This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Recommendation for packages for cell type deconvolution from bulk RNAseq data?

I was wondering, if you have some recommendation on which R package to use for doing cell-type deconvolution from bulk RNAseq data? I would really appreciate your suggestions!

[edit] I am looking for recommendation specifically for brain tissue RNAseq. Any package that you have used in that context and had good experience with...

Thank you very much in advance!

rna-seq cell type deconvolution
• 5,555 views • link
updated by xinyiy027 • xinyiy027 • written by m.sadman.sakib m.sadman.sakib

C: Deconvolution Methods on RNA-Seq Data (Mixed cell types) mentions multiple packages, with good background discussion on input etc.

Pubmed for others. N.B. most newer methods use single cell as a 'guide' to deconvolution of bulk samples.

I've had more success with using GSVA with specific genesets/signatures though, if you can find such signatures.

Usually a good idea to give your specific use-case, e.g. tumour or immune etc so people can be more specific in answering.

• 0 views • link
written by bruce.moran bruce.moran

Thanks a lot! I am looking for recommendation specifically for brain tissue RNAseq.

I will check more extensively then.

• 0 views • link
written by m.sadman.sakib m.sadman.sakib

I have also used GSVA and found it to be really easy and good. Below paper also performed a comparison where GSVA was recommended, you can read for more details.

One of the way one colleague of mine has done is to combine GSVA enrichment score with gene correlation score. For correlation, you can combine all the genes in the gene sets you are going to use for GSVA (my approach), or focus on some set of highly variable genes in target samples. The results were comparable. I am also working on brain tissue, but haven't heard of any thing specific for Brain data. https://f1000research.com/articles/8-296

• 0 views • link
written by piyushjo piyushjo

1 answer

Hi! Perhaps ADAPTS can help. It helps deconvolve bulk samples by generates cell-type specific matrices. The deconvolution correlation can reach above 0.8 for most of the datasets. For more details you can check it out at https://cran.r-project.org/web/packages/ADAPTS/

• 0 views • link
written by xinyiy027 • xinyiy027

Log in to answer this question.

More posts like this