DESeq/DESeq normalization on different experiments

0

Entering edit mode

Guest User ★ 13k

@guest-user-4897

Last seen 11.2 years ago

Hello, I have ~100 RNA seq-samples from the same organism, but they include different experiments (each experiment of 6-16 samples). I would like to put all together into our RNA seq pipeline, that includes mapping, htseq count and DESeq. Differential expression comparisons will be done of course only between samples of a single experiment (due to possible batch effects). However, will it be a good idea to normalize all samples together? The assumption behind the normalization is that most genes are not differentially expressed, but between different experiments the variability might be higher. Therefore I wanted to ask your opinion on doing the normalization on such a large combined data-set. Thank you, Gilgi -- output of sessionInfo(): none -- Sent via the guest posting facility at bioconductor.org.

Normalization Organism DESeq Normalization Organism DESeq • 2.5k views

ADD COMMENT • link updated 11.4 years ago by Ryan C. Thompson ★ 7.9k • written 11.5 years ago by Guest User ★ 13k

0

Entering edit mode

Ryan C. Thompson ★ 7.9k

@ryan-c-thompson-5618

Last seen 12 months ago

Icahn School of Medicine at Mount Sinai…

Hi Gilgi, If you are not going to test for differential expression between experiments, then there is no purpose in normalizing them together. The more worrying problem with analyzing all your experiments as a single data set is that a single dispersion value will be estimated for each gene across all experiments. This is only ok if you believe that every gene has equal biological variability in all your experiments, which is unlikely to be the case. -Ryan On Thu May 22 01:26:55 2014, Gilgi [guest] wrote: > Hello, > > I have ~100 RNA seq-samples from the same organism, but they include different experiments (each experiment of 6-16 samples). I would like to put all together into our RNA seq pipeline, that includes mapping, htseq count and DESeq. Differential expression comparisons will be done of course only between samples of a single experiment (due to possible batch effects). However, will it be a good idea to normalize all samples together? > The assumption behind the normalization is that most genes are not differentially expressed, but between different experiments the variability might be higher. Therefore I wanted to ask your opinion on doing the normalization on such a large combined data-set. > > Thank you, > Gilgi > > -- output of sessionInfo(): > > none > > -- > Sent via the guest posting facility at bioconductor.org. > > _______________________________________________ > Bioconductor mailing list > Bioconductor at r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

ADD COMMENT • link 11.4 years ago Ryan C. Thompson ★ 7.9k

0

Entering edit mode

hi Gilgi, Yes I would second Ryan's recommendation to not normalize all the experiments together. Mike On Thu, May 22, 2014 at 11:06 AM, Ryan <rct at="" thompsonclan.org=""> wrote: > Hi Gilgi, > > If you are not going to test for differential expression between > experiments, then there is no purpose in normalizing them together. The more > worrying problem with analyzing all your experiments as a single data set is > that a single dispersion value will be estimated for each gene across all > experiments. This is only ok if you believe that every gene has equal > biological variability in all your experiments, which is unlikely to be the > case. > > -Ryan > > > On Thu May 22 01:26:55 2014, Gilgi [guest] wrote: >> >> Hello, >> >> I have ~100 RNA seq-samples from the same organism, but they include >> different experiments (each experiment of 6-16 samples). I would like to put >> all together into our RNA seq pipeline, that includes mapping, htseq count >> and DESeq. Differential expression comparisons will be done of course only >> between samples of a single experiment (due to possible batch effects). >> However, will it be a good idea to normalize all samples together? >> The assumption behind the normalization is that most genes are not >> differentially expressed, but between different experiments the variability >> might be higher. Therefore I wanted to ask your opinion on doing the >> normalization on such a large combined data-set. >> >> Thank you, >> Gilgi >> >> -- output of sessionInfo(): >> >> none >> >> -- >> Sent via the guest posting facility at bioconductor.org. >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor at r-project.org >> https://stat.ethz.ch/mailman/listinfo/bioconductor >> Search the archives: >> http://news.gmane.org/gmane.science.biology.informatics.conductor > > > _______________________________________________ > Bioconductor mailing list > Bioconductor at r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor

ADD REPLY • link 11.4 years ago Michael Love 43k

0

Entering edit mode

Thanks a lot! -----Original Message----- From: Michael Love [mailto:michaelisaiahlove@gmail.com] Sent: Thursday, May 22, 2014 6:10 PM To: Ryan Cc: Gilgi [guest]; Gilgi Friedlander; bioconductor at r-project.org Subject: Re: [BioC] DESeq/DESeq normalization on different experiments hi Gilgi, Yes I would second Ryan's recommendation to not normalize all the experiments together. Mike On Thu, May 22, 2014 at 11:06 AM, Ryan <rct at="" thompsonclan.org=""> wrote: > Hi Gilgi, > > If you are not going to test for differential expression between > experiments, then there is no purpose in normalizing them together. > The more worrying problem with analyzing all your experiments as a > single data set is that a single dispersion value will be estimated > for each gene across all experiments. This is only ok if you believe > that every gene has equal biological variability in all your > experiments, which is unlikely to be the case. > > -Ryan > > > On Thu May 22 01:26:55 2014, Gilgi [guest] wrote: >> >> Hello, >> >> I have ~100 RNA seq-samples from the same organism, but they include >> different experiments (each experiment of 6-16 samples). I would like >> to put all together into our RNA seq pipeline, that includes >> mapping, htseq count and DESeq. Differential expression comparisons >> will be done of course only between samples of a single experiment (due to possible batch effects). >> However, will it be a good idea to normalize all samples together? >> The assumption behind the normalization is that most genes are not >> differentially expressed, but between different experiments the >> variability might be higher. Therefore I wanted to ask your opinion >> on doing the normalization on such a large combined data-set. >> >> Thank you, >> Gilgi >> >> -- output of sessionInfo(): >> >> none >> >> -- >> Sent via the guest posting facility at bioconductor.org. >> >> _______________________________________________ >> Bioconductor mailing list >> Bioconductor at r-project.org >> https://stat.ethz.ch/mailman/listinfo/bioconductor >> Search the archives: >> http://news.gmane.org/gmane.science.biology.informatics.conductor > > > _______________________________________________ > Bioconductor mailing list > Bioconductor at r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor

ADD REPLY • link 11.4 years ago Gilgi Friedlander ▴ 90

Login before adding your answer.