[BioC] limma Within Array Normalisation By Controls

[Gordon K Smyth]

Slight correction:

  RG$printer <- list(ngrid.r=1, ngrid.c=1,
     nspot.r=nrow(RG), nspot.c=1)

would be better, so at least that the number of spots is correct.

Gordon

On Wed, 26 Jan 2011, Gordon K Smyth wrote:

> Dear Dario,
>
> In truth, limma shouldn't need to know the layout for "control" 
> normalization.  I'll change the code to remove the check in the future.
>
> In the meantime, you can set
>
>  RG$printer <- list(ngrid.r=1, ngrid.c=1,
>    nspot.r=nrow(RG), nspot.c=ncol(RG))
>
> then the code will work fine.
>
> However, I suspect that the Nimblegen probes marked "random" may actually be 
> negative control probes, whereas control normalization requires positive 
> control probes covering a wide range of intensities.  Nimblegen arrays 
> usually have an annotation column called ControlType, which is equal to 0 for 
> negative controls.  You can also view where the "random" probes fall using 
> plotMA().  If the control probes are all at the lowest intensities, then they 
> are not suitable for use with control normalization.  On the other hand, you 
> could still make use of the fact that these probes shouldn't change between 
> samples, by using "loess" normalization and upweighting these probes.
>
> BTW, the reason why the layout isn't set automatically is that limma doesn't 
> have a read method for Nimblegen arrays, so you probably read the data using 
> source="generic", so limma doesn't know whether it's a new commercial array 
> or an old spotted array.
>
> Best wishes
> Gordon
>
>> Date: Tue, 25 Jan 2011 13:00:29 +1100 (EST)
>> From: Dario Strbenac <D.Strbenac at garvan.org.au>
>> To: bioconductor at r-project.org
>> Subject: [BioC] limma Within Array Normalisation By Controls
>> Message-ID: <20110125130029.BKP14739 at gimr.garvan.unsw.edu.au>
>> Content-Type: text/plain; charset=us-ascii
>> 
>> Hello,
>> 
>> I have an experiment where there are some Nimblegen 2.1 million probe Mouse 
>> promoter arrays with IPs always in the red channel and input DNA always in 
>> the green channel. I notice that the arrays have some probes of type 
>> Random, which I assume should not be changing between channels. So, I'd 
>> like to use normalizeWithinArrays with method = "control". I call the 
>> method as :
>> 
>> normalizeWithinArrays(RG, method = "control", controlspots = 
>> RG$genes$Status == "Random")
>> 
>> and get the error :
>> 
>> Error in normalizeWithinArrays(RG, method = "control", controlspots = 
>> RG$genes$Status ==  :
>>  Layout argument not specified
>> 
>> I'm confused as to why it requires the layout ? I didn't think newer array 
>> designs had print tip groups or replicate spots ?
>> 
>> --------------------------------------
>> Dario Strbenac
>> Research Assistant
>> Cancer Epigenetics
>> Garvan Institute of Medical Research
>> Darlinghurst NSW 2010
>> Australia
>

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