LogoLogo Secondary

A Human Cell Atlas Of Gene Expression During Development

The emergence and differentiation of cell types during human development is of fundamental interest. We applied an assay for single cell profiling of gene expression based on three-level combinatorial indexing (sci-RNA-seq3) to 121 fetal tissues representing 15 organs, altogether profiling transcription in 4-5 million single cells. From these data, we identify and characterize diverse human cell types and annotate them with respect to marker genes, expression and regulatory modules. We focus our initial analyses of these data on cell types spanning multiple organ systems, e.g. epithelial, endothelial and blood cells, as well as on relating these data to a mouse atlas of organogenesis. Intriguing observations include organ-specific endothelial specialization, potentially novel sites of fetal erythropoiesis, and potentially novel cell types or localizations of known cell types. Together with the accompanying human cell atlas of chromatin accessibility during development, these data comprise a rich resource for the exploration of human biology.




Sampled Data

To compare cell types across organs, we randomly sampled 5,000 cells per cell type per organ (or in cases where less than 5,000 cells of a given cell type were represented in a given organ, all cells were taken). The data comprise gene count information including a total of 377,456 cells.
Rows:Correspond to genes
Columns:Correspond to cells
Values:Corresponds to UMI counts


Annotation files for cell by gene count matrices
Sample:Cell id of each single cell with the reverse transcription and ligation barcode attached
Exon reads:The number of UMIs strand specifically mapping to exons region
Intron reads:The number of UMIs strand specifically mapping to gene introns
All reads:The number of UMIs detected per cell
Organ:The organ from which the cell is extracted
Fetus id:The ID of fetus from which the cell is extracted
Assay:Shows whether the sciRNA-seq3 is done from extracted nuclei or cells
Sex:The sex of source fetus
Development day:The gestational ages of the source fetus
Main cluster name:The main cell type name
Main cluster UMAP 1 and UMAP 2:Cell coordinates after UMAP dimension reduction during main clustering analysis
Organ cell lineage:The combined organ and main cell type information
Experiment batch:The sciRNA-seq3 experiment batch information.

Differentially Expressed Gene List

We performed DE gene analysis to identify differentially expressed genes across main cell types within each organ. We the idenfied genes that are differentially expressed across 77 main cell types across organs.
max cluster:Group id with the highest expression (“max.expr”)
second cluster:Group id with the second highest expression (“second.expr”)
fold change:Fold change between the max expression and second max expression
qval:False detection rate (one-sided likelihood ratio test with multiple comparisons adjusted) for the gene differential expression test across different cell groups.
organ:Organ in which DE gene analysis was performed

Garnett Cell Type Classifiers

Garnett faciliates automated cell type classification from single-cell expression data. Garnett works by taking single-cell data, along with a cell type definition (marker) file, and training a regression-based classifier. Once a classifier is trained for a tissue/sample type, it can be applied to classify future datasets from similar tissues.