[BioC] Artifact in RMA-normalized data

Rafael A. Irizarry ririzarr at jhsph.edu
Thu Mar 4 23:07:58 MET 2004


try it again using the default version of rma... quantiles instaed of
quantiles.robus

> allnorm<-expresso(alldata, bgcorrect.method='rma',
> normalize.method='quantiles', pmcorrect.method='pmonly',
> summary.method='me
> dianpolish')


On Thu, 4 Mar 2004, Mary Putt wrote:

> Hi,
> I normalized the data from 13 arrays (6 group H and 7 group P) using
> rma. I found that the arrays from my H group were systematically lower
> than from the P group at the lower end of the expression scale, while my
> arrays from the H group were higher than the arrays from the P group at
> the high end of the scale. The differences are subtle but they show up
> in the MVA plots, as well as in the summary statistics seen below. I had
> a Warning message from Affy during the normalization. It doesn't seem to
> me that RMA should introduce this type of artifact--unless there's
> something about the warning message that I don't understand. I'm
> wondering if anyone has insights on this. Thanks Mary
> 
> ########
> #Program to normalize the data
> ############
> 
> library(affy)
> load('all.Rdata')
> exprs.allnorm<-exprs(allnorm)
> allnorm<-expresso(alldata, bgcorrect.method='rma',
> normalize.method='quantiles.robus', pmcorrect.method='pmonly',
> summary.method='me
> dianpolish')
> save(exprs.allnorm, file='exprs.allnorm.Rdata')
> #########
> #Warning msg following normalization
> ############
> 
> #> source('expresso.all.r')
> #background correction: rma
> #normalization: quantiles.robus
> #PM/MM correction : pmonly
> #expression values: medianpolish
> #background correcting...done.
> #normalizing...Chip weights are  1 1 1 1 1 1 0 1 1 1 1 1 1 1
> #Chip weights are  1 1 1 1 1 1 1 1 0 1 1 1 1 1
> #done.
> #22283 ids to be processed
> #.........
> #Warning messages:
> #1: the condition has length > 1 and only the first element will be used
> in: if (remove.extreme == "variance") {
> #2: the condition has length > 1 and only the first element will be used
> in: if (remove.extreme == "mean") {
> #3: the condition has length > 1 and only the first element will be used
> in: if (remove.extreme == "both") {
> #4: the condition has length > 1 and only the first element will be used
> in: if (remove.extreme == "variance") {
> #5: the condition has length > 1 and only the first element will be used
> in: if (remove.extreme == "mean") {
> #6: the condition has length > 1 and only the first element will be used
> in: if (remove.extreme == "both") {
> 
> #####################
> #descriptive statistics of normalized data
> #
> #note that h and p are different groups
> ##################
> >
> > summary
>             h1     h2     h3     h4     h5     h6     p1     p2
> p3     p4
> Min      2.997  3.008  3.051  3.010  2.967  3.005  3.123  3.057  3.119
> 3.102
> 1stQrtl  4.719  4.679  4.762  4.739  4.771  4.771  4.895  4.717  4.926
> 4.891
> Median   5.924  5.901  5.950  5.970  5.970  5.961  6.015  5.942  6.015
> 6.018
> Mean     6.165  6.143  6.150  6.171  6.182  6.178  6.167  6.163  6.172
> 6.162
> 3rdQrtl  7.291  7.300  7.266  7.358  7.316  7.288  7.201  7.281  7.216
> 7.224
> Max     13.310 13.620 13.760 13.800 13.660 13.660 13.800 13.790 13.670
> 13.660
>             p5     p6     p7
> Min      3.121  3.017  3.041
> 1stQrtl  4.938  4.829  4.835
> Median   6.031  5.993  6.015
> Mean     6.172  6.166  6.168
> 
> >apply(summary[,1:6], 1, median)
>     Min 1stQrtl  Median    Mean 3rdQrtl     Max
>  3.0065  4.7505  5.9555  6.1680  7.2955 13.6600
> > apply(summary[,7:13], 1, median)
>     Min 1stQrtl  Median    Mean 3rdQrtl     Max
>   3.102   4.891   6.015   6.167   7.220  13.660
> > apply(summary[,1:6], 1, mean)
>       Min   1stQrtl    Median      Mean   3rdQrtl       Max
>  3.006333  4.740167  5.946000  6.164833  7.303167 13.635000
> > apply(summary[,7:13], 1, mean)
>       Min   1stQrtl    Median      Mean   3rdQrtl       Max
>  3.082857  4.861571  6.004143  6.167143  7.227143 13.602857
> 
> 
> 
> 
> --
> Mary E. Putt
> Assistant Professor of Biostatistics
> Department of Biostatistics and Epidemiology
> Center for Biostatistics and Epidemiology
> School of Medicine, University of Pennsylvania,
> 621 Blockley Hall
> 423 Guardian Drive
> Philadelphia, PA
> 19104-6021
> 
> Ph. (215) 573-7020
> Fax (215) 573-4865
> 
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor
>



More information about the Bioconductor mailing list