Fig. 3

Hierarchical cluster trees and outlier detection RNA-seq data from a kidney tumor study. Clustering was performed using the ‘single linkage’ method which is generally recommended to identify outliers. Usually, outliers are selected by subjective judgment of a researcher as samples on branches that separate locally or by conspicuous height from the remaining branches. To select outliers in a more objective way, boxplots on the height of the branches can be used. A disadvantage of hierarchical cluster trees is that outlier detection and arrangement of the tree depend on whether experimental groups are displayed separately or as a whole