Normalization of 2 color Nimblegen array

0

Entering edit mode

Khan, Sohail ▴ 490

@khan-sohail-1137

Last seen 9.6 years ago

Dear list, I have a two color Niblegen array which has a slight spatial effect (higher Cy5 in upper left corner). I have applied loess normalization, but it doesn't correct this 'bias". I can't really apply printtiploess. Could VSN or quantile normaliztion be applied here or could suggest a method to normalize this data? Thanks. -Sohail

vsn vsn • 1.3k views

ADD COMMENT • link updated 18.1 years ago by Gustavo Henrique Esteves ▴ 60 • written 18.1 years ago by Khan, Sohail ▴ 490

0

Entering edit mode

Sean Davis 21k

@sean-davis-490

Last seen 3 months ago

United States

On 3/16/06 7:53 AM, "Khan, Sohail" <khan at="" cshl.edu=""> wrote: > Dear list, > > I have a two color Niblegen array which has a slight spatial effect (higher > Cy5 in upper left corner). I have applied loess normalization, but it doesn't > correct this 'bias". I can't really apply printtiploess. Could VSN or > quantile normaliztion be applied here or could suggest a method to normalize > this data? Sohail, A couple of questions, first. What is on the array (what design)? Second, what kind of experiment (chIP/chip, CGH, expression)? In general, we have found nimblegen arrays to be pretty robust against spatial artifacts, simply because one is so often able to use information from neighboring probes, so assuming the probes are randomly distributed on the slide relative to genomic position, there shouldn't be a big problem in a percentage of the slide is missing/bad. Finally, if you think of these spots as "bad", you can set the weights for those specific probes to zero and use limma to do the normalization, but the actual normalization you want to do is going to depend on the questions above. Sean

ADD COMMENT • link 18.1 years ago Sean Davis 21k

0

Entering edit mode

Gustavo Henrique Esteves ▴ 60

@gustavo-henrique-esteves-937

Last seen 8.8 years ago

Brazil

An embedded and charset-unspecified text was scrubbed... Name: not available Url: https://stat.ethz.ch/pipermail/bioconductor/attachments/20060316/ d1d750c1/attachment.pl

ADD COMMENT • link 18.1 years ago Gustavo Henrique Esteves ▴ 60

0

Entering edit mode

On 3/16/06 9:36 AM, "Gustavo H. Esteves" <gesteves at="" gmail.com=""> wrote: > Dear Khan, > > do you know package OLIN from Bioconductor? Have a look at this package. > > It does the loess normalization with a step of estimation of the best span > parameters and then does another loess regression for the x and y location > in the chip. > > The main reference for this normalization method is: > Futschik, M. & Crompton, T. Model selection and efficiency testing for > normalization of cDNA microarray data Genome Biology, 2004, 5, R60 > > Good Luck. Thanks, Gustavo, for the pointer--I learned something! Just to be a bit more explicit on my earlier email, keep in mind here that loess normalization on an M vs. A plot (MA plot) assumes that there is no intensity-dependent signal (ie., that the log-ratios should be centered around 0). In fact, for many nimblegen applications, there IS intensity-dependent signal (ie., one expects higher-intensity spots to have a higher log ratio). In fact, for the "best" arrays, one would expect an MA plot to have a positive slope if fit with a regression line. Therefore, if one uses loess normalization (in the MA plot sense), one is definitely normalizing out the signal! For expression arrays, the usual assumptions apply and this isn't an issue--loess away! That is why it is important to know the application before choosing a normalization method. Sean

ADD REPLY • link 18.1 years ago Sean Davis 21k

0

Entering edit mode

Hi! I just wanted to point out that the oligo package supports nimblegen arrays (you need the xys and ndf) files. It is under development but we need testers. If anybody wants a specific method or function incorporated you can email benitlon or me. Most functions in the affy package can be easily exproted. -r On Thu, 16 Mar 2006, Sean Davis wrote: > > > > On 3/16/06 9:36 AM, "Gustavo H. Esteves" <gesteves at="" gmail.com=""> wrote: > >> Dear Khan, >> >> do you know package OLIN from Bioconductor? Have a look at this package. >> >> It does the loess normalization with a step of estimation of the best span >> parameters and then does another loess regression for the x and y location >> in the chip. >> >> The main reference for this normalization method is: >> Futschik, M. & Crompton, T. Model selection and efficiency testing for >> normalization of cDNA microarray data Genome Biology, 2004, 5, R60 >> >> Good Luck. > > Thanks, Gustavo, for the pointer--I learned something! > > Just to be a bit more explicit on my earlier email, keep in mind here that > loess normalization on an M vs. A plot (MA plot) assumes that there is no > intensity-dependent signal (ie., that the log-ratios should be centered > around 0). In fact, for many nimblegen applications, there IS > intensity-dependent signal (ie., one expects higher-intensity spots to have > a higher log ratio). In fact, for the "best" arrays, one would expect an MA > plot to have a positive slope if fit with a regression line. Therefore, if > one uses loess normalization (in the MA plot sense), one is definitely > normalizing out the signal! For expression arrays, the usual assumptions > apply and this isn't an issue--loess away! That is why it is important to > know the application before choosing a normalization method. > > Sean > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor >

ADD REPLY • link 18.1 years ago Rafael A. Irizarry ★ 2.3k

Login before adding your answer.