Extended Data Fig. 2: Validation of snATAC-seq topic model with gene activity and label transfer and Molecular cartography of the mouse liver.
a. Scaled gene expression (top) and scaled gene activity inferred from the snATAC-seq layer (bottom). b. Correlation between gene activity and gene expression versus a random control (shuffled gene activity) across 1,141 marker genes (log2[FC]> 1.5 and adjusted P < 0.05 in at least one cell type). The top/lower hinge represents the upper/lower quartile and whiskers extend from the hinge to the largest/smallest value no further than 1.5 × interquartile range from the hinge, respectively. The median is used as the centre. A one-sided rank-sum Wilcoxon test was performed to assess if the correlation based on gene activity was greater than random (P: < 10−16). Gene activity values are derived from 4 independent biological replicates. c. snATAC-seq UMAP (22,600 cells) coloured by the label transfer cell type annotation (from snRNA-seq cells to snATAC-seq cells). For the UMAP, cells from 4 biological replicates were combined. d. snATAC-seq UMAP (12,898 cells) coloured by the label given to the cell based on the snRNA-seq clustering. e. Pseudobulk accessibility profiles on representative Differentially Accessible Regions per cell type. The bar plot indicates the number of peaks called by MACS in each pseudobulk. f. Number of selected genes per cell type within the Molecular Cartography gene panel (100 genes). g. snRNA-seq (29,798 cells) UMAP based on the selected 100 genes. h. Cell segmentation using the DAPI signal on the sample with QuPath. Three independent experiments were performed, with similar results. i. UMAP (15,522 spots/cells) of the segmented nuclei based on the number of transcripts measured per gene per spot coloured by their assigned cell type. Within the rectangle, UMAP (15,522 spots/cells) of the segmented nuclei based on the number of transcripts measured per spot coloured by their sample of origin. For the UMAP, cells/spots from 3 biological replicates were combined. j. Molecular Cartography maps (3 replicates, 15,522 spots) coloured by aggregated gene expression using RGB encoding. k. Molecular Cartography maps (3 replicates, 15,522 spots) coloured by assigned cell type. l. Heat map showing the scaled gene expression across cells (grouped by cell type) of selected genes. BEC: biliary epithelial cells, cDC: conventional dendritic cell, HSC: hepatic stellate cells, MSC: mesothelial cells, PC: pericentral, pDC: plasmacytoid dendritic cell, PP: periportal, VEC: vascular endothelial cells. Source numerical data are available in source data.
