Figure 1. High-performing clusters for... | Rockefeller University Press

Figure 1.

$High-performing clusters for 0–1-year-old agglomerative hierarchical clustering experiment. Each patient data point is a 5-dimensional vector reduced through principal component analysis (PCA), followed by further reduction via the t-Distributed Stochastic Neighbor Embedding (t-SNE) algorithm to project on a 2D space. The illustrations show the dominant disease group (DGS) in the enclosed cluster and its fractional proportions. The cluster phenotypes based on diagnostic codes and laboratory data are labeled in red italics for clusters 0, 1, and 4. Clusters 3 and 6 had no immune phenotype of relevance. Refer to the image caption for details.$

High-performing clusters for 0–1-year-old agglomerative hierarchical clustering experiment. Each patient data point is a 5-dimensional vector reduced through principal component analysis (PCA), followed by further reduction via the t-Distributed Stochastic Neighbor Embedding (t-SNE) algorithm to project on a 2D space. The illustrations show the dominant disease group (DGS) in the enclosed cluster and its fractional proportions. The cluster phenotypes based on diagnostic codes and laboratory data are labeled in red italics for clusters 0, 1, and 4. Clusters 3 and 6 had no immune phenotype of relevance.