identify cell of origin from bulk RNAseq data

Question

[sigalottil]

Hi,

I am trying to figure out what could be the cell of origin (a "normal" cell type most similar), for a poorly characterized cancer I am working on, starting from bulk RNAseq data.

I have used ssGSEA to generate scores for MSigDB C8 cell type signature gene sets on my samples, with the idea that cell types with the highest scores should be those more similar to my samples. Unfortunately, no clear enrichment of a specific cell type was evident at the top scores, and apparently similar cell types appear both at the high and low score ends. First question, is my assumption correct (e.g. highest ssGSEA scores should represent more similar cell types)? Second, would you suggest a different strategy?

Thanks for the help

[dsull]

To add to the other answer, finding a "cell of origin" is not easy for cancer cells because they generally dedifferentiate into another cell identity (e.g. resemble embryonic stem cells as I demonstrated in one of my papers). In single-cell sequencing, you might be able to find certain cells that at least partly resemble the original cell type. You might try "cancer cell" signatures to see if your cancer resembles HCC, RCC, AML, etc. to find out which cancer type your sample is most similar to.

[sigalottil]

Absolutely agree. The idea, in this moment, is to try and find a cell type that resembles most the tumor type we are working on. Shure we could not tell that that would be the cell of origin, but at least it would be a staring point. Thanks

Answer 1

[jared.andrews07]

Bulk RNA-seq is likely not the best tool for the job, assuming the tumor/cancer is heterogeneous at all (which most are).

You could consider deconvolution via CIBERSORTx, which will try to infer cell fractions in a given population given reference profiles for each cell type. It will require providing a decent reference dataset, but is otherwise fairly low effort.

[sigalottil]

Thank you. I will give it a try.