Fig. 4

Sub-clusters within premenopausal (preM) estrogen receptor-positive (ER+) tumors. a Clustering result of The Cancer Genome Atlas (TCGA) gene expression data by sparse k-means yielded three clusters (n₁ = 36, n₂ = 29, n₃ = 4). b Sparse k-means clustering in Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) preM ER+ tumors using genes selected by TCGA (n₁ = 56, n₂ = 19, n₃ = 43). c Kaplan-Meier survival curve using clustering result from sparse k-means analysis from METABRIC data. d Semi-supervised clustering of METABRIC gene expression data from preM ER+ tumor. e Kaplan-Meier survival curve of three groups from semi-supervised clustering of gene expression from d. f Kaplan-Meier survival curves of groups resulting from four different clustering algorithms, with the p value from the log-rank test. Lum luminal, Her2 human epidermal growth factor receptor 2