Question: cell type deconvolution with Cibersort
0
gravatar for chuxj1128
5 months ago by
chuxj11280
chuxj11280 wrote:

Dear all:

I have 2 questions using Cibersort deconvoluting cell types. I have single cell RNAseq data from 5 healthy and 5 cases with huge individual variation in some certain cell types.

1) I wonder if I can remove those un-conservative cell types to predict relative proportion with only a subset of cell types presented in single cell RNAseq data. e.g. Cells grouped in to 7 clusters under a certain resolution in single cell data. Then I calculated the relative proportion of each individuals by cell counts (of a certain cell type) dividing all cell counts. Next, I got 3 cell types which are relative conservative across individuals. I wonder if I can only use those 3 cell types to do deconvolution in bulk RNAseq.

2) I observed cell proportion difference between healthy control and cases. Also some genes were differentially expressed. While the strategy of several deconvolution methods is using the same single cell RNAseq profiling as reference to deconvolute cell types in bulk RNAseq from cases and controls. Is that fair ? Cos the prediction models could be different for cases and controls.

Thank you and best wishes!

rna-seq • 443 views
ADD COMMENTlink modified 5 months ago • written 5 months ago by chuxj11280

For your information, you could have edited your previous post to make it a "Question". Because you already posted it as a question I have deleted the previous post.

ADD REPLYlink written 5 months ago by WouterDeCoster41k

I did some small test, and found as for CIBERSORT, it's not able to use a subset of cell types. I wonder if someone know that which method evaluate cell proportion only based on marker genes of its own cell type, but not refer markers genes of other cell types.

Best!

ADD REPLYlink written 5 months ago by chuxj11280

Do you have an appropriate and well-validated ground truth set (=signature matrix)?

ADD REPLYlink modified 5 months ago • written 5 months ago by ATpoint23k

Thanks for your reply, i followed the pipeline of bseq-sc which trains signature matrix from single cell RNAseq. While through looking into my single cell RNAseq data, I found several cell types were not stable (having huge individual variation). So I don't really trust my signature matrix if I included all cell types presented in single cell RNAseq, and that's the reason, I wonder if I could only consider a subset of cell types.

ADD REPLYlink written 5 months ago by chuxj11280
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.3.0
Traffic: 1956 users visited in the last hour