[BioC] default BG subtraction in plotMA...

[Gordon K Smyth]

Hi Jenny,

I can understand your mis-understanding here, but I don't think that documentation in the plotMA
function is the answer.

The limma philosophy is that RGLists always map to MALists using background correction.  This is
true for all limma functions which operate on RGList objects.  The default background correction
method is that defined by backgroundCorrect (or equivalently normalizeWithinArrays).  If you want
to use something other than the default, then you must call backgroundCorrect() explicitly, for
example with method="none".  Then the RGList object is "registered" as having been background
corrected by having its Rb and Gb components removed.

You are thinking of not background correcting as being the default, but I view this action (acting
as if the background is zero) as a deliberate decision which should be declared explicitly.

I will think about how to make this philosophy more obvious in the limma documentation so that
others don't fall into the mistake that you describe.  But I don't want to have to state the
background correction philosophy in the documentation of every function which operates on RGLists
and may convert to an MAList, nor do I want to add background correction options for all of these
functions.  I want the philophy to be stated in one place to cover all.

Cheers
Gordon

On Wed, May 10, 2006 12:58 am, Jenny Drnevich wrote:
> Hi Gordon,
>
> I ran across this non-documentation a while back when trying to visualize
> my raw data, and now after having to point it out to three separate people
> I realized I should tell you about it. When you pass an RGList to 'plotMA'
> or 'plotMA3by2' it gets converted to a MAList by 'MA.RG', which by default
> will do a simple background subtraction if there are $Rb and $Gb items in
> the RGList. However, it's not documented anywhere in 'plotMA' or
> 'plotMA3by2' that the plots will be of background-subtracted values if
> background correction has not been done, nor is there anyway to change this
> behavior within the calls. The other people and myself thought we were
> looking at raw, non-background corrected data and didn't realize that many
> spots were removed because of non-positive values after background
> subtraction. At the very least, would you add a sentence to the help files
> mentioning that background subtraction will be done by default? Being able
> to change the background correction in the call to 'plotMA' or 'plotMA3by2'
> would be nice, but as I've told people, you can do it yourself with one
> extra command:
>
> MA <- MA.RG(RG, bc.method="none")
> plotMA(MA)
>
> Thanks,
> Jenny
>
> BTW - I apologize if this has already been changed since limma 2.4.7 - I'm
> in the middle of several analyses and haven't wanted to upgrade to R 2.3.0 yet.
>
>
>
> Jenny Drnevich, Ph.D.
>
> Functional Genomics Bioinformatics Specialist
> W.M. Keck Center for Comparative and Functional Genomics
> Roy J. Carver Biotechnology Center
> University of Illinois, Urbana-Champaign
>
> 330 ERML
> 1201 W. Gregory Dr.
> Urbana, IL 61801
> USA
>
> ph: 217-244-7355
> fax: 217-265-5066
> e-mail: drnevich at uiuc.edu
>
>