[BioC] Limma: background correction. Use or ignore?

[J.delasHeras at ed.ac.uk]

Quoting Gordon K Smyth <smyth at wehi.EDU.AU>:

> For each channel on each array, plot log-foreground vs log-background.  E.g.,
>
>   plot(log2(RG$Rb[,1]),log2(RG$R[,1]),pch=".")
>   abline(0,1,col="blue")
>
> If background is worth removing, you should see an increasing trend.
>
> This is not dye bias.
>
> Gordon

Unfortunately I am not sure what I am looking at here...
I made the plots (I can attach the two jpegs if you want, I am not sure 
what the rules of the list are about attachments).

I chose two slides (dye swap) for this, from a simple experiment 
comparing two cell lines. They were not the best examples I've had, but 
I think they are reasonably typical of what I can expect and have been 
observing. The data was obtained with Genepix after scanning on an Axon 
4200AL. The background was estimated using Genepix's default: local.

I can see that my background is slightly higher for my green channel, 
centered at around 9, while the red channel bkg is centered at around 
8.5. The range is narrow, about 0.5 in both cases.
I see a slightly denser area at the bottom, separated from the main 
body of the spots. This area takes roughly the shape of a line, 
parallel to the blue abline.
Then the bottom of the main body is also denser and follows a similar 
pattern, also parallel.

I am not sure what to deduce from this, regarding whether it's worth 
bkg-substracting or not...

any pointers based on your experience... much appreciated!

Jose

-- 
Dr. Jose I. de las Heras                      Email: J.delasHeras at ed.ac.uk
The Wellcome Trust Centre for Cell Biology    Phone: +44 (0)131 6513374
Institute for Cell & Molecular Biology        Fax:   +44 (0)131 6507360
Swann Building, Mayfield Road
University of Edinburgh
Edinburgh EH9 3JR
UK