[BioC] arrayQualityMetrics problem with Agilent 2-colour

Thu Jun 24 17:56:06 CEST 2010

Thanks for the reply Audrey

On Wed, 23 Jun 2010 16:51:08 +0200
 Audrey Kauffmann <ak.bergonie at gmail.com> wrote:
> Hi Edwin,
> 
> arrayQualityMetrics is not really appropriate for that, maybe you can
> consider using the clustering and heatmap functions outside from the
> package.
> 

This might be the best solution for the time being. I also tried the
arrayQuality package, but it keeps asking me for my mouse annotations.
Unfortunately, I work with neither mice nor humans, but with
Arabidopsis (those green thingies that photosynthesize).
Some searching on the Internet (sorry, I lost the source) gave the tip
of using Pearson correlation for distance measures of arrays. I came up
with the following code snippet, but I do not know if this is an
appropriate distance measure for RAW intensity values. The Pearson
correlation is 0 to 1, with 1 being perfect correlation between a pair
of arrays.
What additional code is needed to determine outliers by setting a 1.5 *
IQR threshold? I have to do something with object d, correct?
#Generate correlation matrix for green and red channels
  allGR <- cbind(RG$G,RG$R)
  d <- cor(allGR, method="pearson")
#Plot dendrogram of distances
  plot(hclust(dist(1-d)))
#Plot heatmap of all samples against all
  heatmap(d, Rowv=NA, Colv=NA, symm=TRUE)

> Also, regarding using aqm.prepdata on a RGList is not implemented,
> you
> would first need to convert your RGList into a NChannelSet.
> 

I have tried this conversion, but I received an error from aqm.heatmap.
  RGNC <- as(RG, "NChannelSet")
  RGprep = aqm.prepdata(expressionset=RGNC, do.logtransform = TRUE)
  hm <- aqm.heatmap(dataprep=RGNC)
Error in if (n < 2) stop("must have n >= 2 objects to cluster") :
argument is of length zero

That is too bad. The RGprep object has slots for number of channels and
for the sample names.

> A trick, not recommended but which would work, is to convert your
> RGList as an ExpressionSet (usually for one channel) where each R and
> G channels would be treated as separated chips, then use aqm.prepdata
> and aqm.heatmap.
> 

Another reason that I have read against this option is the raw
intensity I read into my RGList is not expression, and therefore not
appropriate for an ExpressionSet. I might open a semantic can of worms,
but I think the 2-colour specialists consider the data as expression
only after background subtraction, normalization and output of the
log-ratios.

Edwin
-- 
> Please let me know if you have more questions regarding this,
> Audrey
> 
> 2010/6/23 Edwin Groot <edwin.groot at biologie.uni-freiburg.de>:
> > Hello all,
> > Would anyone know of a simple way to plot the heatmap in
> > arrayQualityMetrics on a per-channel basis, rather than the
> per-array
> > basis?
> >
> > I have 16 2-colour Agilent arrays read in by the LIMMA package, but
> > arrayQualityMetrics shows only 16 arrays in the heatmap, rather
> than
> > the 32 red and green samples that I would expect. When one array is
> an
> > outlier in the heatmap, I do not know which sample (red or green)
> is at
> > fault.
> >
> > The code that I used is quite simple:
> >> library(limma)
> >> library(arrayQualityMetrics)
> >> targets <- readTargets("qc.exp")
> >> RG <- read.maimages(targets$FileName, source="agilent")
> >> arrayQualityMetrics(expressionset=RG, outdir="default", force =
> TRUE,
> > do.logtransform = TRUE)
> > The directory 'default' has been created.
> > KernSmooth 2.23 loaded
> > Copyright M. P. Wand 1997-2009
> > (loaded the KernSmooth namespace)
> > [[1]]
> > [[2]]
> >
> > I also attempted a custom call to aqm.* functions. However, the
> > preparatory step fails:
> >> RGprep = aqm.prepdata(expressionset=RG, do.logtransform = TRUE)
> > Error in function (classes, fdef, mtable)  :
> >  unable to find an inherited method for function "aqm.prepdata",
> for
> > signature "RGList"
> >
> >> sessionInfo()
> > R version 2.11.1 (2010-05-31)
> > i486-pc-linux-gnu
> >
> > locale:
> >  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
> >  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8
> >  [5] LC_MONETARY=C              LC_MESSAGES=en_US.UTF-8
> >  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
> >  [9] LC_ADDRESS=C               LC_TELEPHONE=C
> > [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
> >
> > attached base packages:
> > [1] stats     graphics  grDevices utils     datasets  methods  
> base
> >
> >
> > other attached packages:
> > [1] arrayQualityMetrics_2.6.0 affyPLM_1.24.0
> > [3] preprocessCore_1.10.0     gcrma_2.20.0
> > [5] affy_1.26.1               Biobase_2.8.0
> > [7] limma_3.4.2
> >
> > loaded via a namespace (and not attached):
> >  [1] affyio_1.16.0        annotate_1.26.0      AnnotationDbi_1.10.1
> >  [4] beadarray_1.16.0     Biostrings_2.16.3    DBI_0.2-5
> >  [7] genefilter_1.30.0    grid_2.11.1          hwriter_1.2
> > [10] IRanges_1.6.4        lattice_0.18-8       latticeExtra_0.6-11
> > [13] marray_1.26.0        RColorBrewer_1.0-2   RSQLite_0.9-1
> > [16] simpleaffy_2.24.0    splines_2.11.1       stats4_2.11.1
> > [19] survival_2.35-8      tools_2.11.1         vsn_3.16.0
> > [22] xtable_1.5-6
> >
> > Thanks in advance,
> > Edwin
> > --
> > Dr. Edwin Groot, postdoctoral associate
> > AG Laux
> > Institut fuer Biologie III
> > Schaenzlestr. 1
> > 79104 Freiburg, Deutschland
> > +49 761-2032945
> >
> > _______________________________________________
> > Bioconductor mailing list
> > Bioconductor at stat.math.ethz.ch
> > https://stat.ethz.ch/mailman/listinfo/bioconductor
> > Search the archives:
> http://news.gmane.org/gmane.science.biology.informatics.conductor
> >
> 
> 
> 
> -- 
> Audrey Kauffmann
> Bergonie Cancer Institute
> 229 Cours de l'Argonne
> 33076 Bordeaux
> France
> +33.5.56.33.04.53