What is the best way to quantile normalize expression matrix ?

Question

[jack]

I have an expression matrix, its rows belong to different genes and columns belong to samples.

I want to do quantile normalization. But for me seems two ways to normalize it. First: row wise; normalize the expression of every gene in all samples. Second: normalize columnwise: normalized expression of all genes in one sample and then go to other sample and ....

Which one makes more sense?

Answer 1

[Devon Ryan]

The point of quantile normalization is to make the signal distribution of the samples as close as possible, so you normalize the columns. Trying to normalize the rows would introduce a whole host of issues that could seriously muck with the results (e.g., since you look for differential expression across rows, how might normalizing each row affect that...).

[jack]

Good point Devon. I'm using normalize.quantiles(x,copy=TRUE) to do quantile normalization, but they didn't explain that, it's normalize by row or column. do you know useful function for quantile normalization?

[Devon Ryan]

The only other one that I know of off-hand (aside from tweaked versions, like cqn) is normalizeBetweenArrays() from the limma package. I wouldn't be surprised if there are others.

Answer 2

[ATCG]

Thank you very much for answering this question Devon, I have a follow-up question: Under what circumstances is it appropriate to do both log transformation and quantile normalization? Thank you in advance.