Question

quantile normalization of one dataset to another

0

Entering edit mode

Daniel Brewer ★ 1.9k

@daniel-brewer-1791

Last seen 10.2 years ago

Hi, I have two different datasets which are both originally from the Affymetrix platform but now I only have the expression matrix. Both datasets have an overall different distribution of expression. What I would like to do is quantile normalisation on each of the experiments in one dataset to the quantile curve of the other dataset i.e. I want to leave the second dataset untouched, but normalise the first to it. ANy ideas how to do this? I looked at normalize.quantiles but it does not appear to do what I want. Thanks Dan -- ************************************************************** Daniel Brewer, Ph.D. Institute of Cancer Research Molecular Carcinogenesis Email: daniel.brewer at icr.ac.uk ************************************************************** The Institute of Cancer Research: Royal Cancer Hospital, a charitable Company Limited by Guarantee, Registered in England under Company No. 534147 with its Registered Office at 123 Old Brompton Road, London SW7 3RP. This e-mail message is confidential and for use by the addre...{{dropped}}

Cancer Cancer • 2.3k views

ADD COMMENT • link updated 17.5 years ago by Ben Bolstad ★ 1.2k • written 17.5 years ago by Daniel Brewer ★ 1.9k

score 0 · Answer 1 · 2007-05-22

0

Entering edit mode

Benilton Carvalho ★ 4.3k

@benilton-carvalho-1375

Last seen 4.7 years ago

Brazil/Campinas/UNICAMP

Assuming that the second dataset is quantile normalized and that its expression matrix is called "expression2", what you want to do is: ref <- sort(expression2[,1]) ## if it is quantile normalized, every column has the same distribution for (i in 1:ncol(expression1)) expression1[order(expression1[,i]),i] <- ref b On May 22, 2007, at 10:17 AM, Daniel Brewer wrote: > Hi, > > I have two different datasets which are both originally from the > Affymetrix platform but now I only have the expression matrix. Both > datasets have an overall different distribution of expression. What I > would like to do is quantile normalisation on each of the > experiments in > one dataset to the quantile curve of the other dataset i.e. I want to > leave the second dataset untouched, but normalise the first to it. > ANy > ideas how to do this? I looked at normalize.quantiles but it does not > appear to do what I want. > > Thanks > > Dan -- Benilton Carvalho PhD Candidate Department of Biostatistics Bloomberg School of Public Health Johns Hopkins University bcarvalh at jhsph.edu

ADD COMMENT • link 17.5 years ago Benilton Carvalho ★ 4.3k

0

Entering edit mode

Just found another complicating factor. Even though both sets are from Affymetrix arrays they are from different types and so have different numbers of probes. Therefore quantile normalisation may not be the way forward. Are there any other cross array normalisation techniques which would be suitable, whilst still retaining that one dataset is left untouched? Dan Benilton Carvalho wrote: > Assuming that the second dataset is quantile normalized and that its > expression matrix is called "expression2", what you want to do is: > > ref <- sort(expression2[,1]) ## if it is quantile normalized, every > column has the same distribution > for (i in 1:ncol(expression1)) > expression1[order(expression1[,i]),i] <- ref > > b > > On May 22, 2007, at 10:17 AM, Daniel Brewer wrote: > >> Hi, >> >> I have two different datasets which are both originally from the >> Affymetrix platform but now I only have the expression matrix. Both >> datasets have an overall different distribution of expression. What I >> would like to do is quantile normalisation on each of the experiments in >> one dataset to the quantile curve of the other dataset i.e. I want to >> leave the second dataset untouched, but normalise the first to it. ANy >> ideas how to do this? I looked at normalize.quantiles but it does not >> appear to do what I want. >> >> Thanks >> >> Dan > > > > -- > Benilton Carvalho > PhD Candidate > Department of Biostatistics > Bloomberg School of Public Health > Johns Hopkins University > bcarvalh at jhsph.edu > > -- ************************************************************** Daniel Brewer, Ph.D. Institute of Cancer Research Molecular Carcinogenesis MUCRC 15 Cotswold Road Sutton, Surrey SM2 5NG United Kingdom Tel: +44 (0) 20 8722 4109 Fax: +44 (0) 20 8722 4141 Email: daniel.brewer at icr.ac.uk ************************************************************** The Institute of Cancer Research: Royal Cancer Hospital, a charitable Company Limited by Guarantee, Registered in England under Company No. 534147 with its Registered Office at 123 Old Brompton Road, London SW7 3RP. This e-mail message is confidential and for use by the addre...{{dropped}}

