Hi Gordon, As per your suggestion I have completed the analysis. The design matrix which I created for my samples (4 biological replicates for test and controls), is in the lines of the example which the limma user guide provided in page 41.The coefficients were in the expected biological line. Still I would like to test whether the data is fine or not? How can I do the same by different plots? Is it possible to obtain a differential expression value for each biological replicate of test against the controls, so that I can create a heatmap of gene expression across samples? Best Abhilash On Thu, Apr 24, 2008 at 4:34 PM, Abhilash Venu <abhivenu@gmail.com> wrote: > Thanks Gordon, > I am using 'normexp', instead of background subtraction is due to agilents > software itself does normalization (I don't know whether it is always true) > and in many times the diagnostic plots also shows the same. Please correct > me if I am wrong. > > >txt_files <- dir(pattern=".txt") >> > > >This will read files in alphabetical order. Will this agree with your > design matrix? You can >specify the order of the files using a targets > file. > In this case can you suggest a alternate method for reading the files or > creating target file? > > Reagrds, > Abhilash > > > On Thu, Apr 24, 2008 at 7:20 AM, Gordon K Smyth <smyth@wehi.edu.au> wrote: > >> >> >> On Thu, 24 Apr 2008, Abhilash Venu wrote: >> >> Hi Gordon, >>> >>> I am a experimentalist, but not with great statistical knowledge and R >>> experiance. Recently we started doing single color experiments using >>> Agilent >>> array. >>> I think that we does not have any specific function to read the single >>> channel data, But as once you suggested I used the dummy values for 'R' >>> and >>> 'Rb' and performed furhter normalization by the following commands. >>> >>>> txt_files <- dir(pattern=".txt") >>>> >>> >> This will read files in alphabetical order. Will this agree with your >> design matrix? You can specify the order of the files using a targets file. >> >> RG<-read.maimages(txt_files, columns = list(G = "gMeanSignal", Gb = >>>> >>> "gBGMeanSignal", >>> R="gMedianSignal",Rb="gBGMedianSignal"), >>> annotation= c("Row", "Col", >>> "ProbeUID","ProbeName", "GeneName",)) >>> >>>> Rgene<-backgroundCorrect(RG$G,method='normexp') >>>> >>> >> RG$G-RG$Gb would be more usual. >> >> MA<-normalizeBetweenArrays(Rgene$G,method="scale") >>>> >>> >> method="quantile" would be more usual. >> >> Is this is fine? >>> >>> In this case I have three samples which are treated and the other three >>> without treatment. I would like to get differentially expressed genes >>> between the treated and untreated. In this scenario what should be the >>> best >>> way to create a desing matrix for my further analysis. >>> >> >> Same considerations apply as for any single-channel microarray. See the >> User's Guide for Affymetrix data for example. >> >> Best wishes >> Gordon >> >> >> Is the following will >>> be fine? >>> design <- cbind(tx=c(1,1,0,0,0,ntx=c(0,0,1,1,1)) >>> Is there anything which I should specifically taken care of in these type >>> of >>> scenario? >>> >>> Thanks >>> -- >>> >>> Regards, >>> Abhilash >>> Graduate student >>> Department of Molecular biology >>> IOB >>> >> >> > > > -- > > Regards, > Abhilash -- Regards, Abhilash [[alternative HTML version deleted]]