Hi,Bioconductor team, > I'm recently doing same work about RNA-seq DGE, I used bowtie2 to align > the reads to reference trancriptome and used eXpress to generate > the read counts, and the data has no biological replicates, the R > command line is as follows, the padjs I get only have 1 and NA, where > did I do wrong? > >countTable<-read.table("eXpresscountmatrix", header=T, row.names=1) > >condition<-factor(c("A1","A2","A3")) > > library( "DESeq" ) > > cds = newCountDataSet( countTable, condition ) > > cds = estimateSizeFactors( cds ) > > cds = estimateDispersions( cds, method="blind", sharingMode="fit- only" ) > > res_A1vsA2 = nbinomTest( cds, "A1", "A2" ) > > res_A1vsA3 = nbinomTest( cds, "A1", "A3" ) > > res_A2vsA3 = nbinomTest( cds, "A2", "A3" ) > > resSig_A1vsA2 = res_A1vsA2[ res_A1vsA2$padj < 0.1, ] > > resSig_A1vsA3 = res_A1vsA3[ res_A1vsA3$padj < 0.1, ] > > resSig_A2vsA3 = res_A2vsA3[ res_A2vsA3$padj < 0.1, ] > Thanks in advance! > Tao 2012-11-20 David Tao [[alternative HTML version deleted]]

It looks like you don't have any biological replicates, so getting no significant differentially expressed genes is probably a valid result. Biological replicates are important for proper dispersion estimates, and without them, your only option is to use estimateDispersions with method="blind" and sharingMode="fit-only", which will greatly limit your power to detect differential expression. See the DESeq manual, section 3.3: "Hence, any attempt to work without replicates will lead to conclusions of very limited reliability. Nevertheless, such experiments are sometimes undertaken, and the DESeq package can deal with them, even though the soundness of the results may depend much on the circumstances." I don't know if there's much you can do to salvage the situation here. Perhaps if two of the three conditions are expected to be similar to each other, with the third one being more different, you can estimate dispersions using factor(c("A1", "A1", "A3")) as your condition vector. Other than that, I'm afraid you may simply be out of luck for detecting and differential expression from these samples. The standard advice would be "do more replicates". Hope this helps, -Ryan Thompson On Mon 19 Nov 2012 04:47:42 PM PST, David Tao wrote: > Hi,Bioconductor team, >> I'm recently doing same work about RNA-seq DGE, I used bowtie2 to align >> the reads to reference trancriptome and used eXpress to generate >> the read counts, and the data has no biological replicates, the R >> command line is as follows, the padjs I get only have 1 and NA, where >> did I do wrong? >> >countTable<-read.table("eXpresscountmatrix", header=T, row.names=1) >> >condition<-factor(c("A1","A2","A3")) >> > library( "DESeq" ) >> > cds = newCountDataSet( countTable, condition ) >> > cds = estimateSizeFactors( cds ) >> > cds = estimateDispersions( cds, method="blind", sharingMode ="fit-only" ) >> > res_A1vsA2 = nbinomTest( cds, "A1", "A2" ) >> > res_A1vsA3 = nbinomTest( cds, "A1", "A3" ) >> > res_A2vsA3 = nbinomTest( cds, "A2", "A3" ) >> > resSig_A1vsA2 = res_A1vsA2[ res_A1vsA2$padj < 0.1, ] >> > resSig_A1vsA3 = res_A1vsA3[ res_A1vsA3$padj < 0.1, ] >> > resSig_A2vsA3 = res_A2vsA3[ res_A2vsA3$padj < 0.1, ] >> Thanks in advance! >> Tao > > > 2012-11-20 > > > > David Tao > [[alternative HTML version deleted]] > > _______________________________________________ > 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