Fig. 6
From: ACES: a machine learning toolbox for clustering analysis and visualization
Type-I diabetes patients and controls shown by PCA 2D scatter. Shown are three clustering methods for the live diabetic samples (DIVID1 through DIVID6), brain dead early-onset diabetics (H911 and H1204), and three brain dead controls (H1778, H1499, and H1530), for genes in the insulin secretion pathway, which should separate by cases and controls (a). By contrast, the complement pathway, which is unrelated to insulin metabolism, separates samples by brain dead versus live patients (b). Sample DIVID6 was identified as outlier in the original study