Hello, I was wondering if anyone could offer me some advice on the best approach for normalizing my two-color expression arrays. I will be processing a large number of arrays, and ideally I would like to develop a semi-automated normalization pipeline. Some of my arrays have issues with spatial effects, and currently the only method that I'm aware of for dealing with such effects is in the Maanova package (the "rlowess" method of transform.madata). However, this method is far from ideal for my purposes because it utilizes grid layout rather than 'X' and 'Y' positions to calculate proximity (which causes some problems with gaps between blocks) and because it is coupled to a intensity-based normalization (which limits the flexibility somewhat). I have a few specific questions: [1] Are there any other methods for spatial normalization of two-color data implemented in R? [2] In my attempts to develop a normalization pipeline I have been stymied by the need to ascertain on a slide-by-slide basis which types of normalization are needed (e.g. pin/intensity/spatial). Do any of you have a "rule-of-thumb", or better yet a quantitative approach to making this decision? Thanks! Oliver Homann

Hi Olivier, Take a look at the OLIN package. It takes xy coordinates and uses a machine learning approach to approximate the smoothing parameters for spatial and intensity normalization. I have the same issue and I've had a great deal of success with it. I wouldn't recommend bypassing the slide-by-slide oversight, but you may find a set of parameters supplied to OLIN that is sufficient for the overwhelming majority of your chips. Cheers, jay On Mar 7, 2007, at 11:47 AM, Oliver Homann wrote: > Hello, > > I was wondering if anyone could offer me some advice on the best > approach for normalizing my two-color expression arrays. I will be > processing a large number of arrays, and ideally I would like to > develop > a semi-automated normalization pipeline. Some of my arrays have > issues > with spatial effects, and currently the only method that I'm aware of > for dealing with such effects is in the Maanova package (the "rlowess" > method of transform.madata). However, this method is far from > ideal for > my purposes because it utilizes grid layout rather than 'X' and 'Y' > positions to calculate proximity (which causes some problems with gaps > between blocks) and because it is coupled to a intensity-based > normalization (which limits the flexibility somewhat). > > I have a few specific questions: > [1] Are there any other methods for spatial normalization of two- color > data implemented in R? > [2] In my attempts to develop a normalization pipeline I have been > stymied by the need to ascertain on a slide-by-slide basis which types > of normalization are needed (e.g. pin/intensity/spatial). Do any > of you > have a "rule-of-thumb", or better yet a quantitative approach to > making > this decision? > > Thanks! > Oliver Homann > > _______________________________________________ > 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