Extended Data Fig. 1: Construction of the HCL.
From: Construction of a human cell landscape at single-cell level
a, Comparison of t-SNE maps for lung data sets from Tabula Muris (10X Genomics), down-sampled Tabula Muris data (10X Genomics), and MCA data (microwell-seq). Note that downsampling in sequencing depth does not affect cell-type clusters in the Tabula Muris data. Notably, lung data from MCA (sequenced at lower depth) detects more cell-type clusters, including important lung epithelial cells such as AT1 cells, club cells and biopotent progenitors. b, Numbers of cells processed by 31 December 2019 at the HCL for each tissue type. c, Venn diagrams of gene numbers detected in bulk RNA sequencing and microwell-seq (genes with fewer than three counts are excluded). Scatter plots on the right show high correlations (more than 0.8) of average gene expression between bulk RNA sequencing and microwell-seq. We analysed 17,058 genes for kidney and 16,910 genes for lung. d, The percentage of cell types recovered in sub-samples of adult lung, kidney and adrenal gland single-cell data. The major cell-type numbers in representative tissues are near plateau at around 8,000 cells; we collected more than 10,000 cells per tissue on average. e, A Seurat analysis of donor batch effect from four fetal kidney samples (n = 22,439 cells) and three adult kidney samples (n = 22,692 cells). The mixing of different donor single cells in each cell-type cluster suggests a relatively low batch effect in the data. The cluster contribution bar charts on the right suggest that one of the fetal kidney donors lacks C19 and one of the adult kidney donors lacks C22.
