[BioC] cDNA array with 2 levels of duplication

[Gordon K Smyth]

Dear Noah,

Only one level of duplication can be handled using the duplicateCorrelation() technique.  I
suggest that you average over the most highly correlated duplicate level, which is the
side-by-side, then use duplicateCorrelation() for the top and bottom half.  E.g.,

  MA2 <- avedups(MA,ndups=2,spacing=1)
  corfit <- duplicateCorrelation(MA2,design,ndups=2,spacing="topbottom")
  fit <- lmFit(MA2,design,ndups=2,spacing="topbottom",cor=corfit$consensus)

etc

Best wishes
Gordon

> Date: Mon, 09 Oct 2006 18:12:53 -0500
> From: "Noah Cohen" <NCOHEN at cvm.tamu.edu>
> Subject: [BioC] cDNA array with 2 levels of duplication
> Content-Type: text/plain
>
> Hi - I am looking for help with how to account for duplication at 2
> levels in analysis using lmFit in limma of data from a 2-channel cDNA
> microarray (for which ImaGene software was used for analysis).  The
> array is a 4 x 4 block of 24 rows x 24 columns of spots.  The array> has
> been duplicated on the top half and the bottom half of the slide.
> Furthermore, the spots themselves have been duplicated side-by-side.
> Thus, each slide has 4 replications of the same EST, occuring in 2
> adjacent pairs that are equally separated.  My reading of the limma
> vignette and elsewhere on within-array duplicate spots hasn't yielded a
> solution.  I've tried using the ndups options in lmFit and
> duplicateCorrelation.  While I have been successful with accounting for
> duplicate spots, I haven't been able to find a solution for the
> replication of the arrays on the same slide, where each array has
> within-array replication.
>
> Does anyone have any experience or advice?  Thanks - Noah Cohen
>
>
> Noah D. Cohen, VMD, MPH, PhD, DACVIM
> Professor
> Department of Large Animal Clinical Sciences
> College of Veterinary Medicine and Biomedical Sciences
> Texas A&M University
> College Station, Texas 77843-4475
> Telephone: 979-845-3541
> Fax: 979-847-8863
> e-mail: ncohen at cvm.tamu.edu