Hi Dave, I assume you have only plotted PC1 vs PC2, you can do the same type of plots PC1 vs PC3, PC1 vs PC4 and so on...., to see if any PC captures the grouping by treatment This is regardless of how much variance each PC explains. I usually don't use not DESeq to do the PCA plots, so I am not sure how you will do this within DESeq I do understand that you are "characterizing" a response to a drug, but your underlying assumption is that part of that response is differences in gene expression that can be observed at the time point your are measuring. It could simply be that the differences between being drug resistant and sensitive have nothing to do with gene expression differences at the steady state, and that's why you don't get any significant p-values. Positive controls assure you that there are differences you can measure. Have you plotted the p-value distribution? you can find how to do it in the Nature protocols tutorial: http://www.nature.com/nprot/journal/v8/n9/full/nprot.2013.099.html Lucia On Fri, Jun 27, 2014 at 11:41 AM, Dave Wettmann <david.wettmann@gmail.com> wrote: > Thanks Lucia; I've checked that PCs 1 and 2 capture 55% and 15% of the > total variance, respectively. Could you explain how, if I did find that > the treatment effect was present in another PC, that would help me please? > I don't have any positive control because it's an experiment to > characterise a response to a drug treatment. > Thanks, > Dave > > > On 27 June 2014 15:06, Lucia Peixoto <luciap@iscb.org> wrote: > >> Hi Dave, >> >> If in your PCA your samples do not cluster by treatment, you likely have >> some sort of unwanted variation or batch effect masking the effect of the >> treatment in your data. I am not sure more samples will help. >> Have you taken a look at the PC loadings past 1 and 2 to see if there is >> any PC that captures your treatment? do you have any positive controls? are >> you sure your treatment actually causes measurable differences in gene >> expression? >> >> The only think I believe will help is RUVSeq: >> >> http://www.bioconductor.org/packages/devel/bioc/html/RUVSeq.html >> >> Lucia >> >> >> On Fri, Jun 27, 2014 at 9:27 AM, Dave Wettmann [guest] < >> guest@bioconductor.org> wrote: >> >>> Hello, >>> >>> I have constructed the following dataset for analysis using DESeq2: >>> >>> class: DESeqDataSet >>> dim: 57396 10 >>> exptData(0): >>> assays(1): counts >>> rownames(57396): ENSG00000223972 ENSG00000227232 ... ENSG00000210195 >>> ENSG00000210196 >>> rowData metadata column names(0): >>> colnames(10): 1 2 ... 10 11 >>> colData names(1): condition >>> >>> >>> > colData(ddsHTSeq) >>> DataFrame with 10 rows and 1 column >>> condition >>> <factor> >>> 1 na >>> 2 na >>> 3 Resistant >>> 4 na >>> 5 Resistant >>> 6 Resistant >>> 7 na >>> 8 na >>> 10 Sensitive >>> 11 Sensitive >>> >>> I am interested in the differential expression between the drug >>> resistant and sensitive samples ('na' are control samples). >>> I've clustered the samples and plotted a PCA as described in the >>> vignette. However, in each of these plots the samples do not cluster by >>> their drug sensitivity but are distributed across the plot. I don't have >>> any more information about the samples with which to model any potential >>> covariates. >>> I was wondering if there were any pointers as to how I could extract >>> some useful meanings from these data please? As might be expected, when I >>> try a DESeq on these data I get no significant p-values. >>> >>> Thanks in advance, >>> Dave >>> >>> -- output of sessionInfo(): >>> >>> R version 3.1.0 (2014-04-10) >>> Platform: x86_64-unknown-linux-gnu (64-bit) >>> >>> locale: >>> [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C >>> [3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8 >>> [5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8 >>> [7] LC_PAPER=en_US.UTF-8 LC_NAME=C >>> [9] LC_ADDRESS=C LC_TELEPHONE=C >>> [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C >>> >>> attached base packages: >>> [1] parallel stats graphics grDevices utils datasets methods >>> [8] base >>> >>> other attached packages: >>> [1] pasilla_0.4.0 matrixStats_0.8.14 gplots_2.13.0 >>> [4] vsn_3.32.0 Biobase_2.24.0 DESeq2_1.4.5 >>> [7] RcppArmadillo_0.4.300.0 Rcpp_0.11.1 GenomicRanges_1.16.3 >>> [10] GenomeInfoDb_1.0.2 IRanges_1.22.7 BiocGenerics_0.10.0 >>> >>> loaded via a namespace (and not attached): >>> [1] affy_1.42.2 affyio_1.32.0 annotate_1.42.0 >>> [4] AnnotationDbi_1.26.0 BiocInstaller_1.14.2 bitops_1.0-6 >>> [7] caTools_1.17 DBI_0.2-7 DESeq_1.16.0 >>> [10] gdata_2.13.3 genefilter_1.46.1 geneplotter_1.42.0 >>> [13] grid_3.1.0 gtools_3.4.0 KernSmooth_2.23-12 >>> [16] lattice_0.20-29 limma_3.20.4 locfit_1.5-9.1 >>> [19] preprocessCore_1.26.1 RColorBrewer_1.0-5 R.methodsS3_1.6.1 >>> [22] RSQLite_0.11.4 splines_3.1.0 stats4_3.1.0 >>> [25] survival_2.37-7 tcltk_3.1.0 tools_3.1.0 >>> [28] XML_3.98-1.1 xtable_1.7-3 XVector_0.4.0 >>> [31] zlibbioc_1.10.0 >>> >>> >>> -- >>> Sent via the guest posting facility at bioconductor.org. >>> >>> _______________________________________________ >>> Bioconductor mailing list >>> Bioconductor@r-project.org >>> https://stat.ethz.ch/mailman/listinfo/bioconductor >>> Search the archives: >>> http://news.gmane.org/gmane.science.biology.informatics.conductor >>> >> >> >> >> -- >> Lucia Peixoto PhD >> Postdoctoral Research Fellow >> Laboratory of Dr. Ted Abel >> Department of Biology >> School of Arts and Sciences >> University of Pennsylvania >> >> "Think boldly, don't be afraid of making mistakes, don't miss small >> details, keep your eyes open, and be modest in everything except your >> aims." >> Albert Szent-Gyorgyi >> > > -- Lucia Peixoto PhD Postdoctoral Research Fellow Laboratory of Dr. Ted Abel Department of Biology School of Arts and Sciences University of Pennsylvania "Think boldly, don't be afraid of making mistakes, don't miss small details, keep your eyes open, and be modest in everything except your aims." Albert Szent-Gyorgyi [[alternative HTML version deleted]]