[BioC] EdgeR: ANOVA-like analysis
Gordon K Smyth
smyth at wehi.EDU.AU
Sat Jun 25 00:20:34 CEST 2011
On Fri, 24 Jun 2011, Christian M. Probst wrote:
> Many thanks, Dr. Gordon.
> I have run the analysis as suggested.
> Two small questions:
> a) I would like to understand deeply the coef option, besides what is
> described in the glmLRT help. Is there a place to look for it?
Options to functions are described on the help page for that function so,
no, there is not somewhere else to look.
> b) If I want to plot the adjusted count values that were used for the GLM
> fit and LR analysis, where can I find the equivalent of $pseudo.alt for the
> CR analysis? I suppose fitted.values is not the correct answer, as its sums
> are not "library size equal".
glmFit doesn't use adjusted counts. It uses the original counts that you
> Thanks in advance.
> On Wed, Jun 22, 2011 at 9:16 PM, Gordon K Smyth <smyth at wehi.edu.au> wrote:
>> Dear Christian,
>> limma syntax doesn't always work in edgeR. We will work to make them as
>> analogous as we can, but generalized linear models in edgeR bring in a few
>> unavoidable complications into edgeR that are not present in limma.
>> At the moment, the contrast argument to glmLRT() must be a vector, not a
>> matrix, as is explained on the help page for that function. In future we
>> will add the matrix capability, but for now you will need to perform
>> F-test-like likelihood ratio tests using the coef argument instead. This
>> means that you must paramatrize your design matrix so that it contains an
>> intercept column:
>> design <- model.matrix(~group)
>> Then later you can use
>> glmLRT(..., coef=2:4)
>> and so on. This will do the combined test that you want on 3 df for
>> differences between the 4 groups.
>> Best wishes
>> Date: Mon, 20 Jun 2011 11:12:39 -0300
>>> From: "Christian M. Probst" <cprobst at fiocruz.br>
>>> To: bioconductor at r-project.org
>>> Subject: [BioC] EdgeR: ANOVA-like analysis
>>> Dear list,
>>> I am using edgeR to analyze a RNA-Seq dataset consisting of 4 samples
>>> (groups), each with 5 biological replicates.
>>> I have read in the Limma manual, as it has an extensive description of
>>> design models. In section 8.6, there is a description for multiple groups
>>> test, and I have tried to follow the steps there. But the edgeR objects
>>> distinct from the Limma, so I am stucked.
>>> Shortly, I want to use edgeR to identify differential expression in all
>>> pair-wise comparisons of the 4 groups, as well as combine all pair-wise
>>> comparisons into one F-test.
>>> I have used the following construction:
>>> deF<-glmLRT(dgecRNAF,glmFit(**dgecRNAF, mmcRNA,
>>> dispersion=dgecRNAF$common.**dispersion), contrast=contrast.matrix)
>>> where dgecRNAF is my DGEList object, mmcRNA is my model matrix, and
>>> contrast.matrix is:
>>> Levels A-B A-C A-D B-C B-D C-D
>>> A 1 1 1 0 0 0
>>> B -1 0 0 1 1 0
>>> C 0 -1 0 -1 0 1
>>> D 0 0 -1 0 -1 -1
>>> Thanks in advance,
>>> Dr. Christian Macagnan Probst
>>> Adjunct Researcher in Public Health
>>> Bioinformatics and Computational Biology Laboratory
>>> Functional Genomics Laboratory
>>> Carlos Chagas Institute - ICC/FIOCRUZ
>>> Curitiba - PR - Brasil
>>> Reseacher ID: http://www.researcherid.com/**rid/B-8678-2009<http://www.researcherid.com/rid/B-8678-2009>
>> The information in this email is confidential and intended solely for the
>> You must not disclose, forward, print or use it without the permission of
>> the sender.
>> Esta mensagem foi verificada pelo sistema de antivírus e
>> acredita-se estar livre de perigo.
> Dr. Christian Macagnan Probst
> Adjunct Researcher in Public Health
> Bioinformatics and Computational Biology Laboratory
> Functional Genomics Laboratory
> Carlos Chagas Institute - ICC/FIOCRUZ
> Curitiba - PR - Brasil
> Reseacher ID: http://www.researcherid.com/rid/B-8678-2009
> Curriculum Lattes: http://lattes.cnpq.br/9371475697752743
More information about the Bioconductor