RNA-seq normalization: How to use TMM and rpkm() in EdgeR??

[alisonewr]

Hi,

I have some RNA-seq samples that I want to normalize and then output RPKM expression, but I am unsure how to do this.

This is my pipeline so far:

1. Normalise raw read counts with TMM in edgeR

expr <- DGEList(counts=data, group=conditions)
expr <- calcNormFactors(expr)

output:

$samples
group lib.size norm.factors
Sample1 F 19770521 1.0462660
Sample2 F 17970679 0.8794805
Sample3 F 19184265 1.0573665

QUESTION: How do I get normalized raw read counts from this? Do I multiply the read counts by the norm.factors?

QUESTION: Ultimately, I want to end up with RPKM values for each gene in each sample. I know I can use the rpkm() function below in edgeR

expr_norm <- rpkm(expr, log=FALSE,gene.length=vector)

but is expr the output from calcNormFactors or something else?

Thanks for your help!

A

[Gordon Smyth]

Yes, you simply use

rpkm(expr, log=FALSE,gene.length=vector)

where expr is the output from calcNormFactors(). edgeR is an object orientated package where the DGEList data object assemblies all the required information.

There is no need for you to compute normalized raw counts (in fact I would argue there isn't any such thing). edgeR uses the normalization factors automatically.

[Gonza]

Hello, does it really matter which normalization you use with edgeR? I guess I do not understand really well the differences.
I am using TMM now but I have no 'solid' argument why i chose that instead of upperquartile or RLE.

Any thoughts?
thanks

normalization:

#y <- calcNormFactors(y, method="TMM")
#y <- calcNormFactors(y, method="upperquartile")
#y <- calcNormFactors(y, method="RLE")
y$samples