Deconvolution of RNAseq (bulk) from skin samples

Question

[BlueSky]

Hello! I have RNAseq data (bulk) from a skin disease, and I would like to do deconvolution. I have understood that for this I need a reference matrix. I found one for skin which is named "Derm22" but the problem here is that this signature matrix is made from microarray data, is that a problem when my gene expression matrix is from RNAseq?

If it is a problem, do anyone know of a reference matrix which contain cells typical in the skin which is from RNAseq?

Answer 1

[Wangering Hedgie]

Hello BlueSky !

I hope my answer is not too late.

From my knowledge and from what I read in articles, using different techniques between reference and bulk is to avoid because of biases between techniques.

However, in Scaden deconvolution method paper, the authors trained the Scaden Deep Learning model with simulated single cell RNA-seq PBMC data and performed deconvolution on PBMC microarray data (See Supplementary Table S13 and Supplementary Figure S8).

Here is the Scaden paper : Deep learning–based cell composition analysis from tissue expression profiles, Menden et al., 2020