Tissue Cell Type – Methods summary

What is the Tissue Cell Type section?

This section provides a cell-type expression specificity prediction for each protein-coding gene, within a given tissue type.

Key Publications:
Norreen-Thorsen M et al. (2022), "A human adipose tissue cell-type transcriptome atlas." Cell Rep 40, (111046)
Öling S et al. (2023), "A human stomach cell type transcriptome atlas." bioRXiv, pre-print
Dusart P et al. (2023), "A tissue centric atlas of cell type transcriptome enrichment signatures." bioRXiv, pre-print

How was the analysis performed?

The analysis was performed on data from RNAseq of unfractionated tissue samples, which contained a mixed cell population (for sample details, see GTEx data). Samples from a single tissue type were analyzed together. A panel of 3 "virtual reference transcripts" with highly cell type specific expression were selected to represent each of the main constituent cell types within each tissue (Figure 1). Reference transcripts were selected based on existing data, including in-house generated immunohistochemically stained tissue sections and published single cell RNAseq data.

Figure 1: Overview of the analysis concept, using visceral adipose tissue as an illustrative example.

Across the respective sample sets the reference transcripts within each cell type panel correlated highly with each other, but not with those in the other panels (Figure 2). An integrative co-expression analysis was performed to determine the expression profile of each gene; genes highly correlated with all transcripts in only one reference panel were classified as enriched in that cell type.

Figure 2: Example reference transcript panels and corresponding correlations, using visceral adipose tissue as an illustrative example.

Benefits of the analysis method

• Artifact generating sample pre-processing is avoided, as cells are effectively analyzed in situ.

• Large numbers of biological replicates are analyzed concurrently.

• Fragile, apoptosis prone or difficult to separate cells can be analyzed.

• Minority cell types do not require prior enrichment processing.

What do the plots show?

For each gene, there is a plot showing the enrichment prediction score for each cell type profiled in the tissue (examples are shown in Figure 3). This score is the mean correlation between the gene and the 3 reference transcripts selected to represent each cell type profiled within the tissue. An enlarged symbol indicates classification of the gene as cell type enriched in the corresponding cell type (see below for details).

Figure 3: Enrichment plots for the gene Glycogenin 2 (GYG2) in visceral adipose tissue, subcutaneous adipose tissue and breast show predicted specificity in adipocytes in all 3 tissue types (indicated by enlarged symbol).

How do the classifications work?

Identified cell type enriched genes are subdivided into 3 specificity categories, based on the difference between the enrichment score in the corresponding cell type, compared to the other cell types profiled in the tissue:

  • Predicted specificity: ´Very high´ - Differential score vs. other profiled cell types within the tissue >0.35 (e.g., GYG2, below)
  • Predicted specificity: ´High´ - Differential score vs. other profiled cell types within the tissue >0.25 (e.g., IDH1, below)
  • Predicted specificity: ´Moderate´ - Differential score vs. other profiled cell types within the tissue >0.15 (e.g., CPM, below)

Figure 4: Example of specificity classifications for genes predicted to be adipocyte enriched in visceral adipose tissue.

Where is the data from?

All data is sourced from the Genotype-Tissue Expression (GTEx) project portal.