Fig. 2
From: EmptyDropsMultiome discriminates real cells from background in single-cell multiomics assays

Simulations to assess the performance of different cell calling approaches. A Scatterplot showing the total number of RNA and ATAC counts for the 5000/2000 simulated dataset with one new cell type. Ground truth of the simulated dataset: all the empty droplets from the PBMC dataset and additional simulated empty droplets, 5000 nuclei containing droplets, 2000 small simulated nuclei by subsampling RNA + ATAC and then scrambling the genomic profile of monocytes. B Result of applying EmptyDropsMultiome at FDR = 0.1% on the 5000/2000 simulation. The nuclei containing droplets are singled out from among the empty ones. The solid line shown is the k-means line used by CellRanger-arc, which misses all the points below it. C ROC curve comparison on the 5000/2000 simulation of EmptyDropsMultiome against a customizable version of CellRanger-arc where we change the intercept of the k-means line (while maintaining its slope)