Genome-wide single cell annotation of the human protein-coding genes

Abstract

An important quest for the life science community is to deliver a complete annotation of the human building-blocks of life, the genes and the proteins. Here, we report on a genome-wide effort to annotate all protein-coding genes based on single cell transcriptomics data representing all major tissues and organs in the human body, integrated with data from bulk transcriptomics and antibody-based tissue profiling. Altogether, 25 tissues have been analyzed with single cell transcriptomics resulting in genome-wide expression in 444 single cell types using a strategy involving pooling data from individual cells to obtain genome-wide expression profiles of individual cell type. We introduce a new genome-wide classification tool based on clustering of similar expression profiles across single cell types, which can be visualized using dimensional reduction maps (UMAP). The clustering classification is integrated with a new “tau” score classification for all protein-coding genes, resulting in a measure of single cell specificity across all cell types for all individual genes. The analysis has allowed us to annotate all human protein-coding genes with regards to function and spatial distribution across individual cell types across all major tissues and organs in the human body. A new version of the open access Human Protein Atlas (www.proteinatlas.org) has been launched to enable researchers to explore the new genome-wide annotation on an individual gene level.

Autocrine role of senescent cardiac fibroblasts-derived extracellular vesicles

“…do not differ between cardiomyocytes and cardiac fibroblasts. On the other hand, it seems likely that the expression level of Fyn kinase, an upstream suppressor of AMPK, in cardiac fibroblasts is higher than that in cardiomyocytes [16].…”

The human protein atlas—Integrated omics for single cell mapping of the human proteome

“…Classification of genes according to specificity of expression within the human body has been implemented previously in the HPA for bulk RNA‐seq data in order to highlight genes with interesting expression patterns across tissues (Ahlen et al, 2015 ). The same strategy was utilized to classify the specificity of gene expression across cell types (Karlsson et al, 2021 , 2022 ). The specificity describes to which degree the expression of a gene is elevated among one or a few cell types compared to all other cell types.…”

“…Elevated expression is divided into three different degrees of specificity, including (a) enriched expression , where a gene has at least a four‐fold higher expression in a cell type compared any other cell types, (b) group enriched , where a gene has at least a four‐fold higher expression in a group of 2–10 cell types compared to any other cell type, or (c) enhanced , where a gene has at least a four‐fold higher expression in 1–10 cell types compared to the mean expression of the other cell types. Most genes ( n = 15,317, 76%) were classified as elevated in one or more cell types (Karlsson et al, 2022 ). This is a marked increase compared to the bulk RNA‐seq data ( n = 10,986, 55%), which is expected since many of the analyzed cell types are found across many tissues.…”

“…Since cells of the same cell type have similar expression profiles compared to other types of cells, they cluster together in the UMAP analysis of scRNA‐seq data, forming the basis for identifying cell type clusters. In addition to analysis of single cell type expression across genes, UMAP was also employed in an opposite fashion to explore the relationship between all protein‐coding genes across tissues, cell types, immune cells, and cell lines, as a help to uncover previously unknown associations between genes (Karlsson et al, 2022 ). This new way of expression clustering constitutes a novel classification of all protein‐coding genes.…”

“…The processed data was added to version 20 of the HPA in the form of a Single Cell Type section (http://www.proteinatlas.org/humanproteome/single%2Bcell%2Btype). In version 21, 12 new tissues were added using the same data processing pipeline, increasing the total coverage to 25 tissues and PBMCs, giving rise to 444 individual cell type clusters (Karlsson et al, 2022). Cell type clusters within a tissue that were annotated with the same name, as well as cell types found in multiple tissues, such as immune cells, connective tissue cells, and endothelial cells, were merged.…”