Figure 5.

Unsupervised hierarchical clustering of immune cell subsets. Flow cytometry data for each timepoint were normalized to diagnosis median values and heatmap was drawn based on Euclidean distance and unsupervised hierarchical clustering methods for each timepoint independently. Annotations were added for Sokal risk at diagnosis, BCR-ABL IS% at each timepoint and study drug. Low Sokal risk patients at diagnosis clustered based on terminally differentiated effector CD8 + T cells, CD57+ CD8 + T cells and GrB+ CD8 + T cells, which are highlighted in each heatmap. At 12-month time-point, the majority of imatinib-treated patients cluster based on their T and NK cells phenotypes. Patients who started on bosutinib and switched to imatinib treatment during the 12-month follow up are marked with an asterisk.