Fig. 3

a Ranking (x-axis) of breast tumor (BRCA) samples according to ESR1 expression (gene TPM) (y-axis). ER+ and ER– subsets were defined as the top and bottom 25 % of the ranking, respectively, leaving out samples in the interquartile range (IQR). b Heat map of PSI values, from 0 (blue) to 1 (yellow), for the top 35 isoforms that separate ER+ and ER– subsets. Isoforms are labeled by gene name (y-axis). Samples are clustered according to the PSI values using Euclidean distance and Ward’s method. c Accuracies in terms of AUC values (y-axis) from a tenfold cross-validation for the transcript isoform signatures for the comparison of ER+ and ER– samples, and for the comparison of early and late N, S, and T stages within ER+ or ER– subsets. The variation on each bar indicates the minimum and maximum AUC values. d PSI distribution of the isoform in TNFRS8 that changes between early and late S stage in ER– samples (Wilcoxon test p value = 0.1046). e Accuracies in terms of AUC values (y-axis) from a tenfold cross-validation for the transcript isoform signatures (Iso-model) and the gene expression signatures OncotypeDX, MammaPrint, and PAM50, indicated in grayscale. Each signature was tested to predict the separation of ER+ and ER– breast tumor samples or the separation between early and late (N, S, and T) stage in ER+ or ER– separately. The variation on each bar indicates the minimum and maximum AUC values