ADD REPLY • link 17.5 years ago Daniel Brewer ★ 1.9k

0

Entering edit mode

Hi Daniel! Even if you had managed to apply a quantile normalization I doubt that it would have really solved your problem. Cominbining micorarray results from different studies (even if the same platform is used) is a tricky thing. A recent publication on this is <http: www.biomedcentral.com="" 1471-2105="" 8="" 80=""> Bayesian meta-analysis models for microarray data: a comparative study Erin M Conlon, Joon J Song, Anna Liu BMC Bioinformatics 2007, 8:80 Hope this helps Claus Daniel Brewer wrote: > Just found another complicating factor. Even though both sets are from > Affymetrix arrays they are from different types and so have different > numbers of probes. Therefore quantile normalisation may not be the way > forward. Are there any other cross array normalisation techniques which > would be suitable, whilst still retaining that one dataset is left > untouched? > > Dan > > Benilton Carvalho wrote: > >> Assuming that the second dataset is quantile normalized and that its >> expression matrix is called "expression2", what you want to do is: >> >> ref <- sort(expression2[,1]) ## if it is quantile normalized, every >> column has the same distribution >> for (i in 1:ncol(expression1)) >> expression1[order(expression1[,i]),i] <- ref >> >> b >> >> On May 22, 2007, at 10:17 AM, Daniel Brewer wrote: >> >> >>> Hi, >>> >>> I have two different datasets which are both originally from the >>> Affymetrix platform but now I only have the expression matrix. Both >>> datasets have an overall different distribution of expression. What I >>> would like to do is quantile normalisation on each of the experiments in >>> one dataset to the quantile curve of the other dataset i.e. I want to >>> leave the second dataset untouched, but normalise the first to it. ANy >>> ideas how to do this? I looked at normalize.quantiles but it does not >>> appear to do what I want. >>> >>> Thanks >>> >>> Dan >>> >> >> -- >> Benilton Carvalho >> PhD Candidate >> Department of Biostatistics >> Bloomberg School of Public Health >> Johns Hopkins University >> bcarvalh at jhsph.edu >> >> >> > > -- ********************************************************************** ************* Dr Claus-D. Mayer | http://www.bioss.ac.uk Biomathematics & Statistics Scotland | email: claus at bioss.ac.uk Rowett Research Institute | Telephone: +44 (0) 1224 716652 Aberdeen AB21 9SB, Scotland, UK. | Fax: +44 (0) 1224 715349

ADD REPLY • link 17.5 years ago Claus Mayer ▴ 340

score 0 · Answer 2 · 2007-05-22

Ignoring the wisdom, or otherwise, of seeking to do a normalization like this there is code for doing this using affyPLM. The following code implements a quantile normalization procedure which determines a target distribution from one dataset and applies it to another: library(affyPLM) X.to.determine.normalization <- matrix(rnorm(1000),100,10) Y.to.normalize <- matrix(rexp(10000),1000,10) target <- normalize.quantiles.determine.target(X.to.determine.normaliz ation,1000) Y.normalized <- normalize.quantiles.use.target(Y.to.normalize,target) boxplot(as.data.frame(X.to.determine.normalization)) boxplot(as.data.frame(Y.to.normalize)) boxplot(as.data.frame(Y.normalized)) On Tue, 2007-05-22 at 15:17 +0100, Daniel Brewer wrote: > Hi, > > I have two different datasets which are both originally from the > Affymetrix platform but now I only have the expression matrix. Both > datasets have an overall different distribution of expression. What I > would like to do is quantile normalisation on each of the experiments in > one dataset to the quantile curve of the other dataset i.e. I want to > leave the second dataset untouched, but normalise the first to it. ANy > ideas how to do this? I looked at normalize.quantiles but it does not > appear to do what I want. > > Thanks > > Dan >