Extended Data Fig. 7: Population analysis of 10x Genomics-based scRNA-seq data and label transfer from CITE-seq transcriptome to scATAC-seq cells.
From: Mouse models of neutropenia reveal progenitor-stage-specific defects
a, Identification of additional cell populations from wild-type scRNA-seq data (ICGS2 unsupervised analysis, male–female CITE-seq datasets). UMAP projections of scRNA-seq data, in which the cell barcodes within the outlined region (top) were analysed for additional heterogeneity through a second ICGS2 analysis (bottom). b, Heat maps of additional CITE-seq captures with cell assignments from cellHarmony. cellHarmony classifications were derived using the refined cluster annotation assignments from a and Extended Data Fig. 6c. Each panel displays a scRNA-seq heat map of MarkerFinder genes from the cellHarmony reference (top), in which each column represents a single cell and each row represents a single gene. Cell barcodes captured from an independent FACS sort of LY6GhighCD11bhigh GMP-P cells are indicated by a black bar (middle). Relative expression of median-normalized ADTs (right) are shown in the bottom heat maps. c, Assigned cell-population frequencies for Gfi1+/− and Gfi1R412X/− CITE-seq (modified GMP gate) cells data sets from cellHarmony. d, Heat maps of differential ADT expression compared to Gfi1+/− for the indicated markers (right). e, t-distributed stochastic neighbour embedding (t-SNE) plot of CITE-seq transcriptome for the Gfi1+/+ sample. f, t-SNE plot of scATAC-seq for the Gfi1+/+ sample. g, t-SNE plot of CITE-seq transcriptome for the Gfi1R412X/R412X sample. h, t-SNE plot of scATAC-seq for the Gfi1R412X/R412X sample. i, Comparison of Seurat-transferred CITE-seq labels to unsupervised scATAC-seq cell population prediction methods (SNAP-ATAC) for Gfi1+/+. Percentage of overlapping cells for all pairwise comparisons between SNAP-ATAC clusters to CITE-seq clusters derived from Seurat label transfer. j, Comparison of Seurat-transferred CITE-seq labels to unsupervised scATAC-seq cell population prediction methods (SNAP-ATAC) for Gfi1R412X/R412X. Percentage of overlapping cells for all pairwise comparisons between SNAP-ATAC clusters to CITE-seq clusters derived from Seurat label transfer. k, Cell-cluster combined scATAC-seq marker peaks associated with Gfi1+/+ CITE-seq annotated cell populations (Seurat 3 label transfer using cicero gene activity scores). Each row is a cluster (left) and each column is a locus within 50 kb of a CITE-seq marker gene for each cluster (bottom), in which the coloured bars represent the normalized read count coverage. l, Heat maps of transcription factor-motif-enrichment probabilities (−log10-transformed) in accessible regions of Gfi1+/+ (blue) or Gfi1R412X/ R412X (red) cell populations (indicated left). Each row represents a transcription factor motif from the Cisbp2 database (indicated right). Each dot in a and e–h represents a single cell barcode that is pseudo-coloured for its ICGS2 predicted cluster.
