[BioC] right normalization?

[Wolfgang Huber]

Dear Andrea

I see three options, in order of complexity:
- add a 'batch' factor into the linear model that you fit with limma in 
order to absorb the batch effect
- do the limma analysis separately for both batches, then combine afterwards
- use the 'sva' package.
http://www.biostat.jhsph.edu/~jleek/sva/index.html

	Best wishes
	Wolfgang


Sep/26/11 10:29 AM, andrea.grilli at ior.it scripsit::
> Hi list,
> I've got 16 gene chip affymetrix arrays of 2 cell lines at different
> time/conditions. Chips were done in 2 different moments, and this bias
> seems to affect my analysis.
>
> I've normalized with RMA and analyzed with eBayes models to compare
> the 2 cell lines at each time point.
> Following analysis are affected by the above mentioned "time bias",
> like in clustering were data are grouped according to when the arrays
> were hybridized.
>
> - Do you suggest a different normalization? Which one could reduce
> this effect?
>
> - could be a good approach normalizing the 2 groups of samples
> separately (will be 8 and 8) and merging the data after normalization?
>
>
> I did quality control steps like in bioconductor book, and all
> parameters are in the right ranges. Also box plot shows a similar
> expression range across samples.
> Any help will be really appreciated,
> thanks in advance,
>
> Andrea
>
>
>
> PS This is a semplification of a previous mail, sorry for the repetition.
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives:
> http://news.gmane.org/gmane.science.biology.informatics.conductor


-- 


Wolfgang Huber
EMBL
http://www.embl.de/research/units/genome_biology/huber