Table 1 Clinical characteristics of k = 2,3,7 clusters: The ER, PR, Her2, Node, stage and grade status of the samples in each cluster are shown for k = 2,3 and 7. ND stands for "Not Determined". At k = 2, the clustering splits the data into normal samples and disease samples (BCA), except for one ADH which is classified with the normals. At k = 3, the BCA samples split into high grade (grade 2 or 3) and low grade (grade 1 or 2) categories. At k = 7, the low grade samples split into two clusters LG1, LG2 and the high grade into four: HG1 – HG4. The HG1 samples are all ER-, PR- and mostly Her2-. The HG3 and HG4 clusters are mostly ER+, PR+, Her2-. The HG2 cluster has mixed ER, PR and Her2 signatures. Using the Sorlie et al classification, we identify HG1 as the Basal-like subtype; LG1 as Luminal A; LG2, HG3 and HG4 as Luminal B and HG2 as the Her2+ subtype. When the sum of the entries for ER/PR/Her2/Node/Grade do not add up to the size of the cluster, it is because the corresponding information was missing in the dataset [10].

From: Portraits of breast cancer progression