[BioC] loess and limma

[Sean Davis]

On Thursday 14 September 2006 09:48, you wrote:
> Sean
>
> Yes, that does help. Thank you.
>
> This worries me as people routinely slap in a loess normalisation. But
> from what you are saying this is not likely to be valid unless you have
> an independent reference. i.e. pooled or from another source.

It worries me, also.  One of the fantastic aspects of bioconductor is that it 
brings VERY powerful tools to even the novice user, so folks can go down the 
wrong path easily if details like assumptions used for normalization are not 
checked.  In the case of normalization of two-color data, this is a 
relatively simple process, at least in principle.  Make scatterplots and MA 
plots and check the quality of the data.  There really isn't a good 
substitute for doing this basic step.  If your data look like they have a 
correlation (a non-zero slope on an MA plot), that is either real biology or 
not.  If you use loess, you are basically saying that you are assuming that 
it is not real biology.  However, you can see that if you never look at the 
data, you will never know.

> This means if you are using your reference channel as in a "Direct
> Two-Color Design" (limma user guide Chapter 7) there is likely to be a
> correlation so you should not use loess?

I don't think a "Direct two-color design" is an absolute contraindication to 
using loess, but I may not be understanding your point here.

I just got Wolfgang's email and I agree completely.  When using ANY 
normalization method, one needs to understand the parameters, the underlying 
assumptions, and the data (by looking at plots and quality metrics).  

Sean