[Bioc-devel] Build error for DESeq2 dev causing error in other package builds
On 04/03/2015 04:37 AM, Andrea Rau wrote:
Hello,
I am the maintainer of a Bioconductor package called 'HTSFilter' that
imports DESeq2. On today's build report, I see that my package (as well
as DESeq2 and all of the other packages that import it) is showing an
error message which seems to arise from a recent change in the DESeq2
code dealing with the dimnames of a DESeqDataSet object. The error
message in my package reads:
Error in `assays<-`(`*tmp*`, value = <S4 object of class structure("SimpleList", package = "S4Vectors")>) :
'dimnames' differ between assay elements
I broke assays<-,SummarizedExperiment,*-method in GenomicRanges 1.19.50. it has
been fixed and problems should clear up in the next build report. The
(incorrect) error about 'dimnames differ between assay elements', seen in
./DESeq2/zin2-buildsrc.html
./BiocGenerics/zin2-checksrc.html
./HTSFilter/zin2-buildsrc.html
./VariantAnnotation/zin2-checksrc.html
./bsseq/zin2-buildsrc.html
./AllelicImbalance/zin2-buildsrc.html
./TCC/zin2-buildsrc.html
./gage/zin2-buildsrc.html
./ReportingTools/zin2-checksrc.html
./MLSeq/zin2-buildsrc.html
./systemPipeR/zin2-buildsrc.html
./DiffBind/zin2-buildsrc.html
./SGSeq/zin2-buildsrc.html
./VariantFiltering/zin2-buildsrc.html
./rgsepd/zin2-buildsrc.html
will not appear.
Off-list, Michael Love asked me what exactly does get checked. The row- and
column names of each incoming assay need either to be NULL, or to match the
current row and column names of the SummarizedExperiment. For instance the
following
> example(SummarizedExperiment)
## ...
> assays(se2) = assays(se2) # incoming dimnames
> assays(se2) = assays(se2, withDimnames=FALSE) # no incoming dimnames
> m = assays(se2)[["counts"]]; rownames(m) = NULL
> assays(se2)[["counts"]] = m # no rownames
> assays(se2)[[1]] = assays(se2)[[1]][, c(2:6, 1)] # out-of-order columns
Error in `assays<-`(`*tmp*`, value = <S4 object of class "SimpleList">) :
current and replacement dimnames() differ
It doesn't seem safe to allow the dimnames to mismatch, even though earlier
versions allowed this.
The documentation page has
assays(x, ..., withDimnames=TRUE) <- value
but says
withDimnames: A 'logical(1)', indicating whether dimnames should be
applied to extracted assay elements (this argument is ignored
for the setter 'assays<-'). Setting 'withDimnames=FALSE'
increases the speed and memory efficiency with which assays
are extracted.
which may make one wonder why withDimnames=TRUE appears in the signature for
assays<-. The reason involves complex assignments like
names(assays(se2)) = "foo"
versus
names(assays(se2, withDimanames=FALSE)) = "bar"
The second version is more efficient than the first, avoiding copying dimnames.
The reason for the efficiency is a little complicated. assays() in these
expressions acts as both getter and setter. The signature of the setter assays<-
needs to be able to 'pass through' the withDimnames argument to the getter, even
if it has no effect on the setter itself. I have updated the documentation to
clarify this, inasmuch as I am able.
assayNames(se2) = "foo"
hides the complex assignment complexity for this common use case. Probably the
default should be FALSE for the setter, but symmetry and the complexity of the
update scenario made me hesitant to do so. Presumably handling of dimnames will
be simplified in the next release cycle, when dimnames are stored redundantly on
all assay elements.
Sorry for the confusion from my flawed commits in version 1.19.50.
Martin
and seems to occur when calling the DESeq function (right after the "gene-wise dispersion estimates" message is printed). I know that today is the deadline for fixing errors prior to release, so the time-frame is somewhat short to fix errors. To ensure that my package isn't withheld from release, should I retrograde to the last stable version of DESeq2 rather than using dev? Thanks for your help, Andrea Here is some code + sessionInfo that illustrate the error:
> library(HTSFilter) > library(DESeq2) > library(Biobase) > data(sultan) > conds <- pData(sultan)$cell.line > dds <- DESeqDataSetFromMatrix(countData = exprs(sultan),
+ colData = data.frame(cell.line = conds), + design = ~ cell.line)
> dds <- DESeq(dds)
estimating size factors
estimating dispersions
gene-wise dispersion estimates
Error in `assays<-`(`*tmp*`, value = <S4 object of class "SimpleList">) :
'dimnames' differ between assay elements
> sessionInfo()
R version 3.2.0 beta (2015-04-01 r68134) Platform: x86_64-w64-mingw32/x64 (64-bit) Running under: Windows 7 x64 (build 7601) Service Pack 1 locale: [1] LC_COLLATE=French_France.1252 LC_CTYPE=French_France.1252 LC_MONETARY=French_France.1252 [4] LC_NUMERIC=C LC_TIME=French_France.1252 attached base packages: [1] stats4 parallel stats graphics grDevices utils datasets methods base other attached packages: [1] DESeq2_1.7.46 RcppArmadillo_0.4.650.1.1 Rcpp_0.11.5 GenomicRanges_1.19.50 [5] GenomeInfoDb_1.3.16 IRanges_2.1.43 S4Vectors_0.5.22 HTSFilter_1.7.0 [9] Biobase_2.27.3 BiocGenerics_0.13.10 loaded via a namespace (and not attached): [1] RColorBrewer_1.1-2 futile.logger_1.4 plyr_1.8.1 XVector_0.7.4 tools_3.2.0 [6] futile.options_1.0.0 rpart_4.1-9 digest_0.6.8 annotate_1.45.4 RSQLite_1.0.0 [11] gtable_0.1.2 lattice_0.20-31 DBI_0.3.1 proto_0.3-10 DESeq_1.19.0 [16] cluster_2.0.1 genefilter_1.49.2 stringr_0.6.2 locfit_1.5-9.1 nnet_7.3-9 [21] grid_3.2.0 AnnotationDbi_1.29.20 XML_3.98-1.1 survival_2.38-1 BiocParallel_1.1.21 [26] foreign_0.8-61 latticeExtra_0.6-26 Formula_1.2-0 limma_3.23.11 geneplotter_1.45.0 [31] ggplot2_1.0.1 reshape2_1.4.1 lambda.r_1.1.7 edgeR_3.9.14 Hmisc_3.15-0 [36] MASS_7.3-40 scales_0.2.4 splines_3.2.0 colorspace_1.2-6 xtable_1.7-4 [41] acepack_1.3-3.3 munsell_0.4.2 [[alternative HTML version deleted]]
_______________________________________________ Bioc-devel at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/bioc-devel
Computational Biology / Fred Hutchinson Cancer Research Center 1100 Fairview Ave. N. PO Box 19024 Seattle, WA 98109 Location: Arnold Building M1 B861 Phone: (206) 667-2793