Start analysis from DESeq2-normalized read counts?

friducha · 07-03-2015, 04:48 PM

Hi all,

I know that DESeq2 relies on using raw read countsd as input. However, I have the following question:

What if the read counts were already normalized using DESeq2? Is there a way to start the analysis after this step?

A colleague did an initial DESeq2 run on our data, and supplied a table with the normalized read counts from DESEq. I do not have the raw counts, and was wondering if I could start from here and skip the normalization method that DESeq perfoms.

Michael Love · 07-04-2015, 06:40 AM

For DESeq2, you shouldn't use any kind of values where the library size has been divided out. The statistical methods use the properties of the sampling distribution of counts, as well as estimation of additional variation (dispersion), for inference.

The input matrix should be: the number of reads/fragments which can be uniquely assigned to a gene (aligning to any of the exons of a gene).

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
DESeq and non-raw counts (but also not rpkm normalized)	wingless	Bioinformatics	0	06-26-2015 12:11 PM
Getting normalized counts	IsBeth	Bioinformatics	2	05-31-2014 10:20 AM
edgeR mean normalized reads counts for each gene	antoza	Introductions	2	02-11-2014 09:22 AM
converting bam files to non-normalized read counts	lpn	Bioinformatics	4	10-09-2012 08:52 PM
DESeq: Read counts vs. BP counts	burkard	Bioinformatics	0	08-06-2010 12:52 AM

07-03-2015, 04:48 PM	#1
friducha Junior Member Location: UK Join Date: Jan 2015 Posts: 9	Start analysis from DESeq2-normalized read counts? Hi all, I know that DESeq2 relies on using raw read countsd as input. However, I have the following question: What if the read counts were already normalized using DESeq2? Is there a way to start the analysis after this step? A colleague did an initial DESeq2 run on our data, and supplied a table with the normalized read counts from DESEq. I do not have the raw counts, and was wondering if I could start from here and skip the normalization method that DESeq perfoms.

07-04-2015, 06:40 AM	#2
Michael Love Senior Member Location: Boston Join Date: Jul 2013 Posts: 333	For DESeq2, you shouldn't use any kind of values where the library size has been divided out. The statistical methods use the properties of the sampling distribution of counts, as well as estimation of additional variation (dispersion), for inference. The input matrix should be: the number of reads/fragments which can be uniquely assigned to a gene (aligning to any of the exons of a gene).