Just wanted to describe a bug I'm getting for the GSVA package.

This little code bit should calculate a little TLR4 signaling pathway on the dataset.

rm(list=ls())
library(GSVA)
library(leukemiasEset)
data(leukemiasEset)
dat <- exprs(leukemiasEset)
TLR4 = 'ENSG00000136869'
MYD88 = 'ENSG00000172936'
plot(x=dat[TLR4,],y=dat[MYD88,])
gset = list(LPS=c('ENSG00000136869','ENSG00000172936'))
pathway <- gsva(dat, gset)$es.obs
plot(x=dat[TLR4,],y=pathway['LPS',])

But this bit of code produces opposite GSVA scores when computed on mac and windows.

Mac was done on R 3.4.2 with Bioconductor 3.6, and Windows was R 3.4.1 with Bioconductor 3.6.

You can see the Windows correlation looks right, but Mac is the opposite.

Windows:

And on Mac OS the same bit of code is this:

I'm not sure how to page the package owner on this forums or if there is a more appropriate place for this post.  But it should be of general interest to anyone who sees that their GSVA results on Mac are opposite of what they expect.. positive and negative switch.

Thanks!

Jason

Edit: I've sent an email to the package maintainer to let them know of the problem.

I would guess that SVA uses singular value decomposition (SVD); in SVD, the sign of the singular vectors is arbitrary and may change between implementations. Here's a quote from help(svd):

     Recall that the singular vectors are only defined up to sign (a
     constant of modulus one in the complex case).  If a left singular
     vector has its sign changed, changing the sign of the
     corresponding right vector gives an equivalent decomposition.

Dear Jason,

thanks for your message. I don't have access to a windows machine but I do have an apple laptop with macOS Sierra (10.12.6) and i cannot reproduce the problem in the release nor in the development version. i get exactly the same numbers in macOS and in linux. I'm providing here below my session information where I have omitted the linux session information to meet the maximum allowed characters per answer (5000 characters). Please check that you are using the latest release version of GSVA in both platforms and verify whether there are different versions of the software between my session information and yours.

cheers,

robert.

** macOS sierra release sessionInfo():

R version 3.4.0 (2017-04-21)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS Sierra 10.12.6

Matrix products: default
BLAS: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRblas.0.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRlapack.dylib

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] parallel  stats     graphics  grDevices utils     datasets  methods
[8] base

other attached packages:
[1] leukemiasEset_1.12.0 Biobase_2.36.2       BiocGenerics_0.22.0
[4] GSVA_1.24.2          BiocInstaller_1.26.1 nvimcom_0.9-28
[7] colorout_1.1-2

loaded via a namespace (and not attached):
 [1] graph_1.54.0         Rcpp_0.12.13         AnnotationDbi_1.38.2
 [4] IRanges_2.10.4       bit_1.1-12           xtable_1.8-2
 [7] rlang_0.1.2          blob_1.1.0           tools_3.4.0
[10] DBI_0.7              bit64_0.9-7          digest_0.6.12
[13] tibble_1.3.4         GSEABase_1.38.2      S4Vectors_0.14.6
[16] bitops_1.0-6         RCurl_1.95-4.8       memoise_1.1.0
[19] RSQLite_2.0          compiler_3.4.0       stats4_3.4.0
[22] XML_3.98-1.9         annotate_1.54.0

** macOS sierra devel sessionInfo():

R version 3.4.0 (2017-04-21)
Platform: x86_64-apple-darwin16.6.0 (64-bit)
Running under: macOS Sierra 10.12.6

Matrix products: default
BLAS: /opt/R/R-3.4.0_BioCdevel/lib/R/lib/libRblas.dylib
LAPACK: /opt/R/R-3.4.0_BioCdevel/lib/R/lib/libRlapack.dylib

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] parallel  stats     graphics  grDevices utils     datasets  methods
[8] base

other attached packages:
[1] leukemiasEset_1.13.0 Biobase_2.37.2       BiocGenerics_0.23.2
[4] GSVA_1.25.6          nvimcom_0.9-28       colorout_1.1-2

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.13                lattice_0.20-35
 [3] prettyunits_1.0.2           Rsamtools_1.29.1
 [5] Biostrings_2.45.4           assertthat_0.2.0
 [7] digest_0.6.12               mime_0.5
 [9] BiocFileCache_1.1.16        R6_2.2.2
[11] GenomeInfoDb_1.13.4         stats4_3.4.0
[13] RSQLite_2.0                 httr_1.3.1
[15] zlibbioc_1.23.0             rlang_0.1.2
[17] GenomicFeatures_1.29.11     progress_1.1.2
[19] lazyeval_0.2.0              annotate_1.55.0
[21] blob_1.1.0                  S4Vectors_0.15.10
[23] Matrix_1.2-11               shinythemes_1.1.1
[25] BiocParallel_1.11.9         geneplotter_1.55.0
[27] stringr_1.2.0               RCurl_1.95-4.8
[29] bit_1.1-12                  biomaRt_2.33.4
[31] shiny_1.0.5                 DelayedArray_0.3.21
[33] compiler_3.4.0              httpuv_1.3.5
[35] rtracklayer_1.37.3          Organism.dplyr_1.3.2
[37] pkgconfig_2.0.1             htmltools_0.3.6
[39] SummarizedExperiment_1.7.10 tibble_1.3.4
[41] GenomeInfoDbData_0.99.1     IRanges_2.11.17
[43] matrixStats_0.52.2          XML_3.98-1.9
[45] dplyr_0.7.4                 dbplyr_1.1.0
[47] GenomicAlignments_1.13.6    bitops_1.0-6
[49] rappdirs_0.3.1              grid_3.4.0
[51] xtable_1.8-2                GSEABase_1.39.4
[53] DBI_0.7                     AnnotationFilter_1.1.9
[55] magrittr_1.5                graph_1.55.0
[57] stringi_1.1.5               XVector_0.17.1
[59] bindrcpp_0.2                RColorBrewer_1.1-2
[61] tools_3.4.0                 bit64_0.9-7
[63] glue_1.1.1                  AnnotationDbi_1.39.3
[65] GenomicRanges_1.29.14       memoise_1.1.0
[67] bindr_0.1