test for differential expresssion

0

Entering edit mode

Rao,Xiayu ▴ 550

@raoxiayu-6003

Last seen 9.5 years ago

United States

Hello, Thank you in advance for your help! I have 16 tumor samples and each one has a corresponding normal sample, which are 16 pairs. For one particular pair, there is a matched metastasis sample. When I analyzed the data using limma package and followed their protocol, is it right that I fit a linear model, and compare two contrasts of Tumor vs. Normal, and Metastasis vs. Normal? Or should I ignore that 1 metastasis sample and just use paired ttest to test between Tumor and Normal? Best, Xiayu [[alternative HTML version deleted]]

• 1.4k views

ADD COMMENT • link updated 11.4 years ago by Gordon Smyth 52k • written 11.4 years ago by Rao,Xiayu ▴ 550

0

Entering edit mode

Michael Breen ▴ 370

@michael-breen-5999

Last seen 10.2 years ago

Xiayu, 1 metastasis sample won't reveal any honest statistics, unless you have others the rule would be to ignore it. Paired T-Test is the most sensitive measure. Use that if you have only 1 treatment (Tumor). . Good luck, Michael On Wed, Jun 26, 2013 at 10:47 AM, Rao,Xiayu <xrao@mdanderson.org> wrote: > Hello, > > > > Thank you in advance for your help! > > > > I have 16 tumor samples and each one has a corresponding normal sample, > which are 16 pairs. For one particular pair, there is a matched metastasis > sample. When I analyzed the data using limma package and followed their > protocol, is it right that I fit a linear model, and compare two contrasts > of Tumor vs. Normal, and Metastasis vs. Normal? Or should I ignore that 1 > metastasis sample and just use paired ttest to test between Tumor and > Normal? > > > > Best, > > Xiayu > > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor@r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > [[alternative HTML version deleted]]

ADD COMMENT • link 11.4 years ago Michael Breen ▴ 370

0

Entering edit mode

Gordon Smyth 52k

@gordon-smyth

Last seen 55 minutes ago

WEHI, Melbourne, Australia

Dear Xiayu, > Date: Wed, 26 Jun 2013 17:47:22 +0000 > From: "Rao,Xiayu" <xrao at="" mdanderson.org=""> > To: "'bioconductor at r-project.org'" <bioconductor at="" r-project.org=""> > Subject: [BioC] test for differential expresssion > > Hello, > > Thank you in advance for your help! > > I have 16 tumor samples and each one has a corresponding normal sample, > which are 16 pairs. For one particular pair, there is a matched > metastasis sample. When I analyzed the data using limma package and > followed their protocol, is it right that I fit a linear model, and > compare two contrasts of Tumor vs. Normal, and Metastasis vs. Normal? This is fine. Of course the genewise variability has to estimated purely from the tumor samples; the metastisized sample doesn't contribute to the variability estimate because there is only one. Also the statistical test of metastasis vs normal will not be very powerful, because it is based on only one sample. Nevertheless, it can all be done. > Or should I ignore that 1 metastasis sample and just use paired ttest to > test between Tumor and Normal? It makes no difference. The limma paired t-test of tumor vs normal will be exactly the same whether or not the metastasis sample is included. The metastasis sample doesn't contribute to the genewise variability estimate, so it makes no difference to the test of tumor vs normal whether or not it is incuded in the data. You may as well include all the data. That is the philosophy behind the limma package. Best wishes Gordon > Best, > Xiayu ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:4}}

