Fig. 1

Schematic summaries of (a) benchmark workflow and (b) clustering stability measure. a Summary of the benchmark workflow. A panel of fourteen scRNA-seq clustering methods that perform the estimation of the number of cell types were evaluated under four main settings for creating different data characteristics via sampling from the Tabula Muris and Tabula Sapiens data. Evaluation includes deviation from the true number of cell types, clustering concordance with predefined cell type labels, and computational time and memory usage of each method. b Illustration of clustering stability, implemented as part of the single-cell Consensus Clusters of Encoded Subspaces (scCCESS) [33], for estimating the number of cell types in a given scRNA-seq dataset