Fig. 6

Combined score for viral oncogene-associated PANDA network and protein interaction network improves enrichment and specificity for identifying cancer genes. a PANDA transcriptional network for the transforming viral oncogenes. Sanger cancer genes are depicted as red nodes. Size of node is proportional to degree of node. b Line plot shows the ROC curves for overlap with cancer drivers for each of three network measures – degree in PANDA network, degree in yeast two-hybrid network, or combined network score. Bar graph (above) shows the odds ratio of the overlap between cancer drivers and the top 10 % (or top 57 out of 571) TFs. Bar graphs (below) show the significance of the enrichment in cancer drivers using three different statistical tests: W = Wilcoxon rank test, K = Kolmogorov-Smirnov test, F = Fisher’s exact test for the overlap between the top 5 % (or top 29 out of 571) ranked TFs and cancer drivers. Black dotted line indicates lower bound for statistical significance