Hi,

I have been running the mnnCorrect from the SCRAN package without issues, and now wanted to run the fastMNN (scripts downloaded form github; https://github.com/MarioniLab/scran/tree/master/R). However, when calling the fastMNN function ( mnn.out <- do.call(fastMNN, c(original, list(k=20, d=50))) ) I end up getting the error:

 Fejl i FUN(X[[i]], ...) : objekt 'cxx_cosine_norm' blev ikke fundet

I can see that "cxx_cosine_norm" is called within the cosineNorm function (out <- .Call(cxx_cosine_norm, X, mode!="l2norm")), which is called within the fastMNN function. I am puzzled as the mnnCorrect calls the same cosineNorm function, without any issues. Does anybody how an suggestion to why the "cxx_cosine_norm" can't be found when called through fastMNN?

All help will be greatly appreciated!

/Tobias

Session info:

> sessionInfo()
R version 3.5.0 (2018-04-23)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS High Sierra 10.13.5

Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.5/Resources/lib/libRlapack.dylib

locale:
[1] da_DK.UTF-8/da_DK.UTF-8/da_DK.UTF-8/C/da_DK.UTF-8/da_DK.UTF-8

attached base packages:
[1] stats4    parallel  stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] BiocManager_1.30.1          dplyr_0.7.6                 Matrix_1.2-14               scran_1.8.2                
 [5] DropletUtils_1.1.5          scater_1.9.7                SingleCellExperiment_1.3.6  SummarizedExperiment_1.11.5
 [9] DelayedArray_0.7.19         BiocParallel_1.14.2         matrixStats_0.53.1          GenomicRanges_1.33.7       
[13] GenomeInfoDb_1.17.1         IRanges_2.15.14             S4Vectors_0.19.17           ggplot2_3.0.0              
[17] Biobase_2.41.1              BiocGenerics_0.27.1        

loaded via a namespace (and not attached):
 [1] viridis_0.5.1            dynamicTreeCut_1.63-1    edgeR_3.23.3             viridisLite_0.3.0        DelayedMatrixStats_1.3.4
 [6] shiny_1.1.0              assertthat_0.2.0         statmod_1.4.30           GenomeInfoDbData_1.1.0   vipor_0.4.5             
[11] pillar_1.3.0             lattice_0.20-35          glue_1.2.0               limma_3.37.3             digest_0.6.15           
[16] promises_1.0.1           XVector_0.21.3           colorspace_1.3-2         htmltools_0.3.6          httpuv_1.4.4.2          
[21] plyr_1.8.4               pkgconfig_2.0.1          zlibbioc_1.27.0          purrr_0.2.5              xtable_1.8-2            
[26] scales_0.5.0             later_0.7.3              tibble_1.4.2             DT_0.4                   withr_2.1.2             
[31] lazyeval_0.2.1           magrittr_1.5             crayon_1.3.4             mime_0.5                 beeswarm_0.2.3          
[36] FNN_1.1                  tools_3.5.0              data.table_1.11.4        stringr_1.3.1            Rhdf5lib_1.3.1          
[41] munsell_0.5.0            locfit_1.5-9.1           bindrcpp_0.2.2           compiler_3.5.0           rlang_0.2.1             
[46] rhdf5_2.25.4             grid_3.5.0               RCurl_1.95-4.10          tximport_1.9.8           rstudioapi_0.7          
[51] rjson_0.2.20             htmlwidgets_1.2          igraph_1.2.1             bitops_1.0-6             gtable_0.2.0            
[56] reshape2_1.4.3           R6_2.2.2                 gridExtra_2.3            bindr_0.1.1              stringi_1.2.3           
[61] ggbeeswarm_0.6.0         Rcpp_0.12.17             tidyselect_0.2.4

I imagine that you're sourceing the fastMNN.R file and trying to run the function. Unfortunately, packages don't work that way. The cxx_cosine_norm object provides a registered entry point to the C++ shared library within scran, which is only visible to other scran functions in the installed package (i.e., following R CMD INSTALL or friends). It's not exported, nor should it be - why would a user ever need to deal with C++ code? You'll face similar issues with several internal R functions that are used by fastMNN but are not exported.

The only solution is to actually install the version of scran that contains fastMNN. This requires switching packages over to BioC-devel, see the instructions at https://www.bioconductor.org/developers/how-to/useDevel/.