Figure 4

Perturbation signature activity scores recapitulate known subtypes of ER+ breast cancer. ConsensusCluster heat map of perturbation activity scores across the 774 ER+ breast cancers of the METABRIC discovery set. Samples clustered using the ConsensusCluster algorithm (the five-cluster solution is shown). Distribution of breast cancer intrinsic subtypes (PAM50) across the five inferred clusters is indicated (sky blue: luminal-A, blue: luminal-B, green: normal-like, pink: HER2+, red: basal). Perturbation signatures have been clustered using hierarchical clustering (dendrogram has been suppressed) according to similarity of activity profiles over samples. Color bars to the right depict the Pearson correlations of the perturbation activity scores with the GGI and MPI. ER+, estrogen receptor positive; GGI, genome grade index; MB-D, METABRIC discovery set; MPI, molecular prognostic index.