Fig. 1

Flowchart summarizing the overall methodology. Flowchart summarizing the overall methodology. The first step depicted in part-a consists of data processing and necessary filtrations of the input databases TCGA and IntAct. The second step depicted in part-b involves generation of pairs of normal/tumor graphs based on expression, mutations, and interactions data. Measures based on graph-centralities are employed on resulting graphs. Ten lists of genes, eight from centrality measures and two from control measures, ordering genes with respect to their computed weights are provided as output. The final step depicted in part-c consists of analyzing the ten lists with regards to ROC, precision/recall (P/R), and GO consistencies (GOC). Two datasets, NCBI BioSystems [37] and COSMIC [38] are employed in all three analysis, whereas for the GOC analysis an additional database, the GO database [39] is also employed. Among all tested centrality-based measures M _bw provides the best performance in all three analyis. The M _bw list is further analyzed in more detail by filtering it based on a maximum weight independent set (MWIS) formulation, and the top genes from the resulting filtration go through a final literature verification step. a Data preparation, b Algorithmic computations, c Analysis and evaluation