Normalization edgeR

[Schelarina]

Dear all,
I know there are thousands of posts regarding this topic.
HOW one can choose the best normalization method in edgeR??
I know that TMM is implemented in edgeR, but you can use other normalizations by entering lib.size and norm.factors in the edgeR DGE list object.
My problem is that when I compare the results I obtained with the TMM, total count and upper quartile normalization methods I obtained different results.
So what is the criteria for choosing the best normalization method?

[dpryan]

Never use "total count", it's not robust. You're going to validate your results on additional samples with qPCR or some other method anyway, so just choose genes appropriately when doing that to see which performed better on your dataset.