[BioC] illumina normalisation using beadarray package

[Matt Ritchie]

Dear Yogi,

We haven't looked in any detail at the effect different normalisations have
on Illumina expression data.  There are a few published studies which choose
quantile normalisation (see reference list below) and this is the approach
we often apply.

You may also want to log2 transform your data before normalising it, i.e.

BSData.quantile = normaliseIllumina(BSData, method = "quantile",
transform="log2")

Of course, it is always advisable to check that quantile is sensible for the
data you're analysing with density plots etc.

Best wishes,

Matt

*********
Papers which use quantile normalisation in the analysis of Illumina
expression data:

Elvidge GP, Glenny L, Appelhoff RJ, Ratcliffe PJ et al. Concordant
regulation of gene expression by hypoxia and 2-oxoglutarate-dependent
dioxygenase inhibition: the role of HIF-1alpha, HIF-2alpha, and other
pathways. J Biol Chem 2006 Jun 2;281(22):15215-26. PMID: 16565084

Golubkov VS, Chekanov AV, Savinov AY, Rozanov DV et al. Membrane type-1
matrix metalloproteinase confers aneuploidy and tumorigenicity on mammary
epithelial cells. Cancer Res 2006 Nov 1;66(21):10460-5. PMID: 17079467

Barnes M, Freudenberg J, Thompson S, Aronow B, Pavlidis P. Experimental
comparison and cross-validation of the Affymetrix and Illumina gene
expression analysis platforms. Nucleic Acids Res. 2005 Oct
19;33(18):5914-23. PMID: 16237126

> Hi all
> 
>  
> 
> I‚m quite new to illumina analysis. I read non-normalised data and did a
> quantile normalisation.
> 
> library(beadarray)
> 
> BSData <- readBeadSummaryData(targets=NULL, header=T, sep="\t",path=
> NULL,
> 
>                                    columns = list(ProbeID = "TargetID",
> 
>                                      AvgSig = "AVG_Signal", Nobeads =
> "Avg_NBEADS",
> 
>                                      Detection="Detection",
> BeadStDev="BEAD_STDEV"),
> 
>                                    other.columns = NULL, skip=7)
> 
> BSData.quantile = normaliseIllumina(BSData, method = "quantile")
> 
> normalised <- exprs(BSData.quantile)
> 
> write.csv(file="normalised.csv", normalised)
> 
>  
> 
> Is quantile normalisation good enough? What normalisation methods are
> considered good for illumina data? Your feedback will be much
> appreciated. 
> 
>  
> 
> Thanks heaps
> 
> Yogi