We have done affy microarrays for wildtype-control treatment, wildtype-toxin treatment, mutant-control treatment and mutant-toxin treatment. The goal is to find diffferentially expressed genes regulated by toxin in wildtype and then find out which of these regulation are mutant dependent. The first goal is typical. So we did R/ bioconductor - SAM and limma. Both could give us a bunch of DEGs. However, I am lost about getting the second aim done. With limma and 2x2 factorial analysis, we could find the DEGs from all kind of pairs, such as wildtype -Toxin/control (this answers the first goal) or mutant/wildtype in control treatment (this tells us how the mutant gene is affecting the basal expression without toxin). But I don't know how to find the wildtype DEGs which have regulation depending on the mutant gene. Say one gene is up-regulated 100 folds by toxin (i.e. toxin treatment/control treatment) in wildtype. Then if this gene is up-regulated 3 folds (toxin treatment/control treatment) in mutant, it will be apparently mutant gene dependent (from 100 folds in wildtype to 3 folds in mutant). However, this gene will be shown as DEG in the mutant analysis as well since it is more than the 2-fold cutoff. Then if I only compared the DEGs in different pair of analysis from limma, I will miss these kind of genes. There must be some dedicated way to analyze this but I could not find it. Any suggestion will be appreciated. Thanks a lot.

Kao, You make a great point about the problem of comparing lists of differentially expressed genes (DEGs) under various conditions. The lists overlap to a certain extent, but it is hard to get a feeling for what these overlaps mean. Certainly, the fact that a gene fails to meet a certain level of significance in some test does not imply that it is not differentially expressed under your experimental conditions. My suggestion is that you visualize various gene sets (say, genes primarily affected by toxin) across all your treatments at once, A heatmap (particularly if your gene sets are around 50-100 genes) is generally very informative. Typically, you will find that mutant status alters the toxic effect in interesting ways depending on the gene: some expression differences will be amplified, some curtailed, etc. A consideration of the genes affected in this way, that is, the roles they play biologically, should yield insight. In fact, you could begin with a heatmap of genes most affected by your conditions in general -- say the 200 genes with most significant by ANOVA. Best Tom On May 25, 2009, at 11:05 AM, Cheng-Yuan Kao wrote: > We have done affy microarrays for wildtype-control treatment, > wildtype-toxin treatment, mutant-control treatment and mutant-toxin > treatment. > > The goal is to find diffferentially expressed genes regulated by toxin > in wildtype and then find out which of these regulation are mutant > dependent. > > > The first goal is typical. So we did R/ bioconductor - SAM and limma. > Both could give us a bunch of DEGs. However, I am lost about getting > the second aim done. > With limma and 2x2 factorial analysis, we could find the DEGs from all > kind of pairs, such as wildtype -Toxin/control (this answers the first > goal) or mutant/wildtype in control treatment (this tells us how the > mutant gene is affecting the basal expression without toxin). > But I don't know how to find the wildtype DEGs which have regulation > depending on the mutant gene. Say one gene is up-regulated 100 folds > by toxin (i.e. toxin treatment/control treatment) in wildtype. Then if > this gene is up-regulated 3 folds (toxin treatment/control treatment) > in mutant, it will be apparently mutant gene dependent (from 100 folds > in wildtype to 3 folds in mutant). However, this gene will be shown as > DEG in the mutant analysis as well since it is more than the 2-fold > cutoff. Then if I only compared the DEGs in different pair of analysis > from limma, I will miss these kind of genes. There must be some > dedicated way to analyze this but I could not find it. Any suggestion > will be appreciated. Thanks a lot. > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/ > gmane.science.biology.informatics.conductor

Hi, In section "8.7 Factorial Designs" of Limma Users Guide, there is a similar situation: 1. which genes respond to stimulation in wild-type cells, 2. which genes respond to stimulation in mutant cells, and 3. which genes respond differently in mutant compared to wild-type cells. The contrast matrix is set up accordingly. Following the same logic, in the current situation, how about such a contrast matrix: > cont.matrix <- makeContrasts( + TOXinWT=WT.TOXIN-WT.CON, + Diff=(MUT.TOXIN-MUT.CON)-(WT.TOXIN-WT.CON), + levels=design) The first one (TOXinWT) is for the effect of toxin in wild-type, while the latter (Diff) is for the genes that show different effect of toxin in mutant (compared to that in wild-type)? Best wishes, Saroj Cheng-Yuan Kao wrote: > We have done affy microarrays for wildtype-control treatment, > wildtype-toxin treatment, mutant-control treatment and mutant-toxin > treatment. > > The goal is to find diffferentially expressed genes regulated by toxin > in wildtype and then find out which of these regulation are mutant > dependent. > > > The first goal is typical. So we did R/ bioconductor - SAM and limma. > Both could give us a bunch of DEGs. However, I am lost about getting > the second aim done. > With limma and 2x2 factorial analysis, we could find the DEGs from all > kind of pairs, such as wildtype -Toxin/control (this answers the first > goal) or mutant/wildtype in control treatment (this tells us how the > mutant gene is affecting the basal expression without toxin). > But I don't know how to find the wildtype DEGs which have regulation > depending on the mutant gene. Say one gene is up-regulated 100 folds > by toxin (i.e. toxin treatment/control treatment) in wildtype. Then if > this gene is up-regulated 3 folds (toxin treatment/control treatment) > in mutant, it will be apparently mutant gene dependent (from 100 folds > in wildtype to 3 folds in mutant). However, this gene will be shown as > DEG in the mutant analysis as well since it is more than the 2-fold > cutoff. Then if I only compared the DEGs in different pair of analysis > from limma, I will miss these kind of genes. There must be some > dedicated way to analyze this but I could not find it. Any suggestion > will be appreciated. Thanks a lot. > > _______________________________________________ > Bioconductor mailing list > Bioconductor at stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor > >