Fig. 7

Application of ZetaSuite to single-cell transcriptomics. aζ-plot at each gene count bin over a full range of gene counts. Raw counts of each gene are plotted against the number of different genes detected. High-quality (orange dots) or low-quality (blue dots) cells are based on the benchmark dataset (E-GEOD-48968). b Same ζ-plot with cells colored based on %mt. c ROC curves deduced using different metrics. The p-values are calculated by plot.roc in pROC R package with default parameters. d UMAP of cells based on the CellRanger cutoff. e–g UMAP of cells based on the cutoff by CellRanger or EmptyDrops software. Colors were labeled by detection software (e), expression of mitochondrial transcripts (f), or levels of ribosomal RNA (g). h Cells’ number detected by both CellRanger and EmptyDrops (red, both), missed by both software (purple, neither), or rescued by EmptyDrops (blue). i Plotting ζ-scores of individual cells against their %mt. Colors label cells as in h. j Same as (i) except plotting the cell density in y-axis. k UMAP of cells that meet the cutoffs of CellRanger, EmptyDrops, and ZetaSuite (light blue) in comparison with those that meet the cutoffs of EmptyDrops and ZetaSuite (red) or the cutoff of only EmptyDrops (dark blue). Colors were labeled by detection software. i Percentage of T cells or monocytes identified by EmptyDrops and ZetaSuite (red) or only EmptyDrops (blue). m Percentages of stripped nuclei (left, characterized by both low %mt and ribosome expression) or broken cells (right, associated with high %mt but low ribosome expression due to selective leakage of cytoplasmic mRNAs from broken membrane) identified by EmptyDrops but discarded by ZetaSuite. n Ridgeline plot showing the distribution of nCount, nFeature, %mt, and ribosome expression for cells detected only by EmptyDrops, by both EmptyDrops and ZetaSuite, or by all three software, showing the ability of ZetaSuite to rescue high-quality cells missed by CellRanger while filter out damaged cells also rescued by EmptyDrops