Options for cell type deconvolution from whole blood?

Question

[penny.lane]

Hello,

My colleagues and I have human, whole blood RNA-Seq data from which we would like to estimate the proportions of various immune cell types within each sample and then see how they are changing between samples. We do not, however, have our own matched single cell RNA-Seq data set to help with the deconvolution. What methods would people recommend we consider? We're particularly interested in any methods that have already been validated to work well on whole blood samples.

Thanks!

Answer 1

[E.Earley]

try their new CIBERSORTx

https://doi.org/10.1038/s41587-019-0114-2

https://cibersortx.stanford.edu/

Also check out SCDC

https://academic.oup.com/bib/advance-article/doi/10.1093/bib/bbz166/5699815

https://meichendong.github.io/SCDC/

[Shicheng Guo]

Great. Thanks for these useful information

Answer 2

[trausch]

Cibersort could be an option.

[penny.lane]

Yeah, this is one that caught my attention, but according to their website the LM22 signature was designed & validated on microarray data and they warn against applying it to RNA-Seq data. They say they are in the process of deriving and validating a signature for RNA-Seq data, but no release dates are given. Any other ideas?

[Shicheng Guo]

I think array data will be more stable. rna-seq data should have high depth to receive same level accuracy.