Dear Zhe, On Thu, July 1, 2010 1:58 pm, ?? wrote: > Dear Gordon, > > I tryed plotMDS.dge function and got good results. But I have two questions: > > 1. What do "Dimension 1" and "Dimension 2" represent, respectively? Think of it as if you were arranging the samples on the floor so that distance apart represents the degree of differential expression between each two samples. Dimension 1 is simply the direction in which the sample are most different, and dimension 2 is the next direction. The dimensions don't have any simple interpretation in terms of metagenes etc. A unit of one on the axis represents a coefficient of variation of 1 between the samples. > 2. I have 12 different samples which were collected from 3 different positions of an organ from 2 > species at 2 stages (3*2*2=12 samples). I did RNA-seq and sequenced them in 12 lanes, > respectively. I want to see the similarity of the 12 samples by clustering and analyze DE between > different species and between different species. Can I separate the 12 samples to 2 groups by the > stage or by the species? I'm not sure if I can consider the samples I grouped as "replicates" and > use edgeR to do these tasks. Just look at the plot and see where the samples fall. Do samples of the same stage tend to group together? Do samples of the same species tend to group together? If they do, then the answer to your question is yes. Best wishes Gordon > Thanks, > Zhe > > > > Dear Zhe, > > To do clustering of RNA-seq profiles using the edgeR packages, you can use the plotMDS.dge > function. See the User's Guide for examples. This function is already designed for RNA-seq data, > so there is no need to worry about normalization factors or variance stabilizing transformations > etc. > > Best wishes > Gordon > > > [BioC] edgeR normalization factors > zhedianyou at yahoo.cn > Mon Jun 28 05:19:19 CEST 2010 > > Hello, > > I have a question about using TMM normalization factors. I want to modify the count for each gene > after normalization. Should I just need to divide the count of each gene by the normalization > factor for its library? Then, I may use the normalized data for DE analysis and other further > analysis (e.g. clustering). > > Thanks a lot, > Zhe ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:4}}

Dear Gordon,  Thank you very much. Your answers help me a lot. Best wishes, Zhe Dear Zhe, On Thu, July 1, 2010 1:58 pm, ?? wrote: > Dear Gordon, > > I tryed plotMDS.dge function and got good results. But I have two questions: > > 1. What do "Dimension 1" and "Dimension 2" represent, respectively? Think of it as if you were arranging the samples on the floor so that distance apart represents the degree of differential expression between each two samples. Dimension 1 is simply the direction in which the sample are most different, and dimension 2 is the next direction.  The dimensions don't have any simple interpretation in terms of metagenes etc.  A unit of one on the axis represents a coefficient of variation of 1 between the samples. > 2. I have 12 different samples which were collected from 3 different positions of an organ from 2 > species at 2 stages (3*2*2=12 samples). I did RNA-seq and sequenced them in 12 lanes, > respectively. I want to see the similarity of the 12 samples by clustering and analyze DE between > different species and between different species. Can I separate the 12 samples to 2 groups by the > stage or by the species? I'm not sure if I can consider the samples I grouped as "replicates" and > use edgeR to do these tasks. Just look at the plot and see where the samples fall.  Do samples of the same stage tend to group together?  Do samples of the same species tend to group together? If they do, then the answer to your question is yes. Best wishes Gordon > Thanks, > Zhe > > > > Dear Zhe, > > To do clustering of RNA-seq profiles using the edgeR packages, you can use the plotMDS.dge > function.  See the User's Guide for examples.  This function is already designed for RNA-seq data, > so there is no need to worry about normalization factors or variance stabilizing transformations > etc. > > Best wishes > Gordon > > > [BioC] edgeR normalization factors > zhedianyou at yahoo.cn > Mon Jun 28 05:19:19 CEST 2010 > > Hello, > > I have a question about using TMM normalization factors. I want to modify the count for each gene > after normalization. Should I just need to divide the count of each gene by the normalization > factor for its library? Then, I may use the normalized data for DE analysis and other further > analysis (e.g. clustering). > > Thanks a lot, > Zhe ______________________________________________________________________ The information in this email is confidential and intend...{{dropped:12}}