Fig. 3

Performance of DiffMut vs other methods. The log₂ fold change in AUPRC when ranking genes using our method, DiffMut, vs MutSigCV [8], the method developed by Youn and Simon (YS) [11], OncodriveCLUST [29], OncodriveFML [30], and MADGiC [10], when evaluating performance in identifying cancer driver genes from the Cancer Gene Census (CGC) [26] (left), the subset of these genes that are oncogenes (middle), and the subset that are TSGs (right). For identifying all cancer genes, differential mutation is computed based on all non-silent mutations, whereas for oncogenes and TSGs, it is computed based on only missense mutations and only nonsense mutations, respectively. Entries with a dash indicate cases where MADGiC could not be run