Fig. 6.

Estimating dimensionality. a Estimated dimensionality using the proportion of variance in each component explained by the true subpopulations. b Difference between “real” dimensionality (from a) and dimensionality estimation methods, along with computing time. c Average silhouette width per subpopulation across a selection of methods, combined with sctransform or standard Seurat normalization. d Clustering accuracy across the normalization/dimensionality estimation methods. The color-mapping scheme is the same as in Fig. 4c