Figure 4

Decision tree generated by training the J48 algorithm on the normal druggability datasets This decision tree was generated by training the J48 algorithm on the normal druggability datasets (see “Methods”). The uppermost ellipse is the node root of tree that represents the most important condition for discriminating druggable genes from non-druggable genes. In this case, such condition is the plasma membrane localization of encoded proteins. The remaining ellipses are internal nodes that represent additional conditions for considering a gene as druggable or non-druggable. In the left branch of tree, such conditions are a central position in a transcriptional regulatory circuitry (inbetreg) and being an enzyme (metin). The rectangles depict genes that, under certain conditions (represented by the root node and internal nodes), are respectively and predominantly classified as druggable (True) and non-druggable (Unknown). In the round brackets inside rectangles, the number before the slash indicates the total number of genes that are actually druggable or non-druggable and the number after the slash indicates how many genes were incorrectly predicted.