Hi,
When I apply fRMA on a dataset using a windows machine and a linux machine I find diffrences when I compare the expressions, Is this expected ,because the diffrences are not small. I am using more or less the same R versions and librarys.
windows:
R version 3.4.4 (2018-03-15)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)
Matrix products: default
locale:
[1] LC_COLLATE=Dutch_Netherlands.1252 LC_CTYPE=Dutch_Netherlands.1252 LC_MONETARY=Dutch_Netherlands.1252 LC_NUMERIC=C
[5] LC_TIME=Dutch_Netherlands.1252
attached base packages:
[1] parallel stats graphics grDevices utils datasets methods base
other attached packages:
[1] hgu133plus2cdf_2.18.0 hgu133plus2frmavecs_1.5.0 affy_1.56.0 frma_1.30.1 Biobase_2.38.0 BiocGenerics_0.24.0
loaded via a namespace (and not attached):
[1] Rcpp_0.12.15 BiocInstaller_1.28.0 compiler_3.4.4 pillar_1.2.1 GenomeInfoDb_1.14.0
[6] XVector_0.18.0 bitops_1.0-6 iterators_1.0.9 tools_3.4.4 zlibbioc_1.24.0
[11] digest_0.6.15 bit_1.1-12 memoise_1.1.0 RSQLite_2.0 preprocessCore_1.40.0
[16] tibble_1.4.2 lattice_0.20-35 ff_2.2-13 rlang_0.2.0 Matrix_1.2-12
[21] foreach_1.4.4 DelayedArray_0.4.1 DBI_0.7 yaml_2.1.17 GenomeInfoDbData_1.0.0
[26] affxparser_1.50.0 Biostrings_2.46.0 S4Vectors_0.16.0 IRanges_2.12.0 stats4_3.4.4
[31] bit64_0.9-7 grid_3.4.4 AnnotationDbi_1.40.0 oligo_1.42.0 blob_1.1.0
[36] codetools_0.2-15 matrixStats_0.53.1 oligoClasses_1.40.0 MASS_7.3-49 GenomicRanges_1.30.0
[41] splines_3.4.4 SummarizedExperiment_1.8.1 RCurl_1.95-4.10 affyio_1.48.0
Linux:
R version 3.3.2 (2016-10-31)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Scientific Linux 7.4 (Nitrogen)
Matrix products:
locale:
[1] C
attached base packages:
[1] parallel methods tools stats graphics grDevices utils datasets base
other attached packages:
[1] hgu133plus2cdf_2.18.0 hgu133plus2frmavecs_1.5.0 frma_1.30.1 affy_1.52.0 Biobase_2.34.0 BiocGenerics_0.20.0
[7] yaml_2.1.14
loaded via a namespace (and not attached):
[1] Rcpp_0.12.13 AnnotationDbi_1.36.0 affxparser_1.46.0 XVector_0.14.0 splines_3.3.2
[6] GenomicRanges_1.26.1 zlibbioc_1.20.0 MASS_7.3-45 IRanges_2.8.1 bit_1.1-12
[11] lattice_0.20-34 foreach_1.4.3 GenomeInfoDb_1.10.1 SummarizedExperiment_1.4.0 grid_3.3.2
[16] ff_2.2-13 DBI_0.7-12 iterators_1.0.8 oligoClasses_1.36.0 digest_0.6.11
[21] preprocessCore_1.36.0 oligo_1.38.0 affyio_1.44.0 Matrix_1.2-7.1 codetools_0.2-15
[26] S4Vectors_0.12.1 memoise_1.0.0 RSQLite_1.1-1 BiocInstaller_1.24.0 Biostrings_2.42.1
[31] stats4_3.3.2
Unfortunately I don't have access to change the linux machine, but I did update the windows machine to the current version but I receive an error when I'm trying to use the frma library (that was the reason I went back to the older version on my windows machine) :
`> library(frma)`
Error: package or namespace load failed for ‘frma’ in loadNamespace(i, c(lib.loc, .libPaths()), versionCheck = vI[[i]]):
there is no package called ‘BiocParallel’
So using the newest version doesn't really solve the problem. BUt maybe you can help me with this problem
There isn't a Windows binary for BiocParallel, so you need to get set up to compile. This used to be a daunting task, but now it's a simple download and install of the Rtool set. You obviously want the one for R-3.5.x, and you need to install as administrator and I usually click the box to add rtools to your path.
Once you have done that you should be able to restart R and then do
library(BiocInstaller) biocLite("BiocParallel")at which time it will say it's a source package and do you want to compile, to which you say heck yeah.
I should also note that unless your sysadmin is cheaping you out on personal space on the Linux box, you can always just compile R in your home dir and add that to your path. Or if you have write access to a larger dir, you could do it there.
Thanks alot, this seems to work!
The reason why we can't just update to the newest R version is that alot of experiments are based on the older version and we find different results for different platforms(Linux vs Windows) as well as differences with old and new librarys(old fRMA vs new fRMA same samples) running the same machine. We can't redo our experiments when we find biases like these.