Figure 3

Comparative driver gene predictions in breast cancer and thyroid cancer. (A) The candidate driver genes predicted by DOTS-Finder in BRCA are compared against four previously reported predictions: MuSiC, MutSig, TUSON Explorer and the TCGA publication. The five-set Venn diagram shows the number of predicted genes in common between the different analyses and those uniquely predicted by each of them. The line delimiting each set and the name of the corresponding method are depicted in the same color. The diagram uses a graduated color ramp from light yellow to dark red to represent the overlap of an increasing number of tools that predict the same drivers. Although the BRCA mutational landscape is highly heterogeneous among patients, all the methods agree on predicting the same 17 genes as drivers (darkest shade of red). In addition, DOTS-Finder is able to predict seven genes that were never found by any method in BRCA. Also MutSig and TUSON Explorer retain unique predictions (11 and 5 possible driver candidates, respectively). This discrepancy is a reflection of the typical 'mountains and hills' landscape of the BRCA genome[4], with few highly mutated genes (predicted by almost all the tools) and hundreds of low-frequency mutations (only identified by a specific tool). (B) Number of genes predicted by TUSON Explorer and DOTS-Finder in the THCA dataset. The former predicts only few driver genes (6); of these, two-thirds are also identified by DOTS-Finder. Notably, our tool shows a much higher sensitivity than TUSON Explorer with 12 new predicted genes.