Skip to main content
Fig. 2 | Genome Biology

Fig. 2

From: MUFFINN: cancer gene discovery via network analysis of somatic mutation data

Fig. 2

Assessment of predictive power of MUFFINN for cancer genes. ROC analyses on prediction of cancer genes annotated by CGC (a) and the 20/20 rule (b) were performed for MUFFINN with various combinations of two network algorithms using direct neighbors (DNmax and DNsum) and two networks (HumanNet and STRING v10) and conventional frequency-based methods, MutSig 2.0, MutsigCV, and MutationAssessor, with mutation data derived from breast cancer type (BRCA). The same analysis was repeated for all 18 cancer type samples and the results are summarized as the distribution of 18 AUC scores for cancer genes annotated by c CGC and d the 20/20 rule. Prediction powers for top candidates were assessed by cumulative numbers of retrieved cancer genes annotated by e CGC and f the 20/20 rule within the top 100, 500, and 1000 with the same analysis setting. Results from all the assessment tests indicate the generally improved performance of MUFFINN over the tested gene-centric cancer gene classifiers. FPR false positive rate, TPR true positive rate

Back to article page