Figure 2

Network graphs derived from six cancer datasets; a) breast carcinoma, b) colorectal carcinoma, c) DLBCL, d) glioma, e) ovarian carcinoma and f) testicular germ cell tumours. Each dataset studied had its own idiosyncrasies owing to the tumour-specific biology each represents and the high degree of inherent variation in gene expression data derived from cancer samples. In order to visualise and analyse a large proportion of the expressed genes in these samples we aimed to construct graphs of data derived from individual cancer types based on just under half the transcripts represented on the chip (18,000-23,000 probesets). For these reasons relatively low correlation thresholds were used for graph construction i.e. between r = 0.65-0.75. The resultant graphs of individual cancer datasets are highly structured and composed of a large number of nodes and edges (18,934 - 23,015 nodes, connected by between 268,471 - 954,082 edges, see Table 1 for details).