Figure 3

Increased entropy in metastatic breast cancer is independent of breast cancer cohort. A) Comparison of local entropies for the proteins with degrees ≥ 10 between the non-metastatic, metastatic and random weighted networks in the three validation cohorts: Mainz (1900 genes), Frid (1270 genes) and LoiUnt (1899 genes). The entropies in the metastatic network exhibit significantly higher values than those in the non-metastatic network: P-value given is from a one-tailed paired Wilcoxon rank sum test. Both non-metastatic and metastatic networks show significantly lower entropies than those of a purely random network obtained by randomisation of expression profiles. B) The top 200 genes with most significant P-values (FDR < 0.001) as identified in the EMC (discovery) data set are grouped according to increases in entropy (1) or decreases (-1). The numbers of genes within each group and represented in the validation set are given just above x-axis. The y-axis labels the statistics (dS _i /σ _i ) of these genes in the corresponding validation set, and the boxplot allows a direct comparison to be made. The P-value is from a one-tailed Wilcox rank sum test as the alternative hypothesis is that the statistics should be larger for those genes identified to undergo increases in entropy in the discovery set.