Fig. 4
From: Cytomulate: accurate and efficient simulation of CyTOF data
Performance of simulation methods against random masking of cell subtypes. For all three panels, randomly cell types of the Levin_32dim were masked and performance of each method was measured using log KL divergence plus 1. Blue bars represent the masked datasets, and yellow bars represent matched, unmasked controls (N = 20). The whisker on top of the bars indicates the variance. a Two cell types are randomly masked in each replicate. b Five cell types are randomly masked in each replicate. c 10 cell types are randomly masked in each replicate