Extended Data Fig. 2: Analysis of cluster stability. | Nature Computational Science

Extended Data Fig. 2: Analysis of cluster stability.

From: Dynamic visualization of high-dimensional data

Extended Data Fig. 2

Analysis of cluster stability. (A) Animated t-SNE visualization of single-cell transcriptomic data from human bone marrow (n=1000). (B) Distribution of cell-wise silhouette coefficients across bootstrap visualizations (B=20) of the CLP cluster and of other candidate cluster identified in Figure 1D on the t-SNE visualizations shown in panel A. Statistical significance was assessed with the two-sided Wilcoxon rank-sum test. Center line represents median, box represents interquartile range (IQR), whiskers represent range up to 1.5 × IQR. (C) Same as in panel A except for animated PCA visualization. (D) Same as in panel B except for animated PCA visualization. (E) Same as in panel A except for animated UMAP visualization of the entire human bone marrow data (n=5780). (F) Same as in panel B except for animated UMAP visualization of the entire human bone marrow data (n=5780). (G) Animated visualization of spectral data of stars, quasars, and galaxies from the Sloan Digital Sky Survey (n=1000) queries stability of clusters identified in one visualization (leftmost panel) across other visualizations. DR visualization settings are outlined in Methods.

Back to article page