ADD COMMENT • link 11.4 years ago Gordon Smyth 52k

0

Entering edit mode

Thank you very much! It is great to know. I am curious that how limma can recognize if our samples are paired or not, and then the corresponding ttest for the contrast comparison would be paired or unpaired? Was it coming from the targets file? This is kind of important. Thanks, Xiayu -----Original Message----- From: Gordon K Smyth [mailto:smyth@wehi.EDU.AU] Sent: Thursday, June 27, 2013 6:53 PM To: Rao,Xiayu Cc: Bioconductor mailing list Subject: test for differential expresssion Dear Xiayu, > Date: Wed, 26 Jun 2013 17:47:22 +0000 > From: "Rao,Xiayu" <xrao at="" mdanderson.org=""> > To: "'bioconductor at r-project.org'" <bioconductor at="" r-project.org=""> > Subject: [BioC] test for differential expresssion > > Hello, > > Thank you in advance for your help! > > I have 16 tumor samples and each one has a corresponding normal > sample, which are 16 pairs. For one particular pair, there is a > matched metastasis sample. When I analyzed the data using limma > package and followed their protocol, is it right that I fit a linear > model, and compare two contrasts of Tumor vs. Normal, and Metastasis vs. Normal? This is fine. Of course the genewise variability has to estimated purely from the tumor samples; the metastisized sample doesn't contribute to the variability estimate because there is only one. Also the statistical test of metastasis vs normal will not be very powerful, because it is based on only one sample. Nevertheless, it can all be done. > Or should I ignore that 1 metastasis sample and just use paired ttest > to test between Tumor and Normal? It makes no difference. The limma paired t-test of tumor vs normal will be exactly the same whether or not the metastasis sample is included. The metastasis sample doesn't contribute to the genewise variability estimate, so it makes no difference to the test of tumor vs normal whether or not it is incuded in the data. You may as well include all the data. That is the philosophy behind the limma package. Best wishes Gordon > Best, > Xiayu ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:6}}

ADD REPLY • link 11.4 years ago Rao,Xiayu ▴ 550

0

Entering edit mode

Could you please read the section in the User's Guide on "Paired Samples". Gordon On Fri, 28 Jun 2013, Rao,Xiayu wrote: > Thank you very much! It is great to know. > > I am curious that how limma can recognize if our samples are paired or > not, and then the corresponding ttest for the contrast comparison would > be paired or unpaired? Was it coming from the targets file? This is kind > of important. > > Thanks, > Xiayu > > -----Original Message----- > From: Gordon K Smyth [mailto:smyth at wehi.EDU.AU] > Sent: Thursday, June 27, 2013 6:53 PM > To: Rao,Xiayu > Cc: Bioconductor mailing list > Subject: test for differential expresssion > > Dear Xiayu, > >> Date: Wed, 26 Jun 2013 17:47:22 +0000 >> From: "Rao,Xiayu" <xrao at="" mdanderson.org=""> >> To: "'bioconductor at r-project.org'" <bioconductor at="" r-project.org=""> >> Subject: [BioC] test for differential expresssion >> >> Hello, >> >> Thank you in advance for your help! >> >> I have 16 tumor samples and each one has a corresponding normal >> sample, which are 16 pairs. For one particular pair, there is a >> matched metastasis sample. When I analyzed the data using limma >> package and followed their protocol, is it right that I fit a linear >> model, and compare two contrasts of Tumor vs. Normal, and Metastasis vs. Normal? > > This is fine. > > Of course the genewise variability has to estimated purely from the tumor samples; the metastisized sample doesn't contribute to the variability estimate because there is only one. Also the statistical test of metastasis vs normal will not be very powerful, because it is based on only one sample. Nevertheless, it can all be done. > >> Or should I ignore that 1 metastasis sample and just use paired ttest >> to test between Tumor and Normal? > > It makes no difference. The limma paired t-test of tumor vs normal will be exactly the same whether or not the metastasis sample is included. > The metastasis sample doesn't contribute to the genewise variability estimate, so it makes no difference to the test of tumor vs normal whether or not it is incuded in the data. > > You may as well include all the data. That is the philosophy behind the limma package. > > Best wishes > Gordon > >> Best, >> Xiayu ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:4}}

ADD REPLY • link 11.4 years ago Gordon Smyth 52k

Login before adding your answer.