[BioC] limma and marray data import problem

Gordon K Smyth smyth at wehi.EDU.AU
Thu Jun 5 04:15:12 CEST 2008



On Wed, 4 Jun 2008, Piotr Stêpniak wrote:
> Dear Gordon,
>
> Thank you again for your prompt reply.
>
>> This shows that the first gpr file has read correctly.  Try putting
>> file=targets$FileName[1] to read one file at a time.
>
> If I do that, how do I put them together later on to have all the
> files in one object for further analysis?

   cbind()
   ?cbind.RGList

>> This shows that your gpr files have a problem after the last line of data.
>
> Is it possible it is because the files have different eof sign? Would
> changing them to unix like or windows like be of any help and would it
> be considered as data manipulation?

Sounds unlikely, but strange things can happen.  Storing files in 
appropriate platform mode is sensible to me.

>> There's nothing wrong with the object, the printer layout is simply not set.
>> Since limma doesn't "know" the ScanArray csv format, you have to set the
>> printer layout yourself manually.
>>
> OK, I do this:
> bialkoRaw$printer$ngrid.r=10
> bialkoRaw$printer$ngrid.c=4
> bialkoRaw$printer$nspot.c=9
> bialkoRaw$printer$nspot.r=9
> bialkoRaw$printer$spacing=1
> bialkoRaw$printer$ndups=3
>
> Is there anything else I should add? When I try to use e.g.
> maQualityPlots() I get:
>> maQualityPlots(bialkoRaw)
> Error: dims [product 2641] do not match the length of object [3240]
> In addition: Warning message:
> In samplesub & which & subset & good :
>  longer object length is not a multiple of shorter object length
>
> Which is because the array has 36 grids instead of 40 (it is not a
> full rectangle). Is there a way to go around this situation? Maybe I
> can set a grid number value?

There is no way to set the print layout in limma.  limma strictly assumes 
complete arrays.  You'd have to pad out your data object with NAs for the 
other 4 grids if you want to use print layout functions in limma.

You do not need the layout to create MA-plots or to loess normalize.

I can't give advice on maQualityPlots().

Best wishes
Gordon


More information about the Bioconductor mailing list