I am trying to perform a CIBERSORTx analysis on bulk data using single cell data as a reference. The single cell data has almost 59k cells, therefore the reference file for creating the signature matrix is coming out to be quiet large, more than the CIBERSORTx web application can handle. Is pseudo bulking the reference file on the basis of samples a suitable option in this case? Or should I use the original matrix and perform deconvolution using the CIBERSORTx image in docker? Also after deconvolution, the result generated is in the form of fractions or proportions. What will be a suitable statistical test to use on these fractions to check the difference in cell types in different conditions. Thankyou!
•
written
by
yadavsoumya647 •
1 answer
My team and I ran into a similar issue. Easiest solution is just downsampling the matrix while accounting for compositional differences in cell type populations. Also if you want to compare cell type distributions across conditions you can probably use a statistical test like Wilcoxon Rank sum to compare them
Log in to answer this question.
-
CIBERSORTx Fractions/HiRes: which mixture/signature-matrix combination is actually used internally …
written by Francesco •
-
Statistical test for proportion values.
written by yadavsoumya647 •
-
Confidence in CIBERSORT (LM22) Deconvolution Output for Rare Cell Types — Best Practices?
written by Pumla •
-
Error in deconvolution using CIBERSORTx
written by yadavsoumya647 •
-
How to create a signature matrix, using scRNA-seq data, for downstream CIBERSORTx deconvolution of …
written by Fossil •
-
Creating a signature matrix using CibersortX
written by Aspire
-
Creating a Single-Cell Signature Matrix for Bulk RNA-seq Deconvolution
written by Lloyd98 •
-
How to carry out cell deconvolution GIVEN known cell proportions already?
written by Rashid •
-
CIBERSORTx error (error in rep(2, size * length(cells) -1)) : invalid 'times' argument
written by Mr. George •
-
[SOLVED] How to build a table with gene expression per cell type with Seurat ?
written by Evan •
am also doing similler type of experiment, i got similar issue. Easiest solutionfor this time try with less no of sample is just down sampling the matrix while accounting for compositional differences in cell type populations. Also if you want to compare cell type distributions across conditions you can probably use a statistical test like Wilcoxon Rank sum to compare them Try with docker if it working ? because now Cibersortx guideline changes, and if you want to check further cell states and differential expressed genes, We can discuss.