[BioC] RMA/QuantileNormalization results difference between oligo and aroma.affymetrix for Hugene
Benilton Carvalho
beniltoncarvalho at gmail.com
Fri Feb 26 12:05:08 CET 2010
Quantile normalization is already one step in the RMA workflow.
Therefore, there's no need to normalize the data again once you've
gone RMA, ie. (regarding oligo) your call
"normalize.quantiles(exprs(rmadata))" should be dropped.
Using the defaults, rma() in oligo will:
1) Background correct (via the RMA convolution model)
2) Quantile normalize
3) Summarize via median-polish.
b
On Fri, Feb 26, 2010 at 10:46 AM, Mikhail Pachkov <pachkov at gmail.com> wrote:
> Dear All,
>
> I am new in microarray analysis and need your expertise.
> I need to develop procedure for producing expression values from CEL
> files. Data should processed with RMA and quantile normalized. I have
> tried two packages - oligo and aroma.affymetrix. Obtained results are
> quite different form my point of view. Moreover
> aroma.affymetrix::QuantileNormalization function produce dta which do
> not look like they were quantile normalized.
> I have made density plots of data after RMA and after quantile
> normalization which are available here
> http://test.swissregulon.unibas.ch/bioc/index.html There are also
> links to two CEL files I have used.
>
> I have a few questions:
> Why RMA results are so different?
> Which RMA implementation is correct?
> Why does quantile normalization in aroma.affymetrics produce two
> different distributions?
>
> Thank you in advance!
>
> Here are R scripts I have used:
>
> ################################
> #aroma.affymetrix
> library(aroma.affymetrix);
> verbose <- Arguments$getVerbose(-8, timestamp=TRUE);
>
> # read files
> cdf <- AffymetrixCdfFile('annotationData/chipTypes/HuGene-1_0-st-v1/HuGene-1_0-st-v1.cdf');
> cs <- AffymetrixCelSet$byPath("rawData/mine/HuGene-1_0-st-v1/");
>
> # RMA
> bc <- RmaBackgroundCorrection(cs);
> csBC <- process(bc,verbose=verbose);
>
> # QuantileNormalization
> qn <- QuantileNormalization(csBC, typesToUpdate="pm");
> csN <- process(qn);
>
> # Plots
> image_file <- ("aroma.affymetrix.RMA.png");
> png(image_file,width=1028,height=768);
> plotDensity(csBC);
> title("aroma.affymetrix RMA data");
> dev.off();
>
> image_file <- ("aroma.affymetrix.QN.png");
> png(image_file,width=1028,height=768);
> plotDensity(csN);
> title("aroma.affymetrix QN data");
> dev.off()
> ################################
>
> ################################
> # oligo
> library(oligo);
> rawdata=read.celfiles(c("rawData/mine/HuGene-1_0-st-v1/sample1.CEL","rawData/mine/HuGene-1_0-st-v1/sample2.CEL"));
> rmadata=rma(rawdata);
> qndata=normalize.quantiles(exprs(rmadata))
>
> library(affy)
> # Plots
> image_file <- ("oligo.RMA.png");
> png(image_file,width=1028,height=768);
> plotDensity(exprs(rmadata));
> title("oligo RMA data");
> dev.off();
>
> image_file <- ("oligo.QN.png");
> png(image_file,width=1028,height=768);
> plotDensity(qndata);
> title("oligo QN data");
> dev.off()
> ###############################
>
> Kind regards,
>
> Mikhail Pachkov
>
> _______________________________________________
> 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
>
More information about the Bioconductor
mailing list