Cell Type Composition In A Tissue Based On Gene Marker Expression

Question

[k.nirmalraman]

I am not sure if the following would even make sense....

Tissues are composed of composite cell types, and often there are studies such as microarray/NGS where we perform a collective sampling of cells from these tissues. Information about the composition (say percentage of cell type) is not taken into consideration. In some case (such as brain/cancer), there are many studies where specific cell type specific marker genes are identified.

I was wondering if there are any methods, where we can use these cell type marker gene expression from RNA-Seq, to estimate the percentage composition of the individual cells in tissue (or say differential percentage)... Do you even think this method would make sense...?

If there are any references towards this line of thought, pls do share?

Thanks!

Answer 1

[Mikael Huss]

In fact, there are many tools for this, although most of them were developed for microarray data.

There are some recent ones like DeconRNASeq which is, obviously, meant for RNA-seq data.

One other recent example that is supposed to work for both platforms: http://www.biomedcentral.com/content/pdf/1471-2105-14-89.pdf

CellMix is a very nice package that collects a lot of published methods for cell-type deconvolution based on expression data. The manual: http://web.cbio.uct.ac.za/~renaud/CRAN/web/CellMix/vignettes/SampleAnalysis.pdf

This is just scratching the surface. Look at the references in the CellMix manual.

By the way, I suspect that normalizing RNA-seq data with voom() (in limma) gives you something that you can use with the methods that were developed for microarray data.