Fig. 7
From: Euclidean distance-optimized data transformation for cluster analysis in biomedical data (EDOtrans)
available at https://wis.kuleuven.be/statdatascience/robust/Programs/pooledVariableScaling/pvs-r.zip with addition of code implementing the present EDO transformation
The effect of variable scaling on the gene expression data. The data set comprised 65 surgical samples of human breast tumors in which hierarchical clustering of the expression of 496 intrinsic genes that showed significantly greater variation between different tumors than variation between paired samples of the same tumor had resulted in four distinct tumor types [32]. The dendrogram colors correspond to the tumor type: basal-like in red, Erb-B2þ in green, normal-breast-like in dark blue and luminal epithelial/ERþ in cyan. The EDO transformation generally yields superior recovery of the true clusters, comparable with the PVS transformation. The experiment is a re-run of the experiment performed for Fig. 3 in [6], using the R script