Hi all, I was checking DESeq normalization success for my samples following the suggestions in this post : https://stat.ethz.ch/pipermail/bioconductor/2010-October/035933.html. I observed that filtering down to the genes that are present in both case and control samples improved normalization  (visualized by plotting as per the post), but for a few samples ,say  3 samples these procedures did not help, and I tried calculating shorth estimator (again as per the post) and this did n't help either. My question is what should I follow for the normalization of these samples ? Regards, Sudeep. [[alternative HTML version deleted]]

Hi Sudeep On 2012-08-01 11:47, sudeep s wrote: > I was checking DESeq normalization success for my samples following > the suggestions in this post : > https://stat.ethz.ch/pipermail/bioconductor/2010-October/035933.html. > I observed that filtering down to the genes that are present in both > case and control samples improved normalization (visualized by > plotting as per the post), but for a few samples ,say 3 samples > these procedures did not help, and I tried calculating shorth > estimator (again as per the post) and this did n't help either. My > question is what should I follow for the normalization of these > samples ? You would need to tell us what exactly you did not like about the histograms. Anyway, I revised my opinion a bit, and I now think that MA plots are mroe helpful. Starting as in the post that you cited, plot an MA plot instead of a histogram, with plot( geomeans, log2( counts(cds)[,j] / geomeans ), log="x", pch=".", col="#00000060" ) abline( h=log2( sizeFactors(cds)[ j ] ), col="red" ) Simon

Hi Simon, Thank you for the reply As you said in the post, I was plotting to see if the size factor was hitting the median, in some samples (3), the size factor was way off the median and hence I tried shorth, but this didn't help either, and I tried filtering down the genes (filtering genes on row sums for 10,15,20, 25...) but again in the plot size factor was way off, so I was wondering what I should follow. regards, Sudeep. ________________________________ From: Simon Anders <anders@embl.de> To: bioconductor@r-project.org Sent: Tuesday, 7 August 2012 12:59 PM Subject: Re: [BioC] DESeq normalization Hi Sudeep On 2012-08-01 11:47, sudeep s wrote: > I was checking DESeq normalization success for my samples following > the suggestions in this post : > https://stat.ethz.ch/pipermail/bioconductor/2010-October/035933.html. > I observed that filtering down to the genes that are present in both > case and control samples improved normalization  (visualized by > plotting as per the post), but for a few samples ,say  3 samples > these procedures did not help, and I tried calculating shorth > estimator (again as per the post) and this did n't help either. My > question is what should I follow for the normalization of these > samples ? You would need to tell us what exactly you did not like about the histograms. Anyway, I revised my opinion a bit, and I now think that MA plots are mroe helpful. Starting as in the post that you cited, plot an MA plot instead of a histogram, with plot( geomeans, log2( counts(cds)[,j] / geomeans ), log="x",   pch=".", col="#00000060" ) abline( h=log2( sizeFactors(cds)[ j ] ), col="red" )   Simon _______________________________________________ Bioconductor mailing list Bioconductor@r-project.org https://stat.ethz.ch/mailman/listinfo/bioconductor Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor [[alternative HTML version deleted]]