Fig 1. Relative importance of mutation statuses of oncogenes for predicting drug activity against cancer cell lines.

(A) Gini impurity indices calculated for entire data set consisting of 225 drugs and 990 cancer cell lines. The 50 most important oncogenes are shown. (B) The top-ranking oncogenes for each individual drug were computed as those whose relative importance is >2 standard deviations above the average oncogene importance for that drug. The 20 oncogenes that are top-ranking for the greatest number of drugs are shown. Gini impurity indices for the mutation statuses were computed from random forest classification models at an IC₅₀ activity cutoff of 1 μM.