Differential expression with CQN normalized RNA-seq data

Question

[swkim]

Hello,

I am pretty much new to RNA-seq and wanted to ask a really basic question about differential expression analysis.

I have a set of CQN (conditional quantile normalization) normalized RNA-seq RPKM data (the are all matched-pairs like tumor vs. normal).

When we compare expression of a gene with control, we usually use a log-based fold change like log2(T/N). However, CQN values are already log2 based. So calculating a log-based fold change seems like taking logarithm twice.

So, in this case, do we still take a log fold-change? or do different things like just using differences of those values?

Thank you all~

Answer 1

[Devon Ryan]

If the output of CQN is on the log2 scale, then just subtract the values (log2(A/B)==log2(A)-log2(B)).

Having said that, I've only ever used CQN with edgeR/DESeq2/limma, which will produce the log2(foldchange) for you. I wouldn't recommend inventing your own analysis without good reason.

[swkim]

Thank you! I agree with you. Someone just gave me the log fold change result from CQN log2 scale data, so I wanted to double-check the result is inappropriately processed.

[grzegorz.maciag]

Hi. You mention here that you use CQN together with limma. Could you maybe share how you do it, I don't think it's a default functionality, as in this post: https://support.bioconductor.org/p/56238/ Do you know any kind of resource that would help me get the two tools working together?

I know I'm resurrecting a very old post, but it's one of the only ones I found googling for using CQN together with limma.