TMM - Normalized FPKM values for RNASeq data ?

Question

[gangireddy]

Hi People,

I want to generate TMM- Normalized fpkm values RNASeq data 5 time point data for male and female. I am very much confused how to generate these values.

1) Is it applying TMM to fpkm or apply TMM to raw counts and log transform the values to fpkm ?

2) Is it acceptable to use TMM- Normalized fpkm or should people use TMM- normalized raw count directly ?

and finally

3) TMM- normalized fpkm does it mean it normalized twice and is it necessary.

Can anyone provide the step wise commands in R to generate TMM- normalized fpkm matrix

Thanks in advance

[Benn]

What would you like to do with these TMM-normalized fpkm values? There are many discussions on this site over why not to use fpkm or rpkm values...

If you know what you are doing (and are sure that you need these values), you can use edgeR in R to do it.

[WouterDeCoster]

Most often the conclusion is not to use fpkm and just use raw counts in a statistical framework like DESeq2 and edgeR.

[gangireddy]

we want to compare log2 male/female fpkm autosomal vs sex linked genes density distribution. for this we are comparing two different samples that differ in sample size (read counts).

is it preferred to go with raw counts or fpkm ?

[WouterDeCoster]

If you are going to use edgeR or DESeq2 then you need raw counts, not normalized.

Answer 1

[EVR]

hi,

Try using egdeR or Deseq2. take a look at trinity's pipeline for normalizatio which make use of edgeR's TMM normalization methid and outputs the TMM values.

[lay_0]

I have the same question as gangireddy, but how can you enter FPKMs in edgeR?, just as you would do with raw counts?, never mind the non-integers?

[Benn]

Use raw read counts!

Answer 2

[Gordon Smyth]

Assuming gene length is set as a column in your DGEList object, then

y <- calcNormFactors(y)
FPKM <- rpkm(y)