[BioC] Balanced Block design in LIMMA

Gordon K Smyth smyth at wehi.EDU.AU
Sun Sep 28 03:26:13 CEST 2008


Dear Erika,

Not quite sure why you are worried about this design.  It is the simplest 
possible microarray experimental design, and is covered in Section 8.1.2 
of the limma User's Guide.

> Date: Fri, 26 Sep 2008 18:24:23 +0200
> From: "Erika Melissari" <erika.melissari at bioclinica.unipi.it>
> Subject: [BioC] Balanced Block design in LIMMA
> To: <bioconductor at stat.math.ethz.ch>, "Gordon K Smyth"
> 	<smyth at wehi.EDU.AU>
> Message-ID: <017d01c91ff4$56b42e60$ba517283 at maanalysis>
> Content-Type: text/plain
>
> Hello all,
>
> I am planning the experimental design for a new microarray experiment. 
> We are interested in studying the effect of a new drug on treated mice 
> respect to untreated mice.

> In order to obtain an efficient experiment, we would like to use a 
> balanced block design, that is to balance the samples respect to the 
> dyes, as the following:
>
>        Red Green
>      array_1 treated_mice _1 untreated_mice_1
>      array_2 untreated_mice_2 treated_mice _2
>      array_3 treated_mice _3 untreated_mice_3
>      array_4 untreated_mice_4 treated_mice _4
>      array_5 treated_mice _5 untreated_mice_5
>      array_6 untreated_mice_6 treated_mice _6
>      array_7 treated_mice _7 untreated_mice_7
>      array_8 untreated_mice_8 treated_mice _8
>      array_9 treated_mice _9 untreated_mice_9
>      array_10 untreated_mice_10 treated_mice _10
>
>
> Usually I use LIMMA package to perform statistical analysis but I looked 
> LIMMA userguide up not finding anything...

> Does someone knows if LIMMA supports this experimental design?

> How do I have to consider (biological replicates, etc.) the arrays?...I 
> think they are indipendent? Is it right?

No you don't have to consider blocks.  Yes, your arrays are independent, 
because you have different mice on every array.

> Do I have to calculate duplicate correlation among spot replicates on 
> different arrays (I use arrays with only one spot per gene)?

No.

> How can I correct for the dye bias if I do not have any pair of arrays 
> with the same pair of samples dye-swapped?

Same way as always.  There is no need whatsover to have the same samples 
dye-swapped.

   Treated <- rep(c(-1,1),5)
   design <- cbind(Dye=t,Treated=Treated)
   fit <- lmFit(MA, design)
   fit <- eBayes(fit)
   topTable(fit, coef="Treated")

Best wishes
Gordon

> Any suggestions are appeciated.
>
> Best Regards
>
> Erika



More information about the Bioconductor mailing list