I am reposting to the mailing list as it was bounced the first time. Sorry if you received this more than once. Best, Norman ---------- Forwarded message ---------- From: Norman Pavelka <normanpavelka@gmail.com> Date: Thu, Sep 22, 2011 at 9:07 AM Subject: Re: [BioC] plgem for spectral counts To: Amin Moghaddasi <a.moghaddasi at="" gmail.com=""> Cc: "bioconductor at stat.math.ethz.ch" <bioconductor at="" stat.math.ethz.ch=""> Dear Amin, Thank you for your interest in PLGEM! Currently the package supports only pairwise comparisons. I've always had on my to-do-list a further development to extend the functionality to multivariate analysis, but never got to do it. My lazy excuse has always been that in most experimental designs people are usually interested in knowing which genes/proteins change under which condition. For example, in a time-course experiment you want to know which genes/proteins change at what timepoint. Or if you have a set of healthy controls and various sets of patients with different diseases, you want to know which genes/proteins change in which disease. So even if you do first an F-test type of analysis to see which genes/proteins changes overall in your dataset, you would still have to do some kind of post-hoc test to see under which particular condition they change most significantly. In a sense PLGEM gets directly at the second question. Of course, from a statistical point of view, PLGEM is testing 'n' x 'm' hypothesis (where 'n' is the number of genes or proteins, and 'm' is the number of conditions) instead of just 'n', so we have a bigger multiple testing problem. That said, if you or anyone else has suggestions on how to extend the signal-to-noise-ratio statistic to the multiple-conditions case, I'd be happy to implement it in a future version of the package. Cheers, Norman P.S.: I changed the subject line as we're not talking about 'limma' anymore... :-) On Wed, Sep 21, 2011 at 5:06 AM, Amin Moghaddasi <a.moghaddasi at="" gmail.com=""> wrote: > Dear Norman / all, > > Many thanks for making PLGEM available to the community. It's a very > straightforward package to use. I'm using this for spectral counts measured > on four conditions. > As far as I understand, PLGEM can fit the model to one comparison of > interest at a time (Control_vs_condirion1, control_vs_condition2, ...) to > detect differential expression. . I'm wondering if there is anyway to > calculate p-values for multi-group comparison? Basically what I'm after is > to make all pair-wise comparisons to detect the significance of changes over > all conditions. The same as what limma does with f-statistics. > > Many thanks in advance, > Amin > > > > > > > On Sun, Oct 24, 2010 at 6:34 PM, Pavelka, Norman <nxp at="" stowers.org=""> wrote: >> >> Hi Yolande, >> >> The error message is telling you that there is no condition called 'C' in >> your ExpressionSet. In fact, if you look at your 'pData' you only have two >> conditions, either condition 'M' or 'F'. Try running it again changing the >> value of argument 'fitCondition' to either 'M' or 'F'. >> >> On a separate note, if the only thing you want to change compared to the >> default behaviour is the significance level 'delta', you don't have to use >> the step-by-step mode. You can use the wrapper mode, and simply change the >> value of argument 'signLev'. >> >> Let me know how it works. I'll be happy to help more. >> BTW, if you reply through the Bioconductor mailing list, also others can >> benefit from the discussion! ;-) >> >> Thanks! >> Norman >> >> >> -----Original Message----- >> From: Yolande Tra [mailto:yolande.tra at gmail.com] >> Sent: Saturday, October 23, 2010 1:19 PM >> To: Pavelka, Norman >> Subject: Re: [BioC] limma for spectral counts >> >> Hi Norman, >> >> Thank you for your reply. I tried the method using the step-by-step mode, >> since I want to use delta = 0.05 (not 0.001) but it did not work. Here is >> all the code I run. I built an expressionset for the data using pData1 >> (attached file). I have 4 replicates for condition C and 5 replicates for >> PLS. I used the same notation as in the tutorial. >> >> library(plgem) >> library("Biobase") >> exprs <- as.matrix(read.table("phtn102210.txt", header = TRUE, sep = "\t", >> row.names = 1, as.is = TRUE)) pData <- read.table('pData1.txt', row.names = >> 1, header = TRUE, sep = "\t") >> rownames(pData) >> ?all(rownames(pData) == colnames(exprs)) phenoData <- >> new("AnnotatedDataFrame", data = pData) exampleSet <- new("ExpressionSet", >> exprs = exprs, phenoData = phenoData) >> > exampleSet >> ExpressionSet (storageMode: lockedEnvironment) >> assayData: 865 features, 9 samples >> ?element names: exprs >> protocolData: none >> phenoData >> ?sampleNames: C1, C2, ..., LPS5 ?(9 total) >> ?varLabels and varMetadata description: >> ? ?conditionName: NA >> featureData: none >> experimentData: use 'experimentData(object)' >> Annotation: >> >> > phenoData(exampleSet) >> An object of class "AnnotatedDataFrame" >> ?sampleNames: C1, C2, ..., LPS5 ?(9 total) >> ?varLabels and varMetadata description: >> ? ?conditionName: NA >> >> It seems that the same description is outputed for your data LPSeset and >> my data exampleSet, but still gave me an error. >> >> LPSfit <- plgem.fit(data = exampleSet, covariate = 1, fitCondition = "C", >> p = 10, q = 0.5, plot.file = FALSE, fittingEval = TRUE, verbose = >> TRUE) >> Error in .checkCondition(fitCondition, "fitCondition", covariate, >> pData(data)) : condition 'C' is not defined in the input ExpressionSet for >> function 'plgem.fit'. >> >> Thank you for your help, >> Yolande >> >> On Fri, Oct 22, 2010 at 7:49 PM, Pavelka, Norman <nxp at="" stowers.org=""> wrote: >> > Hi Yolande, >> > >> > You can try normalizing your specral counts following the NSAF >> > (Normalized Spectral Abundance Factor) approach and then you can use package >> > 'plgem' to detect your differentially abundant proteins. You can have a look >> > at this publication to get an idea and then let me know if you need any >> > help: >> > >> > http://www.ncbi.nlm.nih.gov/pubmed/18029349 >> > >> > Thanks and good luck! >> > Norman >> > >> > >> > On 20 October 2010 14:20, Yolande Tra <yolande.tra at="" gmail.com=""> wrote: >> >> Hello list members, >> >> >> >> I was wondering if limma method can be used for spectral counts of >> >> proteins from mass spectrometry. If yes, is there a function in >> >> Bioconductor that normalizes these counts.before running limma. >> >> >> >> Thank you for your help, >> >> >> >> Yolande >> >> >> >> _______________________________________________ >> >> 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 >> >> >> > >> > Norman Pavelka, Ph.D. >> > Postdoctoral Research Associate >> > Rong Li lab >> > Stowers Institute for Medical Research 1000 E. 50th St. >> > Kansas City, MO 64110 >> > U.S.A. >> > >> > phone: +1 (816) 926-4103 >> > fax: +1 (816) 926-4658 >> > e-mail: nxp at stowers.org >> > >> > _______________________________________________ >> > 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 >> > >> >> _______________________________________________ >> 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 > >