[BioC] tilingArray normalizeByReference Data Scaling

[Wolfgang Huber]

Dear Dario

thank you for the feedback. Sorry that this is so much trouble for you. 
The method implemented in the normalizeByReference function is described 
in the "Methods" part, Section 2.3 of the article Huber W, Toedling J, 
Steinmetz, L. Transcript mapping with high-density oligonucleotide 
tiling arrays. Bioinformatics 22, 1963-1970 (2006), see 
http://www-huber.embl.de/pub/pdf/Huber-Bioinf-2006.pdf
A single method is described there.
Did you maybe refer to Section 3 of the paper, where the method is 
applied and compared to potential alternatives, or Fig.5, where some 
results of the comparison are visualised? In any case, I have added a 
more specific reference to Section 2.3 to the function's manual page.

Regarding the subtraction of the 5% quantile from the final values in 
order to make the lower range of the data be at around 0, this is 
mentioned in the description of Fig.5, but it is not described in the 
Methods part, and it is not part of the function. It is trivial enough 
that you can do it yourself if you find it useful.

So, to summarise, what the function does is described in Section 2.3, 
and applications of the function are described in other parts of that 
paper and in various other papers. If you can tell me exactly what is 
unclear I'll be happy to add a clarification to the function's manual page.

Regarding logarithms, base 2 is used throughout. I am not sure I 
understand your question, can you point out what you think the problem is?

	Best wishes
	Wolfgang

May/10/12 2:00 AM, Dario Strbenac scripsit::
> Hello,
>
> Could the description about what steps are done be made clearer ?
> Looking at the output, it seems that the data has been divided by the
> DNA values and log2 scaled after DNA reference normalisation. But
> there's no mention of this in ?normalizeByReference and the
> Bioinformatics article gives a number of different ways to normalise
> the data. I also notice that the 5% quantiles of the normalised
> results aren't scaled to 0. Is this an additional step ?
>
> Looking at the source code, I also noticed that it log2 scales the
> reference probes and natural log scales the background probes. Is
> this intentional ?
>
> Thanks, Dario.
>
> -------------------------------------- Dario Strbenac Research
> Assistant Cancer Epigenetics Garvan Institute of Medical Research
> Darlinghurst NSW 2010 Australia
>
> _______________________________________________ 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


-- 
Best wishes
	Wolfgang

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