[BioC] Limma with homemade chip, multiple probes per gene, no replicates?

[Gordon K Smyth]

normalizeWithinArrays() produces an MA object.  You can plot the results 
using

     plotMA(MA)

The M-values are the log2-fold changes.  Just select genes with 
large M-values.  For example,

     o <- order(abs(MA$M),decreasing=TRUE)
     tab <- data.frame(MA$genes,A=MA$A,M=MA$M)
     tab[o[1:20],]

Best wishes
Gordon

---------------------------------------------
Professor Gordon K Smyth,
Bioinformatics Division,
Walter and Eliza Hall Institute of Medical Research,
1G Royal Parade, Parkville, Vic 3052, Australia.
http://www.statsci.org/smyth


> Date: Tue, 26 Nov 2013 05:01:06 -0800 (PST)
> From: "Gordon [guest]" <guest at bioconductor.org>
> To: bioconductor at r-project.org, assafgordon at gmail.com
> Subject: [BioC] Limma with homemade chip, multiple probes per gene,	no
> 	replicates?
>
>
> Hello,
>
> I have data from a two-color, homemade, microchip array (one GPR file).
> The chip contains multiple probes for each gene (just to be sure I'm using the correct terms: the GPR file has multiple lines with the same NAME and readings of F635 and F532 for each gene of interest). There are also many lines marked with "EMPTY" .
>
> There is only one GPR file, and no replicates at the moment.
>
> (I'm trying to help a friend deal with an existing situation, which is less than ideal - but that's what we've got.)
>
> First,
> How do I deal with the multiple probes for each gene?
> Should I pre-process the data and calculate the mean/median for each, then feed "limma" only the aggregated values? or is there a better way (taking the variability within each group into account) ?
>
> Second,
> The data works fine with limma's "read.maimages()", "normalizaeWithinArrays()", and "lmFit()".
>
> However, "eBAyes" and "topTable" do not work (as explained in many other threads, they will not work without replicates).
>
> Despite the fact that it's inadvisable to continue without replicates, is there a way to perform some sort of differential expression analysis?
> (Akin to EdgeR's method of working when there are no replicates?)
>
> Many thanks,
> -gordon

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}