Fig. 2

Overall performance and impact of cell type numbers on the number of cell type estimation. a Deviation of the estimated and the true number of cell types for each of the 14 clustering methods. Positive deviation represents over-estimation and negative deviation represents under-estimation. The true number of cell types ranges from 5 to 20 and each was repeated 10 times to capture the estimation variability. b Density plot summarising overall deviation across all numbers of cell types in a. c Concordance of clustering output and predefined cell type labels as quantified by four concordance measures. Each bar represents the average performance across datasets with 5 to 20 cell types, and error bars represent the standard deviation