[BioC] edgeR factorial experiment contrasts

[Jahn Davik]

Hi Jim,
Yes, I believe I can see that (except for possibly my third question). It's just that I tend to think along these lines. Are they wrong?

jahn

________________________________
Fra: James W. MacDonald [jmacdon at uw.edu]
Sendt: 3. september 2014 22:06
Til: Jahn Davik [guest]
Kopi: bioconductor at r-project.org; Jahn Davik
Emne: Re: [BioC] edgeR factorial experiment contrasts

Hi Jahn,

Things will go much easier for you if you just parameterize as cell means rather than a conventional factorial model.

grps <- factor(paste(Group, Time, sep = "_"))

design <- model.matrix (~0+grps)

Then all the comparisons are easy to do.

Best,

Jim



On Wed, Sep 3, 2014 at 12:39 PM, Jahn Davik [guest] <guest at bioconductor.org<mailto:guest at bioconductor.org>> wrote:
Hi,
I am trying to figure out how to do comparisons using edgeR but have to admit that I am on unsure ice.
My experiment is abiotic stress on two genotypes (E and J) and these have been sampled for RNA-seq at five time points (00, 01, 05, 48, and 96).
I would like to analyze this experiment as a factorial (which I am sure could be argued), and when I set up the following design:

> design<-model.matrix(~0 + Group + Time + Group*Time),

I get the following design matrix (I've taken out the 3 three replicates at experimental cell.):

> design
      GroupE GroupJ Time01 Time05 Time48 Time96 GroupJ:Time01 GroupJ:Time05 GroupJ:Time48 GroupJ:Time96
E00      1      0      0      0      0      0             0             0             0             0
J00      0      1      0      0      0      0             0             0             0             0
E01      1      0      1      0      0      0             0             0             0             0
J01      0      1      1      0      0      0             1             0             0             0
E05      1      0      0      1      0      0             0             0             0             0
J05      0      1      0      1      0      0             0             1             0             0
E48      1      0      0      0      1      0             0             0             0             0
J48      0      1      0      0      1      0             0             0             1             0
E96      1      0      0      0      0      1             0             0             0             0
J96      0      1      0      0      0      1             0             0             0             1
attr(,"assign")
 [1] 1 1 2 2 2 2 3 3 3 3
attr(,"contrasts")
attr(,"contrasts")$Group
[1] "contr.treatment"

attr(,"contrasts")$Time
[1] "contr.treatment"


I have some questions regarding testing contrasts:
First though, I realize that I can pull out the data for E00 and J00, say, and compare these using exactTest, but then I guess that you are disregarding all information that lies in the remaining data set? That does not seem to me to be a good way to do it. But I may be wrong here?

Second, if I want to do the comparison E00 vs J00 using the design indicated above, how should that be set up? Would

lrt <- glmLRT(fit,contrast=c(1,-1,0,0,0,0,0,0,0,0)) do it?

Third, if I want to compare (E00 + E01 + E05) vs (E48 + E96), how would I set up that contrast?
For my own pedagogic reasons I can show how I was thinking in this case:

E00 = L1
E01 = L1 + L4   x 2  -->  6L1 + 2L3 + 2L4
E05 = L1 + L4

versus                                       --> 2L3 + 2L4 - 3L5 -3L6

E48 = L1 + L5   x 3  -->  6L1 + 3L5 + 3L6
E96 = L1 + L6


resulting in the contrast:

lrt <- glmLRT(fit,contrast=c(0,0,2,2,-3,-3,0,0,0,0))

Then I seem to remember that contrasts should sum up to zero? If anybody could tell me where I am loosing it, I would be very obliged.

Thank you

jd





 -- output of sessionInfo():

R version 3.1.0 (2014-04-10)
Platform: x86_64-w64-mingw32/x64 (64-bit)

locale:
[1] LC_COLLATE=Norwegian (Bokmål)_Norway.1252  LC_CTYPE=Norwegian (Bokmål)_Norway.1252
[3] LC_MONETARY=Norwegian (Bokmål)_Norway.1252 LC_NUMERIC=C
[5] LC_TIME=Norwegian (Bokmål)_Norway.1252

attached base packages:
[1] splines   stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] edgeR_3.6.8          limma_3.20.9         BiocInstaller_1.14.2

loaded via a namespace (and not attached):
[1] tools_3.1.0
>

--
Sent via the guest posting facility at bioconductor.org<http://bioconductor.org>.

_______________________________________________
Bioconductor mailing list
Bioconductor at r-project.org<mailto:Bioconductor at r-project.org>
https://stat.ethz.ch/mailman/listinfo/bioconductor
Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor



--
James W. MacDonald, M.S.
Biostatistician
University of Washington
Environmental and Occupational Health Sciences
4225 Roosevelt Way NE, # 100
Seattle WA 98105-6099

	[[alternative HTML version deleted]]