filtering lowly expreed genes after or before normalization ?

[atasub]

Hi, I need to figure out which approach is more appropriate regarding filtering lowly expressed genes. According to tximport manual, it is recommended to follow following commands for EdgeR analysis:
library(edgeR)

cts <- txi$counts
normMat <- txi$length
normMat <- normMat/exp(rowMeans(log(normMat)))
library(edgeR)
o <- log(calcNormFactors(cts/normMat)) + log(colSums(cts/normMat))
y <- DGEList(cts)
y$offset <- t(t(log(normMat)) + o)

and to continue with y as a DGE object. In my analysis I filtered out the lowly expressed genes based on the cpm value (for instance, cpm value is greater than 1 in at least the number of small group of samples) using "keep.lib.sizes=FALSE" after doing above mentioned normalization.
I am now confused if my approach is appropriate and if I should do the normalization after filtering?

Thanks for your help.
Best,

[mikelove]

hi, please see this note:

#19