Hi Francesco, If I understand your question correctly: Normalization removes background, probe and sample effects. Often batch effects are often unaddressed by normalization. In contrast, most batch adjustment methods do not address probe and background effects--so both tools are needed for some datasets. For normalization (for Affy arrays) we strongly recommend either SCAN or fRMA for normalization. In terms of distortion, yes there is a chance that the signal could be a little distorted by normalization and/or batch adjustment. Yes, a normalization method that addresses batch at the same time might be a little less likely to cause signal distortion. However in practice, a two-step normalization/batch adjustment works just fine. Hope this helps. Evan On Oct 10, 2013, at 5:17 AM, francesco.brundu@gmail.com<mailto:francesco.brundu@gmail.com> wrote: Hi all, I have a question about running combat on microarray data. In http://biostatistics.oxfordjournals.org/content/8/1/118.full it is said that (3.1) 'We assume that the data have been normalized' before batch correction. My question is, what is intended here as normalization? Does not normalization before batch correction distort the signal? Thanks ~Francesco Brundu [[alternative HTML version deleted]]

Hi all, I have a question about running combat on microarray data. In http://biostatistics.oxfordjournals.org/content/8/1/118.full it is said that (3.1) 'We assume that the data have been normalized' before batch correction. My question is, what is intended here as normalization? Does not normalization before batch correction distort the signal? Thanks ~Francesco Brundu [[alternative HTML version deleted]]

Thanks Evan, I am working on Illumina arrays and, if possible, I'll try also SCAN and fRMA as you recommend. I used to think that normalization has to be done after batch effects removal, in order not to flatten the signal. I use loess normalization after ComBat for this particular reason. >From your experience the order in which normalization and batch effects removal are done is not important? Thanks ~Francesco Brundu On 10 October 2013 16:57, Johnson, William Evan <wej@bu.edu> wrote: > Hi Francesco, > > If I understand your question correctly: Normalization removes > background, probe and sample effects. Often batch effects are often > unaddressed by normalization. In contrast, most batch adjustment methods do > not address probe and background effects--so both tools are needed for some > datasets. For normalization (for Affy arrays) we strongly recommend either > SCAN or fRMA for normalization. > > In terms of distortion, yes there is a chance that the signal could be a > little distorted by normalization and/or batch adjustment. Yes, a > normalization method that addresses batch at the same time might be a > little less likely to cause signal distortion. However in practice, a > two-step normalization/batch adjustment works just fine. > > Hope this helps. > > Evan > > > > On Oct 10, 2013, at 5:17 AM, francesco.brundu@gmail.com wrote: > > Hi all, > I have a question about running combat on microarray data. > In http://biostatistics.oxfordjournals.org/content/8/1/118.full it is > said that (3.1) 'We assume that the data have been normalized' before > batch correction. > My question is, what is intended here as normalization? Does not > normalization before batch correction distort the signal? > Thanks > > ~Francesco Brundu > > > [[alternative HTML version deleted]]