Fig. 3

Cell type sorting and progressive overloading to assess overloading and deconvolution. A Overview of the workflow for generating the progressively overloaded dataset. PBMCs purified from a Ficoll gradient were sorted into three populations using FACS. Each population was split and stained with a hashing antibody. All samples were then pooled, and were loaded at increasing cell count into the wells of a 10x 3′ scRNA-seq chip. B Bar plot showing the cell counts from each population (x-axis) in each well (y-axis). C UMAP plot of all singlets (n = 105,395 cells) colored based on which well the mixed samples were loaded into. D Heatmap colored based on the fraction of cells from each hashed population assigned to each cell type using Seurat v4.0 label transfer methods. E UMAP plot, as in C, colored based on the sorted and hashed replicates. F UMAP plot, as in C, colored based on cell type assignments from Seurat v4.0 label transfer (legend to the right). G Dot plot showing the expression of well-known, population-specific marker genes for each replicate. Size corresponds to the fraction of cells in each group with > 0 expression. Color corresponds to the log-transformed median of all non-zero, normalized values in each group. H Overlay of the marker gene expression values per cell on the UMAP plot used in panel C. Color corresponds to normalized expression values. Color scales are independent per panel, with ranges indicated by the legend at the top-right of each plot