Figure 2.

Unsupervised clustering of PanNET samples (primary and metastasis/recurrency) of four patients with multiple samples and three normal pancreases. Hierarchical clustering shows the segregation of samples according to their grade based on Ki67. estimation in the first step and to the belonging patient in the second. Major clusters are A (all PanNETs) and B (all normal tissues). Cluster A is further subdivided into subgroup A1 (all G3 samples) and subgroup A2 (all G2 samples). Samples from the same patient within clusters A1 and A2 are most similar to one another (smaller dendrogram distance).