The target of interest was a gene isoform which is deposited in the extracellular matrix (ECM). As such, it is unclear from IHC data alone if tumor cells themselves can also produce the protein.
Off-the-shelf single-cell quantification algorithms cannot quantify isoform expression at the single cell level. Indeed, there are typically not enough reads in individual cells to be able to quantify expression at the complexity of gene isoforms.
A Novel Pseudo-Bulk Approach
A pseudo-bulk approach to quantification was able to pull together enough transcriptomic data across hundreds of cells to overcome some of the limitations of single-cell RNA-seq data.
We developed a novel algorithm and single-cell pipeline that first classified cells into cell types, and then aggregated across all reads mapping to the cells of a cell type in order to perform "pseudo-bulk" isoform-level quantification, making the assumption that junction usage across cells of a cell type are likely to be similar.
Although junction usage was not uniform across such groupings, the variation could be captured and a junction usage distribution could be calculated and compared across cell types.
Key Results
In a proof-of-concept analysis of single cell data, we were able to apply this pipeline to show that tumor cells were unlikely to be producers of the protein, as the transcriptional activity was very low in tumor cells compared to cancer-associated fibroblasts.
At the RNA level, the primary producers at the RNA level were demonstrably the CAFs, with sub-clusters also showing significantly different transcriptional activity.