vladpetyuk / msnid Goto Github PK

Hi Vlad,
I've recently run into an error related to non-standard evaluation in your apply_filter method. The problem appeared when building the Bioconductor "proteomics" workflow vignette by a call to rmarkdown::render from inside of another function, see at the bottom of http://docbuilder.bioconductor.org:8080/job/proteomics/20/label=master/console

It appears to me that this is some sort of scoping issue caused by the fact that eval(exprssn) evaluates exprssn in the calling environment, in this case your apply_filter method. Things seem to work fine provided that the objects are defined in the global environment, which is usually the case when working interactively, or for code chunks inside a vignette. However, things break once the rmarkdown::render function is called from another function, which is the case for the BioC workflow builder scripts. This is because by default render evaluates code chunks in its calling environment and not in the global environment; in such a case object names from the expression passed to the apply_filter method are not resolved properly.

The workaround which I've implemented on the BioC workflow builders is to run render in the global environment by specifying envir = parent.frame(2L). However, it would be good to make your code bullet proof in first place. I've already tried replacing the following line in R/MSnID-methods.R

msnidObj@psms <- msnidObj@psms[eval(exprssn),]

by

msnidObj@psms <- msnidObj@psms[eval(exprssn, msnidObj@psms, parent.frame()),]

but strangely enough, this doesn't seem to make any difference. I would expect that specifying the enclosing environments should alleviate the issue, but somehow this doesn't work as expected.

Any thoughts?

Cheers,
Andrzej

need to add a way to combine two (and thus more) MSnID objects.
candidate approaches are: c, +, combine

Would you consider a patch to support mzR::openIDfile to read the ident files. It is really fast!

Turns out FDR calculation at the PSM level ignores peptide-to-protein redundancy. That is some PSMs are counted more then others in the FDR calculation. Not that it results in substantially different value, but still technically this is not accurate. Note, that this does not affect more FDR calculation at peptide and proteins/accession levels. This should be fixed the throwing away protein/accession column and taking unique rows before PSM FDR calculation.

I'm running assess_missed_cleavages() and it is returning the following error:

Error: all(grepl("^.\\..+\\..$", peptide)) is not TRUE

I've made sure that "peptide" %in% names(msnid) is true. Also, all(grepl("^.\\..+\\..$", msnid_out$peptide)) is true, where msnid_out is the output of psms(msnid).

Any help is greatly appreciated!

Here is my session information:

R version 3.4.3 (2017-11-30)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows >= 8 x64 (build 9200)

Matrix products: default

locale:
[1] LC_COLLATE=English_Canada.1252 LC_CTYPE=English_Canada.1252 LC_MONETARY=English_Canada.1252
[4] LC_NUMERIC=C LC_TIME=English_Canada.1252

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

other attached packages:
[1] ggplot2_2.2.1 magrittr_1.5 MSGFplus_1.12.0 MSnbase_2.4.2 ProtGenerics_1.10.0 BiocParallel_1.12.0
[7] Biobase_2.38.0 BiocGenerics_0.24.0 MSnID_1.12.1 mzID_1.16.0 mzR_2.12.0 Rcpp_0.12.16

loaded via a namespace (and not attached):
[1] reshape2_1.4.3 lattice_0.20-35 colorspace_1.3-2 stats4_3.4.3 vsn_3.46.0
[6] XML_3.98-1.11 rlang_0.2.0 R.oo_1.22.0 pillar_1.2.2 glue_1.2.0
[11] R.utils_2.6.0 affy_1.56.0 bindrcpp_0.2.2 affyio_1.48.0 foreach_1.4.4
[16] R.cache_0.13.0 bindr_0.1.1 plyr_1.8.4 stringr_1.3.1 zlibbioc_1.24.0
[21] munsell_0.4.3 pcaMethods_1.70.0 gtable_0.2.0 R.methodsS3_1.7.1 codetools_0.2-15
[26] labeling_0.3 IRanges_2.12.0 doParallel_1.0.11 BiocInstaller_1.28.0 preprocessCore_1.40.0
[31] scales_0.5.0 limma_3.34.9 S4Vectors_0.16.0 impute_1.52.0 digest_0.6.15
[36] stringi_1.2.2 dplyr_0.7.4 grid_3.4.3 tools_3.4.3 lazyeval_0.2.1
[41] tibble_1.4.2 pkgconfig_2.0.1 data.table_1.11.2 assertthat_0.2.0 iterators_1.0.9
[46] R6_2.2.2 MALDIquant_1.17 compiler_3.4.3

Would it be possible to ditch the peptides generic from MSnID and re-use the one defined in ProtGenerics instead?