[BioC] limma and gpr flags?

[Gordon Smyth]

At 07:42 PM 23/12/2003, Simon Melov wrote:
>I have been flagging various spots in gpr files for qc reasons in genepix, 
>and they are appropriately given a negative number in the gpr file. 
>However, after reading in the gpr files via the following, carrying out 
>analysis as described in the limma guide, and generating a list of the top 
>100 DE genes via Bayes, I am still getting flagged genes in the final output.
>
>Shouldnt these be excluded from the entire analysis by the wt.fun?

Not exactly, they're just down-weighted to make it harder for them to 
appear in the top table.  Use wt.fun=wtflags(0) if you want to exclude them 
completely from the analysis.

BTW we often find it better not to background correct Genpix data. Since 
your flagged spots are almost certainly dull spots, not background 
correcting would greatly damp down on the variability of the corresponding 
log-ratios.

Gordon

>files <- dir(pattern="*.gpr")
>RG <- read.maimages(files,source="genepix",wt.fun=wtflags(0.1))
>
>
>thanks
>
>Simon.