EdgeR Normalization Question

R_user · 03-31-2012, 03:07 AM

I am looking to do TMM normalization in an experiment where you have to do pairwise comparisons between 20+ samples in replicates of three (A1, A2, A3 vs B1, B2, B3; A1, A2, A3 vs C1, C2, C3; B1, B2, B3 vs C1,C2, C3 and so on).
I was thinking that I should normalize all the samples together and then do pairwise comparisons.
I was wondering how I can get EdgeR to give me the TMM normalized values in a csv file.
Thanks!

Gordon Smyth · 05-07-2012, 11:14 PM

There's actually no such thing as TMM normalized values. Normalization in edgeR is something that affects the fitted models rather than the data values. As the edgeR User's Guide says:

"There are methods implemented in edgeR to normalize the counts for compositional bias in sequenced libraries and for differences between libraries in sequencing depth. These adjustments are offsets in the models used for testing DE and do not transform the counts in any way."

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
RNA seq data normalization question	slny	Bioinformatics	35	10-19-2016 06:32 AM
edgeR	ECHo	Bioinformatics	13	04-18-2013 02:01 PM
edgeR Normalization question	polsum	Bioinformatics	1	05-07-2012 01:21 AM
edgeR	Puva	Bioinformatics	2	05-19-2011 10:04 AM
Normalization: DESeq vs. EdgeR with method="RLE"	aquila	RNA Sequencing	3	04-26-2011 10:00 AM

03-31-2012, 03:07 AM	#1
R_user Junior Member Location: Springfield Join Date: Mar 2012 Posts: 1	EdgeR Normalization Question I am looking to do TMM normalization in an experiment where you have to do pairwise comparisons between 20+ samples in replicates of three (A1, A2, A3 vs B1, B2, B3; A1, A2, A3 vs C1, C2, C3; B1, B2, B3 vs C1,C2, C3 and so on). I was thinking that I should normalize all the samples together and then do pairwise comparisons. I was wondering how I can get EdgeR to give me the TMM normalized values in a csv file. Thanks!

05-07-2012, 11:14 PM	#2
Gordon Smyth Member Location: Melbourne, Australia Join Date: Apr 2011 Posts: 91	There's actually no such thing as TMM normalized values. Normalization in edgeR is something that affects the fitted models rather than the data values. As the edgeR User's Guide says: "There are methods implemented in edgeR to normalize the counts for compositional bias in sequenced libraries and for differences between libraries in sequencing depth. These adjustments are offsets in the models used for testing DE and do not transform the counts in any way."