[BioC] BioC normalisations for small array 2 colour data?

[Wolfgang Huber]

Hi Dan,

shameless self-promotion, you could try "vsn" because it only estimates 
4 parameters in total in your case, which should be possible with good 
enough precision from  404 data points.

If you do not expect many genes to be differentially expressed, please 
set the parameter 'lts.quantile' (it controls the degree of robustness 
or resistance of the estimator) to a higher value than the default, e.g. 
0.95. You can use the spike control to see whether the result is plausible.

Sean - I agree that normalization methods that are based on assumptions 
of invariance of 'something' between the different colors or arrays can 
(but need not) fail if a large part of genes is differentially expressed 
- but I am not following the argument why 'single-channel' methods would 
be fundamentally different in this respect.

  Best wishes
  Wolfgang



Sean Davis wrote:
> On Thursday 07 September 2006 08:44, Dan Swan wrote:
>> Hi,
>>
>> I have some data from a small specialised microarray - 200 genes, 1
>> spiked control, 1 negative control.  This is 2 colour data, with dye
>> swaps.  I was wondering what an appropriate normalisation for this
>> scenario is within Bioconductor given that Lowess is unreliable for
>> <1000 genes?
> 
> There is no "correct" answer here.  You will need to look at the data and 
> determine what needs to be done.  Scatterplots, density plots/histograms, and 
> M vs. A plots can help. 
> 
> If your genes were chosen because they were all thought to be differentially 
> expressed, then any normalization method for two-color arrays will be 
> inappropriate and you should probably think about single-channel 
> normalization.  
> 
> Sean
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor