TMM normalization with DESeq2

Question

[marco.barr]

Hello everyone,

I wanted to know how to computationally use TMM normalization (typically used with edgeR) with DESeq2. I've read that it's possible to do so. In my analyses, I've always used tximport initially, DESeq (ddTxi), and then worked on the results of DEG. To use TMM normalization, what do I need to convert? Would an approach similar to this work:

N <- colSums(geneCount) (vector of library sizes) 
tmm.counts <- Ntmm/exp(mean(log(Ntmm))).

But in my case, since I don't use edgeR, what would geneCount be? Perhaps txi$counts?

Thank you very much for any advice and help.

Answer 1

[ATpoint]

Why so complicated? The normalizations of edgeR and DESeq2, while different in how they tackle the problem, perform almost identically for the end user. Just use DESeq2 and go along.

DESeq2 does not calculate the Trimmed Mean of the M-values (TMM). If you already have what in edgeR is called "norm.factor" and you want to convert it into a DESeq2 size factor then you can use this code, assuming your norm.factor is nf:

# libsize is the column sum of the raw counts
ls <- colSums(raw_counts)

# size factor combines libsize with normalization factor into a single value
sf <- nf * ls / 1000000

# stabilize size factors to have geometric mean of 1
sf <- sf / exp(mean(log(sf)))

sizeFactors(dds) <- sf

But again, why bother?

[marco.barr]

Thanks for your response, but What I'm missing is precisely the norm.factor. Do I need to execute this to obtain the norm.factor?

TMM <- calcNormFactors(txi$counts, method="TMM")

and then proceed with what you confirmed with what I was thinking?

[ATpoint]

Yes, or the newer version of the function called normLibSizes().