[BioC] limma model to treat replicate spots

[Matthew Ritchie]

Hi Joyce,

The normalizeWithinArrays() function doesn't average across the duplicate
spots (as its name suggests, it normalizes the log-ratios of all spots on
each array).

You can average over the duplicate spots by running the normalized
log-ratios through lmFit() and specifying ndups=2, as well as a value for
the 'spacing' argument (if the duplicate spots are side by side in the
rows of the print-tip groups, spacing=1.  If they are below each other,
the spacing will be the number of spots in the row of a print-tip). 
You'll also need a design matrix, which will depend on your experiment,
and what you're interested in (check out the limma user's guide for some
examples).  You might also want to use the duplicateCorrelation() function
to estimate the spatial correlation between the duplicate spots.  I've put
an example of the code you might use below (if the duplicate spots were
printed side-by-side).

Best wishes,

Matt Ritchie


M <- normalizeWithinArrays(RG)
cor <- duplicateCorrelation(M, design, ndups=2, spacing=1) # assuming you
have specified a design matrix
fit <- lmFit(M, design, ndups=2, spacing=1, correlation=cor$cor)
aveM <- fit$coef


> Hello,
> I have cDNA, I tried limma to do normalizeWithinArrays, but the results
> looks  like they did not aveage spots. Each spot was printed twice in my
> cDNA array,  how limma average them? I have all the log ratio from every
> printed spot(twice  printed). When I did correlation between replicated,
> their coef is very low,  which compared to arrayTool analysis, is
> significant different. When I looked  at the log ratio, limma has twice
> as many as arrayTool log ratio, so I think  limma did not average my
> spots.
>
> Any idea and suggestion is greatly appreciated!
>
> Thanks
>
> ********************************
> Joyce Gu   M.S.
> Scientific Programmer
>
> Cancer Genomics Group
> Baylor College of Medicine
> Phone: 832-824-4680
> Fax: 832-825-4038
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor