All, A collaborator has approached me with a situation in which a batch is completely confounded with treatment for affy gene expression data. I am wondering if there is anything reasonable to try? For example, I'm wondering if anything can be done with control genes. Does anyone have any suggestions, pointers to papers, or a template script? The key feature is, of course, the complete confounding, so that it seems Combat or SVA are not applicable (correct me if this is incorrect). I am looking for a way to use the data to generate hypotheses and make some guesses with these limitations. Thanks, Juliet

Juliet, You could use fRMA, which down-weights probes that are particular susceptible to batch-effects, to preprocess the data. Paper: McCall MN, Bolstad BM, and Irizarry RA* (2010). Frozen Robust Multi-Array Analysis (fRMA), Biostatistics, 11(2):242-253. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2830579/ BioC package vignette: http://bioconductor.org/packages/devel/bioc/vignettes/frma/inst/doc/fr ma.pdf Best, Matt On Mon, Aug 13, 2012 at 2:46 PM, Juliet Hannah <juliet.hannah at="" gmail.com=""> wrote: > All, > > A collaborator has approached me with a situation in which a batch is > completely confounded with treatment for > affy gene expression data. > > I am wondering if there is anything reasonable to try? For example, > I'm wondering if anything can be done with > control genes. > > Does anyone have any suggestions, pointers to papers, or a template > script? The key feature is, of course, the complete > confounding, so that it seems Combat or SVA are not applicable > (correct me if this is incorrect). > > I am looking for a way to use the data to generate hypotheses and make > some guesses > with these limitations. > > Thanks, > > Juliet > > _______________________________________________ > 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 -- Matthew N McCall, PhD 112 Arvine Heights Rochester, NY 14611 Cell: 202-222-5880

Juliet, Two thoughts and ideas for you: First off, another new alternative to fRMA is SCAN, which is now in press at the journal Genomics. SCAN also works on single samples and it works as well as fRMA without the need to design/use frozen vectors for your particular array platform. Software is available at: http://jlab.bu.edu/software/scan-upc/ Also, if you are still worried about batch effects, I have already solved the batch/confounding problem. I now have a ComBat script that can adjust for Batch Effects in confounded designs. However, I have not published this work yet. I'm working on the manuscript now and hope to have it published within a few months, therefore: a. You can wait for the publication and then use the ComBat for confounded variables. b. Alternatively, I can help you adjust your data using my beta version of my confounded ComBat, but I have a policy to not release my unpublished methods unless you are willing to include me as a collaborator on your manuscript (assuming that the confounded ComBat works well on your data). If you don't want to wait, send me a personal email and we can discuss options for collaboration and for getting your confounded data adjusted. Hope this helps. Thanks! Evan -- W. Evan Johnson, PhD Assistant Professor of Medicine Division of Computational Biomedicine Boston University Department of Medicine 72 East Concord St., E-645 Boston, MA 02118 [[alternative HTML version deleted]]