[BioC] Read single channel GenePix in limma [was: Analyze miRNA experiment in Bioconductor]

[Gordon K Smyth]

Dear Paul,

The limma User's Guide doesn't discuss how to read single channel data, 
but how to do this has been described half a dozen times on this mailing 
list.  Since limma is designed for two colours, you can fool it by giving 
two column names and ignoring the one you don't need.  If you only have 
the Cy3 channel foreground for example you might use

   Cy3 <- "F532 Mean"
   RG <- read.maimages(source="genepix",columns=list(R=Cy3,G=Cy3))

then

   RG$R <- NULL

to remove the extraneous values.

Then RG$G could be given as input to vsnMatrix() and the output analysed 
with lmFit().

Please don't edit your GenePix files manually, there's no need.  It's 
prone to introducing errors and is non-reproducible.

The error message "number of items to replace is not a multiple of 
replacement length" is not caused by having only one channel.  limma gives 
a far more informative message in that case.  The most likely explanation 
is that your GenePix files are not of equal lengths.  If that is indeed 
the problem, then the limma package doesn't offer any easy solution.  Your 
only approach would be to read the files in individually, then align the 
expression values yourself.

You cannot use read.maimages() with source="imagene" because you do not 
have ImaGene files.

Best wishes
Gordon


> Date: Fri, 9 May 2008 15:54:39 +0100
> From: "Paul Geeleher" <paulgeeleher at gmail.com>
> Subject: Re: [BioC] Analyze miRNA experiment in Bioconductor
> To: "Wolfgang Huber" <huber at ebi.ac.uk>
> Cc: bioconductor at stat.math.ethz.ch
>
> Doesn't seem to be anything in the users guide specific to this kind
> of analysis unfortunately.
>
> -Paul
>
> On Thu, May 8, 2008 at 10:31 AM, Wolfgang Huber <huber at ebi.ac.uk> wrote:
>> Dear Paul,
>>
>>> Hmm interesting. I might try introducing the extra columns into the
>>> files and specifying all the values as 0. I can't see why that
>>> shouldn't work?
>>
>> It might, but Narendra's suggestion of reading the limma users guide is  a
>> worthwhile option to consider.
>>
>>  Best wishes
>>        Wolfgang
>>
>> ------------------------------------------------------------------
>> Wolfgang Huber  EBI/EMBL  Cambridge UK  http://www.ebi.ac.uk/huber
>>
>>>
>>> -Paul
>>>
>>> On Wed, May 7, 2008 at 1:39 PM, Narendra Kaushik
>>> <kaushiknk at cardiff.ac.uk> wrote:
>>>>
>>>> You can specify your red channel like this:
>>>>
>>>>  RG <- read.maimages(files,source="genepix",  columns=list(R="F635
>>>> Median",G="F532
>>>>  Median",Rb="B635",Gb="B532"))
>>>>
>>>>  I will suggest you read limma guide.
>>>>
>>>>  But I think your have data from Imagene package which gives one file for
>>>> each channel, you can:
>>>>
>>>>  files <- targets[,c("FileNameCy3","FileNameCy5")]
>>>>  RG <- read.maimages(files, source="imagene")
>>>>
>>>>  Hope, this helps
>>>>
>>>>  Narendra
>>>>
>>>> >>> "Paul Geeleher" <paulgeeleher at gmail.com> 07/05/2008 13:24:01 >>>
>>>>
>>>>
>>>> Hi Deepayan,
>>>>
>>>>  Thanks for your reply. I suppose my main concern is how I should read
>>>>  in the data initially in order to be able to use the normal tools to
>>>>  analyze the data. Reading the data normally like this:
>>>>
>>>>  RG <- read.maimages( files, source="genepix")
>>>>
>>>>  Gives the following error:
>>>>
>>>>  Error in RG[[a]][, i] <- obj[, columns[[a]]] :
>>>>  number of items to replace is not a multiple of replacement length
>>>>
>>>>
>>>>  I'm assuming this is down to the fact that the files only contain
>>>>  intensity data for one color rather than two?
>>>>
>>>>  How should I go about reading the data?
>>>>
>>>>  Thanks alot,
>>>>
>>>>  -Paul.
>>>>
>>>>  On Tue, May 6, 2008 at 10:15 PM, Deepayan Sarkar
>>>>  <deepayan.sarkar at gmail.com> wrote:
>>>> > On 5/6/08, Paul Geeleher <paulgeeleher at gmail.com> wrote:
>>>> > > Dear Members,
>>>> > >
>>>> > >  I've inherited a bunch of GenePix files from an miRNA experiment.
>>>> They
>>>> > >  are single color arrays, ( as opposed to 2 color as is the norm
>>>> for
>>>> > >  GenePix I think). There is a subset of 7 arrays and I wish to
>>>> compare
>>>> > >  a group of 4 of these to the other group of 3 and analyze
>>>> differential
>>>> > >  expression between the two groups. I was hoping somebody could
>>>> point
>>>> > >  me in the right direction of how I'd go about doing this with
>>>> > >  Bioconductor? Is it possible using the Limma package? Is there any
>>>> > >  code out there to assist me?
>>>> > >
>>>> > >  I've experience in analyzing Affymetrix data using Limma and PUMA,
>>>> but
>>>> > >  not GenePix, and the Limma Users Guide seems to focus on analyzing
>>>> two
>>>> > >  dye experiments.
>>>> >
>>>> >  Any analysis ultimately boils down to some sort of normalization, and
>>>> >  the actual differential expression analysis. The second part in limma
>>>> >  (lmFit, etc.) can work with any expression matrix, irrespective of
>>>> >  whether it's 2-color or 1-color (or affy).
>>>> >
>>>> >  We have been working with a miRNA array dataset recently, and we used
>>>> >  limma to read in the GPR files and do the differential expression
>>>> >  analysis (on one channel). For normalization, many of the standard
>>>> >  microarray algorithms probably don't make much sense, but VSN seems
>>>> to
>>>> >  work fine.
>>>> >
>>>> >  We don't really have code (beyond what's already in limma and vsn)
>>>> >  that is generally useful; most of the work is in figuring out which
>>>> >  rows are of interest (i.e., those representing human miRNAs),
>>>> >  combining the replicates (you seem to have four of each), etc. I'm
>>>> >  happy to give you more details if you are interested.
>>>> >
>>>> >  -Deepayan