[BioC] limma: paired samples, before and after treatment

[James W. MacDonald]

Hi Klemens,

On 12/11/2010 12:53 PM, Vierlinger Klemens wrote:
> Dear all,
>
> I have a single colour experiment with paired samples (samples before and after treatment from the same mouse) and different treatments. I read the section in the limma user guide dealing with paired samples, but am not convinced that this is the best approach for my case.
>

It is functionally identical to doing the subtraction first. As an example:

First using a blocking variable...

 > fakedat <- matrix(rnorm(1000), ncol = 10))
 > pair <- factor(rep(1:5, each=2))
 > prepost <- factor(rep(1:2, times = 5))
 > design1 <- model.matrix(~pair+prepost)
 > topTable(eBayes(lmFit(fakedat, design1)), coef="prepost2")
         logFC         t    P.Value adj.P.Val         B
18  1.5097296  2.306015 0.02110979 0.9982365 -4.546033
47  1.4465578  2.209524 0.02713820 0.9982365 -4.551000
67 -1.2536337 -1.914845 0.05551227 0.9982365 -4.564854
9   1.2044666  1.839745 0.06580563 0.9982365 -4.568068
84  1.1536478  1.762123 0.07804852 0.9982365 -4.571255
54 -1.1369380 -1.736600 0.08245784 0.9982365 -4.572273
96 -1.0943923 -1.671614 0.09460047 0.9982365 -4.574798
4  -1.0334116 -1.578470 0.11445772 0.9982365 -4.578249
51 -0.9355558 -1.429001 0.15300381 0.9982365 -4.583373
61  0.9143791  1.396655 0.16251725 0.9982365 -4.584415

Now take the same data, and pair by hand...


 > fakedat2 <- fakedat[,seq(2,10,2)] - fakedat[,seq(1,9,2)]
 > topTable(eBayes(lmFit(fakedat2)))
         logFC         t    P.Value adj.P.Val         B
18  1.5097296  2.306015 0.02110979 0.9982365 -4.546033
47  1.4465578  2.209524 0.02713820 0.9982365 -4.551000
67 -1.2536337 -1.914845 0.05551227 0.9982365 -4.564854
9   1.2044666  1.839745 0.06580563 0.9982365 -4.568068
84  1.1536478  1.762123 0.07804852 0.9982365 -4.571255
54 -1.1369380 -1.736600 0.08245784 0.9982365 -4.572273
96 -1.0943923 -1.671614 0.09460047 0.9982365 -4.574798
4  -1.0334116 -1.578470 0.11445772 0.9982365 -4.578249
51 -0.9355558 -1.429001 0.15300381 0.9982365 -4.583373
61  0.9143791  1.396655 0.16251725 0.9982365 -4.584415

Best,

Jim



> Would it be correct to simply subtract the expression values after-before for each mouse and then use limma on these subtracted values to estimate the effect of the treatment?
>
> Many thanks in advance for your ideas!
> Best
> Klemens
>
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> 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

-- 
James W. MacDonald, M.S.
Biostatistician
Douglas Lab
University of Michigan
Department of Human Genetics
5912 Buhl
1241 E. Catherine St.
Ann Arbor MI 48109-5618
734-615-7826
**********************************************************
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues