randomly selected subsets of genes of size 1000. (Black) Variability close to the average for the whole dendrogram; (blue) a lower than average variability; (red) an increased variability. The interpretation of the increased variability indicated by the red squares is that the corresponding organs are close in some of the HC graphs, while they are further away in other HC graphs. (C) The variability index for the whole dendrogram is plotted as a function of the size of the sets of genes drawn randomly from the genes on the microarray. The black line represents the variability index for randomly generated gene expression data (see Methods), the red line for the entire data set (19 organs), and the green when ovary and stomach, which showed the highest variability, are removed from the analysis.
