Hello, I've a question regarding the removeBatchEffect function in limma. I basically have a model expression_signal ~ organ + species + organ:species in which species can be considered a batch effect (I expect differences between species due to mainly due to array platform differences). I'd like to derive an expression signal matrix that's "corrected" for the species effect (but not for the organ effect). I've come across two problems with removeBatchEffect that I'm trying to get my head around: 1) basically there could be a species specific organ effect, i.e. an interaction which I'd like to keep in my data. 2) my data may contain some NAs which probably won't work for removeBatchEffect (and functions called within) Do you know of any alternatives that can deal with NAs (maybe impute values?), and would it be possible and does it make sense at all to consider an interaction as an experimental factor for which one part is "batch" and the other isn't? thanks for your thoughts, Arne [[alternative HTML version deleted]]

Hi Arne, On 7/31/2013 3:24 AM, Mueller, Arne wrote: > Hello, > > I've a question regarding the removeBatchEffect function in limma. I basically have a model expression_signal ~ organ + species + organ:species in which species can be considered a batch effect (I expect differences between species due to mainly due to array platform differences). I'd like to derive an expression signal matrix that's "corrected" for the species effect (but not for the organ effect). From ?removeBatchEffect: Details This function is useful for removing batch effects, associated with hybridization time or other tech- nical variables, prior to clustering or unsupervised analysis such as PCA or heatmaps. It is not intended to use with linear modelling. For linear modelling, it is better to include the batch factors in the linear model. So as the help says, fit the batch effects as part of the linear model and construct whatever interactions you want. Best, Jim > > I've come across two problems with removeBatchEffect that I'm trying to get my head around: > > 1) basically there could be a species specific organ effect, i.e. an interaction which I'd like to keep in my data. > > 2) my data may contain some NAs which probably won't work for removeBatchEffect (and functions called within) > > Do you know of any alternatives that can deal with NAs (maybe impute values?), and would it be possible and does it make sense at all to consider an interaction as an experimental factor for which one part is "batch" and the other isn't? > > thanks for your thoughts, > > Arne > > > [[alternative HTML version deleted]] > > _______________________________________________ > Bioconductor mailing list > Bioconductor at r-project.org > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor -- James W. MacDonald, M.S. Biostatistician University of Washington Environmental and Occupational Health Sciences 4225 Roosevelt Way NE, # 100 Seattle WA 98105-6099