Skip to main content
Fig. 2 | Genome Biology

Fig. 2

From: Phiclust: a clusterability measure for single-cell transcriptomics reveals phenotypic subpopulations

Fig. 2

Application of phiclust to a BMNC data set drives the discovery of biologically meaningful sub-clusters. a UMAP of BMNC data set. b Phiclust for the BMNC data set. Error bars indicate the uncertainty obtained by resampling the noise. Inset: UMAP of clusters with low, intermediate, and high values of phiclust. c Singular value distribution, MP distribution (red line), and TW threshold (green line) of clusters with low, intermediate, and high values of phiclust. Significant singular values are highlighted with asterisks. In the gdT cluster, the singular vectors corresponding to the outlying singular values had normal distributed entries and were thus not significant. d First three graphs: first singular vector of the red blood cell progenitor cluster in the BMNC data set versus normalized total counts per cell, normalized expression of ribosomal genes, and normalized expression of mitochondrial genes. Rightmost graph: second singular vector versus normalized G2M score. The dashed line indicates the linear regression and the grey area indicates the standard deviation. e Left: UMAP of the MAIT cell cluster in BMNC data set. The color indicates the normalized total counts per cell. Middle: singular value distribution, MP distribution (red line), and TW threshold (green line) for the MAIT cell cluster. The only significant singular value is indicated by an asterisk. Right: normalized total counts per cell versus the singular vector associated with the significant singular value (here: first singular vector) in the MAIT cluster

Back to article page