[BioC] RobustSpline normalization method applied to agilent data

[Gordon K Smyth]

Dear Sylvia,

My original motivation for creating the robustspline method was to 
moderate the normalization curves between multiple print tip groups. 
This was an important issue for the spotted two-colour microarrays that 
were popular ten years ago.  With modern Agilent two colour arrays, there 
are no print tips and no print tip groups, so much of the motivation for 
robustspline normalization is no longer relevant.  However the method can 
be used even when there are no print tips (or just one print tip group), 
in which case robustspline simply fits a single curve.  In this situation, 
the method is closely analogous to global loess, with a regression spline 
replacing the loess curve and with high-breakdown robust regression 
replacing the loess robustifying weights.  The method has potential 
advantages over loess normalization when there a lot of differential 
expression or the differential expression is assymetric.  I do not 
actively promote it however.

Priscilla got an error previously because she tried to set the printer 
layout herself and the settings did not agree with the number of probes on 
your arrays.  If one specifies a single tip group with:

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

then nr*nc must equal the total number of probes on your arrays.  It is 
usually best not to over-ride the settings that the software does itself 
unless you're very sure of what they are doing.

I don't understand your question about error messages depending on which 
package is used, because robustspline normalization is something I created 
and is only available in the limma package.

The software you are using is two Bioconductor releases out of date. 
Please update to current software.  See the posting guide, which asks you 
to ensure that your packages are up-to-date.

Is there any reason why you are setting the column names yourself instead 
of just using source="agilent.median"?

Best wishes
Gordon



> Date: Tue, 27 Nov 2012 22:59:14 +0000
> From: sylvia <sylvia.shiah at oicr.on.ca>
> To: <bioconductor at stat.math.ethz.ch>
> Subject: Re: [BioC] RobustSpline normalization method applied to
> 	agilent data
>
> Priscila Grynberg <priscilag at ...> writes:
>
>>
>> Dear Gordon,
>>
>> Thanks for the explanation. I didnt realize robust spline was dependent of
>> print-tips.
>>
>> Sincerely,
>>
>> Priscila
>>
> Dear all,
>
> I also applied the robust spline with the Agilent two colour array data,
> however, I did not get any errors.... are the error messages depend on the
> package that I use?
>
> This is what I did:
>
> targets        <- readTargets(file = cfg$target.files);
>
> RG <- read.maimages(
> 	files = targets$FileName,
> 	source = "agilent",
> 	columns = list(G = "gMedianSignal", Gb = "gBGMedianSignal", R =
> "rMedianSignal", Rb = "rBGMedianSignal"),
> 	annotation = c("Row", "Col", "FeatureNum", "ControlType", "ProbeName",
> "GeneName")
> 	);
>
> RG <- backgroundCorrect(RG, method="minimum", offset=1)
> MA <- normalizeWithinArrays(RG, method="robustspline")
>
>
> ### Session Info
> ################################################################################
> ###
> R version 2.14.2 (2012-02-29)
> Platform: x86_64-unknown-linux-gnu (64-bit)
>
> locale:
> [1] C
>
> attached base packages:
> [1] splines   grid      stats     graphics  grDevices datasets  methods
> [8] utils     base
>
> other attached packages:
> MASS_7.3-17
> latticeExtra_0.6-19
> lattice_0.20-0
> limma_3.10.3
>
>
>
>
>
>
>
>
>
>> On Thursday, March 8, 2012, Gordon K Smyth <smyth at ...> wrote:
>>> Dear Priscila,
>>>
>>> You can't apply robust-spline normalization to an Agilent array.
>> Robust-spline in limma is intended to moderate the loess curves over
>> different print-tip groups.  But Agilent arrays don't have print-tips. For
>> an Agilent arrays, only global loess normalization is meaningful, and that
>> does not require the layout.
>>>
>>> Best wishes
>>> Gordon
>>>
>>>> Date: Wed, 7 Mar 2012 09:16:22 -0300
>>>> From: Priscila Grynberg <priscilag at ...>
>>>> To: <bioconductor at ...>
>>>> Subject: [BioC] RobustSpline normalization method applied to agilent
>>>>        data
>>>>
>>>> Dear BioCs,
>>>>
>>>> I received some agilent data to analyse using Limma, and I'm facing a
>>>> problem when I tried to apply robustspline normalization within arrays
>>>> method.
>>>>
>>>> ##Reading the files:
>>>>
>>>> RG = read.maimages(targets,,source="agilent", columns = list(G =
>>>> "gMedianSignal", Gb = "gBGMedianSignal", R = "rMedianSignal", Rb =
>>>> "rBGMedianSignal"), annotation = c("Row", "Col","FeatureNum",
>>>> "ControlType","ProbeName","SystematicName"))
>>>>
>>>> ##Background correction:
>>>>
>>>> RGn <- backgroundCorrect (RG, method="normexp", offset=50)
>>>>
>>>> #Normalization step
>>>>
>>>> MA <- normalizeWithinArrays(RGn, method="robustspline")
>>>>
>>>> Error in normalizeWithinArrays(RGn, method = "robustspline") :  Layout
>>>> argument not specified
>>>>
>>>> ##Setting the printer information:
>>>>
>>>> nr <- length(unique(RG$genes$Row))
>>>>
>>>> nc <- length(unique(RG$genes$Col))
>>>>
>>>> RG$printer <- list(ngrid.r=1,ngrid.c=1,nspot.r=nr,nspot.c=nc)
>>>>
>>>> ## Trying again
>>>>
>>>> MA <- normalizeWithinArrays(RGn, RG$printer, method="robustspline")
>>>>
>>>> Error in X[O, ] <- ns(A[O], df = df, intercept = TRUE) : number of 
>>>> items to replace is not a multiple of replacement length
>>>>
>>>> I don't have the gal file like for this agilent slides. Can this be 
>>>> exactly the problem? Do you have any sugestions?
>>>>
>>>>
>>>> Cheers,
>>>>
>>>> Priscila
>>>>
>>>>
>>>> --
>>>> Priscila Grynberg, D.Sc.- Bioinformatics
>>>> Laborat?rio de Gen?tica Bioqu?mica
>>>> Universidade Federal de Minas Gerais
>>>> Tel: +55 31 3409-2628
>>>> CV: http://lattes.cnpq.br/8808643075395963
>>>

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