Hi Jim, Ok, thank you very much for your input. Makes much more sense to filter after variability estimation. Thanks again for your suggestion. Jose El dic 20, 2012, a las 11:18 p.m., James W. MacDonald escribi?: > Hi Jose, > > On 12/20/2012 4:55 PM, Jos? L?pez wrote: >> Hi Jim, >> >> Thank you for your quick reply and your useful explanation. >> In fact, I was using the function to subset the eset since I didn?t >> know other way to do it (before). >> >> By the way, I would like to know your opinion about whether is a >> good option to filter controls before doing moderate t-statistics >> (limma) or should I rather perform statistic on the whole variable >> dataset (including controls)? > > My understanding is that you should filter after fitting the model > when using limma, because you will bias the empirical Bayes estimate > if you filter first. So that is what I do with these arrays. > >> >> I have read in that in contrast to standard t-statistic, common >> filter/test pairs does not necessarily translate in power gains >> when moderated t-statistics is performed (Bourgon, Gentleman and >> Huber, 2009), so following their indications, I don?t use to apply >> any filter on genes. The question is whether you think it is a good >> practice to remove control probes (not genes) or not. > > I filter the control probes, mainly because they have a bad habit of > popping up in a list of differentially expressed genes. This was > much rarer with the 3' biased arrays, and way more obvious since > those controls had a big AFFX appended to the probeset ID. > > What I generally see is that the intronic controls often appear to > be differentially expressed. I can come up with several hypotheses > as to why this is so (the primary one being that total mRNA will > likely also include mRNA that hasn't yet been processed to remove > the introns, so if one sample is more actively expressing a gene, > you may well end up with introns being processed into cDNA and then > hybridized to the chip). > > Regardless, these are supposed to be controls, and are not really > annotated, and it is hard to explain when they pop up in lists of > differentially expressed genes. So I take the easy way out and nuke > them right after fitting the model. > > Best, > > Jim > > > > >> >> Sorry if the question is not clearly exposed. >> >> Thank you in advance for your time and your help. >> >> Best, >> >> Jose >> >> El dic 20, 2012, a las 7:00 p.m., James W. MacDonald escribi?: >> >>> Hi Jose, >>> >>> On 12/20/2012 12:20 PM, Jos? L?PEZ wrote: >>>> Dear all, >>>> >>>> I was trying to use preProcessGeneST ArrayTools to get rid of >>>> control probes in Mouse Gene 1.0ST arrays, but it dosent work in >>>> last R/BioC version. >>>> It was working perfectly in previous R/BioC version. Do I >>>> downgrade to previous version to continue to use the ArrayTools >>>> package? >>>> It is possible that the error has a different cause? >>> >>> This looks like a bug in the current version of the >>> mogene10sttranscriptcluster.db package, as the MAP object appears >>> to be missing: >>> >>> > ls(2) >>> [1] "mogene10sttranscriptcluster" >>> <snip> >>> [20] "mogene10sttranscriptclusterGO2ALLPROBES" >>> [21] "mogene10sttranscriptclusterGO2PROBE" >>> [22] "mogene10sttranscriptclusterMAPCOUNTS" >>> [23] "mogene10sttranscriptclusterMGI" >>> <snip> >>> >>> So I don't think downgrading anything will help - we just need to >>> rebuild this package. >>> >>> But this brings me to a different question. The function you are >>> using is intended to annotate things and then output in the >>> current directory, and removing control probes is just a side >>> effect of one argument. So are you trying to annotate, or to >>> remove control probes? >>> >>> If you just want to remove control probes, note that you can do >>> >>> > data(mogene10stCONTROL) >>> >>> and then you can subset your eset using the resulting data.frame: >>> >>> eset_no_control <- eset_norm[!featureNames(eset_norm) %in% >>> mogene10stCONTROL$probeset_id,] >>> >>> Note the use of the bang (!) preceding featureNames - we want to >>> remove these things, not select for them. >>> >>> Best, >>> >>> Jim >>> >>> >>>> >>>> Thank you for your kind help, >>>> >>>>> eset_process = preProcessGeneST(eset_norm, output = TRUE) >>>> Warning message: >>>> In chkPkgs(chip) : >>>> The mogene10sttranscriptcluster.db package does not appear to >>>> contain annotation data. >>>> Error in function (x, envir, mode = "any", ifnotfound = >>>> list(function(x) stop(paste0("value for '", : >>>> error in evaluating the argument 'envir' in selecting a method >>>> for function 'mget': Error: object >>>> 'mogene10sttranscriptclusterMAP' not found >>>>> sessionInfo() >>>> R version 2.15.2 (2012-10-26) >>>> Platform: i386-apple-darwin9.8.0/i386 (32-bit) >>>> >>>> locale: >>>> [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8 >>>> >>>> attached base packages: >>>> [1] stats graphics grDevices utils datasets methods base >>>> >>>> other attached packages: >>>> [1] mogene10stv1cdf_2.11.0 annaffy_1.30.0 KEGG.db_2.8.0 >>>> [4] GO.db_2.8.0 arrayQualityMetrics_3.14.0 ArrayTools_1.18.0 >>>> [7] mogene10sttranscriptcluster.db_8.0.1 org.Mm.eg.db_2.8.0 >>>> RSQLite_0.11.2 >>>> [10] DBI_0.2-5 affy_1.36.0 annotate_1.36.0 >>>> [13] AnnotationDbi_1.20.3 vsn_3.26.0 Biobase_2.18.0 >>>> [16] BiocGenerics_0.4.0 limma_3.14.3 >>>> >>>> loaded via a namespace (and not attached): >>>> [1] affyio_1.26.0 affyPLM_1.34.0 beadarray_2.8.1 >>>> BeadDataPackR_1.10.0 BiocInstaller_1.8.3 >>>> [6] Biostrings_2.26.2 Cairo_1.5-2 cluster_1.14.3 colorspace_1.2-0 >>>> gcrma_2.30.0 >>>> [11] genefilter_1.40.0 grid_2.15.2 Hmisc_3.10-1 hwriter_1.3 >>>> IRanges_1.16.4 >>>> [16] lattice_0.20-10 latticeExtra_0.6-24 parallel_2.15.2 plyr_1.8 >>>> preprocessCore_1.20.0 >>>> [21] RColorBrewer_1.0-5 reshape2_1.2.2 setRNG_2011.11-2 >>>> splines_2.15.2 stats4_2.15.2 >>>> [26] stringr_0.6.2 survival_2.37-2 SVGAnnotation_0.93-1 >>>> tools_2.15.2 XML_3.95-0.1 >>>> [31] xtable_1.7-0 zlibbioc_1.4.0 >>>>> class(H2Bgfp_norm) >>>> [1] "ExpressionSet" >>>> attr(,"package") >>>> [1] "Biobase" >>>> >>>> _______________________________________________ >>>> Bioconductor mailing list >>>> Bioconductor at r-project.org >>>> https://stat.ethz.ch/mailman/listinfo/bioconductor >>>> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor >>> >>> -- >>> James W. MacDonald, M.S. >>> Biostatistician >>> University of Washington >>> Environmental and Occupational Health Sciences >>> 4225 Roosevelt Way NE, # 100 >>> Seattle WA 98105-6099 >>> >> > > -- > James W. MacDonald, M.S. > Biostatistician > University of Washington > Environmental and Occupational Health Sciences > 4225 Roosevelt Way NE, # 100 > Seattle WA 98105-6099 >