I am using a scRNA-seq data as a reference matrix to deconvolute cellular landscape in bulk RNA-seq data. To do that, I am following CIBERSORTX tutorial for cross-platform deconvolution here. Before running the cell fraction imputation job, there is a checkbox "disable quantile normalization".
In the tutorial, they say:
Quantile normalization: Uncheck this box to enable quantile normalization. Disabling quantile normalization is recommended for RNA-Seq data. As we are working with RNA-Seq data in this tutorial, we leave this box unchecked.
So, disabling is recommended as I am going to deconvolute RNA-seq data. But they say "we leave the box unchecked" which will enable quantile normalization. Could somebody help me understand this? Should I leave it checked or unchecked?
I don't get why this is not clear for you. Checking the box will disable the quantile normalization (set to FALSE). By default they leave the box unchecked (so, with the quantile normalization set to TRUE) to deconvolute microarray data, where the quantile normalization is suggested.
From the website
Quantile normalization is recommended for gene expression data generated using microarray, it is not recommended for RNA-Seq data. As we are working with RNA-Seq data, check this option to disable quantile normalization.
The data in both the tutorial and my analysis is RNA-seq, not microarray.
They say "Disabling quantile normalization is recommended for RNA-Seq data (set to FALSE is recommended). As we are working with RNA-Seq data in this tutorial, we leave this box unchecked (they set it to TRUE)."
It's just a typo in that specific part of the tutorial. This is likely a holdover from when this option was originally implemented and not checked by default.
Leave it checked.
thank you!
Log in to answer this question.
