Brain cell type deconvolution

Question

[rodd]

Hi all!

I am interested in understanding the % of main cell types composing my bulk RNA-seq data from brain (e.g., astrocytes, microglia, neurons, oligodendrocytes). However, from what I gathered, this appears to be an extremely complicated task. I came across CiberSortX, but I couldn't find a good tutorial for it. There are no accessible instructions about what the input files should look like, how to treat the RNA-seq data, etc. And even the idea of selecting and processing the appropriate dataset (single-cell RNA-seq, or RNA-seq performed on sorted cells) seems a bit daunting.

Has any of you conducted this type of analysis for bulk RNA-seq tissue from brain? Do you know any tutorial or tool that could help me achieve this more easily, or it's not gonna be that easy either way?

Thank you!

Answer 1

[jv]

There are several detailed tutorials available for CiberSortX, but you have to register for the software in order to access them.

For additional info on prepping your bulk RNA-seq data you can also review the Methods section of the CibserSortX paper https://doi.org/10.1038/s41587-019-0114-2.

Answer 2

[LChart]

The PsychEncode consortium performed cell type deconvolution using non-negative least squares (https://www.science.org/doi/10.1126/science.aat8464) and they provide the cell-by-gene expression targets as a resource (http://resource.psychencode.org/). DeconRNAseq implements this for you; though it underperforms its weighted version (MuSiC) which performs competitively with CiberSort (source: https://www.nature.com/articles/s41467-022-28655-4). Check the supplementary material of this paper for their approach to running these alternative (open-source) deconvolution methods.