Dear Evan, I ran Combat on my log2 transformed quantile-normalized data (44 samples, Illumina HT12) and a density plot of the Combat corrected data showed that a second quantile normalization might be in order. The peaks were slightly unaligned. Does that sound like a good idea? Chris [[alternative HTML version deleted]]

Hey Chris, My preliminary thought is that its NOT a good idea. Are your batches balanced WRT treatment/control? How far are they off? Can you send me a plot? Evan On Nov 3, 2012, at 7:00 AM, bioconductor-request@r-project.org wrote: > Message: 8 > Date: Thu, 1 Nov 2012 14:35:48 -0400 > From: chris briggs <cebriggs7135@gmail.com> > To: bioconductor@r-project.org > Subject: [BioC] ComBat - pre and post Combat quantile normalization? > Message-ID: > <camuhhe_0r+n4fdpxv+a45c_8nk1rlxqhu89t6irc8fqz3jegfa@mail.gmail.com> > Content-Type: text/plain > > Dear Evan, > I ran Combat on my log2 transformed quantile-normalized data (44 samples, > Illumina HT12) and a density plot of the Combat corrected data showed that > a second quantile normalization might be in order. The peaks were slightly > unaligned. Does that sound like a good idea? > Chris [[alternative HTML version deleted]]

Chris, Looking at your plots: Don't re-quantile normalize your data. It turns out that QQ norm can actually introduce artifacts into the data on occasion. Your plots look okay to me, so I strongly recommend that you use the data 'as is' after ComBat. Evan On Nov 5, 2012, at 11:12 AM, chris briggs wrote: > I have 3 different tumor categories (Cap, Ece and R1) each with an equal number of -T (tumor derived mRNA) and -N (adjacent normal-tissue derived mRNA) for each tumor category. Each array is a batch, as shown in the attached Combat Sample Table. Does it look correct? I've also attached the density plot for my Combat corrected data. Recall that I used log2 transformed, quant. normalized data for the Combat correction. > > Thanks, > Chris > > > > On Sat, Nov 3, 2012 at 4:05 PM, W. Evan Johnson <wej@bu.edu> wrote: > Hey Chris, > > My preliminary thought is that its NOT a good idea. > > Are your batches balanced WRT treatment/control? How far are they off? Can you send me a plot? > > Evan > > > On Nov 3, 2012, at 7:00 AM, bioconductor-request@r-project.org wrote: > >> Message: 8 >> Date: Thu, 1 Nov 2012 14:35:48 -0400 >> From: chris briggs <cebriggs7135@gmail.com> >> To: bioconductor@r-project.org >> Subject: [BioC] ComBat - pre and post Combat quantile normalization? >> Message-ID: >> <camuhhe_0r+n4fdpxv+a45c_8nk1rlxqhu89t6irc8fqz3jegfa@mail.gmail.com> >> Content-Type: text/plain >> >> >> Dear Evan, >> I ran Combat on my log2 transformed quantile-normalized data (44 samples, >> Illumina HT12) and a density plot of the Combat corrected data showed that >> a second quantile normalization might be in order. The peaks were slightly >> unaligned. Does that sound like a good idea? >> Chris > > > > > -- > cell phone: 617 905-9301 > office phone: 617 779-6535 > FAX: 617 779-6572 > office emails: christine.briggs@steward.org > > > > <cmbat_sampletable_103012.txt><densityplot_lg2qncombat_values.pdf> [[alternative HTML version deleted]]