Fig. 1

Outlier detection. Outlier detection method is reported. a Statistical detection: for each kinase, gene expression level in all the analysed samples belonging to a specific tumor type is reported as an histogram (left panel) and as boxplot (right panel). “Rare events” (a kinase over-expressed in one or a few cell lines and low/not expressed in the others) are identified by mean of the Grubb test and reported as a red circle. b Prioritization and filtering: the most relevant outlier kinases are selected applying specific filter criteria (minimal expression threshold; maximum median level of expression over the tumor type; minimum distance from the 75th percentile of the tissue-specific distribution; proportion of the number of outliers with respect to the whole dataset of outlier occurrences). Samples that do not consistently pass the imposed filters are removed (reported in the figure as red crosses)