Multidimensional scaling to visualize classes. Scatter plots visualizing the two classes in each of the data sets. Multidimensional scaling was performed on the correlation based distance matrices producing scatter plots that try to preserve the distances between observations. The two classes are plotted as either red diamonds and blue circles, corresponding for each data set to the classes; Alizadeh: DLCL and other, Finak: epithelial and stromal tissue; Galland non-invasive and invasive; Herschkowitz: ER+ and ER-, Jones: cancer and not cancer, Sørlie: ER+ and ER-, and Ye: P and PN. These figures illustrate how easy it is to separate the classes (in two dimensions) e.g., Alizadeh have quite well separated classes, whereas Sørlie have two classes that are not easily separated.