[BioC] help in normalization cDNA arrays

J.delasHeras at ed.ac.uk J.delasHeras at ed.ac.uk
Wed Jan 24 17:16:46 CET 2007


If you're going to be looking at the ratios only, normalising between  
arrays is rarely going to change anything (as the method aims to scale  
the range of intensities while leaving the ratios relatively unchanged).

You can easily do a quick test and compare the results you get if you  
don't normalise between arrays, and using different methods (scale and  
Aquantile, for instance). When I try, I don't see much difference at  
all. I try to avoid altering the data (normalising procedures alter  
the data) unless I have a good reason to do so.

On teh other hand, if you want to do some sort of single channel  
analysis, where you compare signal from one slide to signal from  
another, then you do need to make sure the slides are all in the same  
scale, and normalising between arrays makes sense.

Also, as someone else said, you should not normalise between arrays  
experiments that by their own nature will behave very differently.

I think that when we start doing arrays we tend to think  
"we-must-normalise... we-must-normalise" (imagine that with a Dalek  
voice ;-) and that anything that uses teh word "normalisation" is a  
good thing. However, it all depends on the biology of our experiment.  
Try to understand what your experiments are about and what is  
expected, and then think how it would make sense to normalise the  
data. Then see which of the methods available to you do what is  
necessary for your data. I find it useful to explore teh raw data  
(scatter plots, MA plots) and see what the trends are... and when I  
decide to normaise in a particular way, copmpare the effect of the  
normalisation to see if it did what I (roughly) wanted it to do. This  
allows you to build up a good feel for what the limitations are with  
any method, and you truly get to know your data.

Jose



Quoting Marcelo Laia <marcelolaia at gmail.com>:

> Hi,
>
> I did a
>
> normalizeWithinArrays(RG,method="printtiploess")
>
> and get this boxplot
>
> http://200.145.102.65:9102/normalized.data.png
>
> Is needed to do a
>
> normalizeBetweenArrays(MA.2,method="scale")
>
> ???
>
> What you suggest me?
>
> I am not sure if a little difference  between box is big or not.
>
> Thank you
>
> --
> Marcelo Luiz de Laia
> Ph.D Candidate
> São Paulo State University (http://www.unesp.br/eng/)
> School of Agricultural and Veterinary Sciences
> Department of  Technology
> Via de Acesso Prof. Paulo Donato Castellane s/n
> 14884-900   Jaboticabal - SP - Brazil
> Phone: +55-016-3209-2675
> Cell: +55-016-97098526
>
> _______________________________________________
> 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



-- 
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



More information about the Bioconductor mailing list