Changing FDR in DESeq2

rajarapupriya · 03-24-2014, 01:06 PM

Could anyone please tell me how to change the FDR from default (0.1) to 0.05 in DESeq2.

Thanks
Priya

Jeremy · 03-26-2014, 08:38 PM

at which step?

rajarapupriya · 03-27-2014, 08:36 AM

Thanks Jeremy for replying to my question. I would like to change FDR before/while calculating the differential gene expression. I'm using DESeq2 1.2.10 Vignette. The entire differential expression steps are distilled to one step DESeq(). That is where I'm having problem.

Priya

Michael Love · 03-27-2014, 01:43 PM

You can choose whatever adjusted p-value cutoff you like after DESeq() and results(). DESeq() does not perform multiple test correction.

dds <- DESeq(dds)
res <- results(dds)
resSig <- res[which(res$padj < 0.05),]

The independent filtering in results() has an argument 'alpha', which is used to optimize a cutoff on mean normalized count, to maximize the number of genes with padj < alpha.

So you can do:

res <- results(dds, alpha=.05)
resSig <- res[which(res$padj < 0.05),]

Lots of useful information in ?DESeq and ?results.

rajarapupriya · 03-27-2014, 01:59 PM

Thanks, Michael !! Think I missed it somewhere.

Priya

Jeremy · 03-28-2014, 12:27 AM

The object res has the results for all of the genes and their FDR adjusted p-values. I always just merge res with the corrected count data and then write it all out to a tab delimited table then you have all the information that you can put in as a supplementary file.

rajarapupriya · 03-28-2014, 08:21 AM

Thank you, Jeremy !!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
edgeR: How important is the FDR value?	polsum	Bioinformatics	23	02-07-2014 09:58 AM
FDR and underpowered p-values	rskr	Bioinformatics	0	11-11-2013 07:09 AM
R mateys! Which is Best FDR ?	Richard Finney	Bioinformatics	9	11-05-2013 07:56 AM
Changing RGSM without changing any RGID, RGPL etc	akhattri	Bioinformatics	2	10-25-2013 11:26 PM
Nice p-value, bad FDR	Jcl	Epigenetics	5	08-30-2013 07:41 AM

03-24-2014, 01:06 PM	#1
rajarapupriya Member Location: OH, USA Join Date: Oct 2013 Posts: 18	Changing FDR in DESeq2 Could anyone please tell me how to change the FDR from default (0.1) to 0.05 in DESeq2. Thanks Priya

03-26-2014, 08:38 PM	#2
Jeremy Senior Member Location: Pathum Thani, Thailand Join Date: Nov 2009 Posts: 190	at which step?

03-27-2014, 08:36 AM	#3
rajarapupriya Member Location: OH, USA Join Date: Oct 2013 Posts: 18	Thanks Jeremy for replying to my question. I would like to change FDR before/while calculating the differential gene expression. I'm using DESeq2 1.2.10 Vignette. The entire differential expression steps are distilled to one step DESeq(). That is where I'm having problem. Priya

03-27-2014, 01:43 PM	#4
Michael Love Senior Member Location: Boston Join Date: Jul 2013 Posts: 333	You can choose whatever adjusted p-value cutoff you like after DESeq() and results(). DESeq() does not perform multiple test correction. dds <- DESeq(dds) res <- results(dds) resSig <- res[which(res$padj < 0.05),] The independent filtering in results() has an argument 'alpha', which is used to optimize a cutoff on mean normalized count, to maximize the number of genes with padj < alpha. So you can do: res <- results(dds, alpha=.05) resSig <- res[which(res$padj < 0.05),] Lots of useful information in ?DESeq and ?results.

03-27-2014, 01:59 PM	#5
rajarapupriya Member Location: OH, USA Join Date: Oct 2013 Posts: 18	Thanks, Michael !! Think I missed it somewhere. Priya

03-28-2014, 12:27 AM	#6
Jeremy Senior Member Location: Pathum Thani, Thailand Join Date: Nov 2009 Posts: 190	The object res has the results for all of the genes and their FDR adjusted p-values. I always just merge res with the corrected count data and then write it all out to a tab delimited table then you have all the information that you can put in as a supplementary file.

03-28-2014, 08:21 AM	#7
rajarapupriya Member Location: OH, USA Join Date: Oct 2013 Posts: 18	Thank you, Jeremy !!