On Tue, Mar 6, 2012 at 3:28 PM, Wolfgang Huber <whuber at="" embl.de=""> wrote: > > Hi Julian, Kasper > > in this case you will need to (and as I am sure, be able to) come up with > another method for estimating size factors that works on these data. Why not > try with the 'colSums'? Hey, I am just reporting. It might be worthwhile adding some check in estimateSizeFactorsForMatrix to at least warn about the case where the number of non-infinity rows is small, say less than 100. This could be an issue with data types where the number of features is small, eg. small RNAs or other types. I am sure people are using DESeq for these types as well. Kasper > > ? ? ? ?Best wishes > ? ? ? ?Wolfgang > > Mar/6/12 3:25 PM, Kasper Daniel Hansen scripsit:: > >> I got the data from Julian. ?It is not a standard RNA-seq experiment. >> He only has 150 "genes" (rows), but 76 samples (columns). ?There are >> no NA's. ?The problem arises from the fact that >> estimateSizeFactorsForMatrix only uses rows (genes) where all samples >> have counts> ?0. ?In Julian's case, all rows have at least one column >> with zero counts, implying that loggeomeans is -Inf for all rows. >> >> Kasper >> >> On Mon, Mar 5, 2012 at 10:59 AM, Wolfgang Huber<whuber at="" embl.de=""> ?wrote: >>> >>> Dear Julian >>> >>> do your data contain NA? >>> What is the output of sessionInfo()? >>> >>> Can you provide a code example (incl. input data, potentially subset) >>> that >>> reproduces your error? >>> >>> ? ? ? ?Best wishes >>> ? ? ? ?Wolfgang >>> >>> >>> Mar/5/12 11:50 AM, Julian [guest] scripsit:: >>> >>>> >>>> I'm using deseq with 454 data and it worked for one set of data but the >>>> same script is failing me the second time around with a different set of >>>> experimental data. >>>> >>>> The input data is a matrix of counts of 454 seqs per sample, I have 36 >>>> pre >>>> and 36 post samples. >>>> >>>> When I run the estmateSizeFactors I get all my samples as NA. >>>> >>>> Any ideas why? >>>> >>>> ?-- output of sessionInfo(): >>>> >>>> >>>>> cds<- estimateSizeFactors(cds) ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? # >>>>> Estimates >>>>> size factors based on the count data >>>>> sizeFactors( cds ) >>>> >>>> >>>> ? X01_MA_1 ?X02_MA_10 X03_MA_100 X04_MA_102 ?X05_MA_11 ?X06_MA_13 >>>> ?X07_MA_14 ?X08_MA_15 ?X09_MA_17 ?X10_MA_18 ?X11_MA_19 >>>> ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> ? X12_MA_2 ?X13_MA_20 ?X14_MA_22 ?X15_MA_23 ?X16_MA_24 ?X17_MA_25 >>>> X18_MA_4 ?X19_MA_47 ? X20_MA_5 ?X21_MA_69 ? X22_MA_7 >>>> ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> ?X23_MA_71 ?X24_MA_73 ?X25_MA_75 ?X26_MA_77 ?X27_MA_79 ? X28_MA_8 >>>> ?X29_MA_81 ?X30_MA_83 ?X31_MA_86 ?X32_MA_88 ? X33_MA_9 >>>> ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> ?X34_MA_90 ?X35_MA_92 ?X36_MA_94 ?X37_MA_96 ?X38_MA_98 X39_MA_101 >>>> X40_MA_103 ?X41_MA_26 ?X42_MA_27 ?X43_MA_29 ?X44_MA_30 >>>> ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> ?X45_MA_31 ?X46_MA_33 ?X47_MA_34 ?X48_MA_36 ?X49_MA_37 ?X50_MA_40 >>>> ?X51_MA_41 ?X52_MA_42 ?X53_MA_43 ?X54_MA_44 ?X55_MA_45 >>>> ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> ?X56_MA_46 ?X57_MA_49 ?X58_MA_50 ?X59_MA_52 ?X60_MA_54 ?X61_MA_55 >>>> ?X62_MA_70 ?X63_MA_72 ?X64_MA_74 ?X65_MA_76 ?X66_MA_78 >>>> ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> ?X67_MA_80 ?X68_MA_82 ?X69_MA_84 ?X70_MA_87 ?X71_MA_89 ?X72_MA_91 >>>> ?X73_MA_93 ?X74_MA_95 ?X75_MA_97 ?X76_MA_99 >>>> ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> NA ? ? ? ? NA ? ? ? ? NA ? ? ? ? NA >>>> >>>> >>>> -- >>>> Sent via the guest posting facility at bioconductor.org. >>>> >>>> _______________________________________________ >>>> 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 >>> >>> >>> >>> >>> -- >>> Best wishes >>> ? ? ? ?Wolfgang >>> >>> Wolfgang Huber >>> EMBL >>> http://www.embl.de/research/units/genome_biology/huber >>> >>> >>> _______________________________________________ >>> 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 > > > > -- > Best wishes > ? ? ? ?Wolfgang > > Wolfgang Huber > EMBL > http://www.embl.de/research/units/genome_biology/huber > >

Hi, On Tue, Mar 6, 2012 at 3:28 PM, Wolfgang Huber <whuber at="" embl.de=""> wrote: > > Hi Julian, Kasper > > in this case you will need to (and as I am sure, be able to) come up with > another method for estimating size factors that works on these data. Why not > try with the 'colSums'? Couldn't we just-as-well adapt the estimateSizesFactorsForMatrix function to step over the (row,col) bins that have the 0 counts instead of skipping over rows that only have 1 0 element? Something like so: ######### esfm <- function(counts, locfunc=median) { ## Can't use info from rows that are all 0, so axe them counts <- counts[rowSums(counts) > 0,] logc <- log(counts) loggeomeans <- apply(logc, 1, function(x) { use <- is.finite(x) sum(x[use]) / sum(use) }) apply(logc, 2, function(lc) { use <- is.finite(lc) exp(locfunc(lc[use] - loggeomeans[use])) }) } ########## ?mas o menos? -steve -- Steve Lianoglou Graduate Student: Computational Systems Biology ?| Memorial Sloan-Kettering Cancer Center ?| Weill Medical College of Cornell University Contact Info: http://cbio.mskcc.org/~lianos/contact