mzR::get3Dmap only return zeros

[joerg-b]

I've recently updated my computer (macOS Monterey, R 4.2, Bioconductor 3.15) and in the process I've also installed mzR.
Now, while running some scripts that worked perfectly fine on R 3.6 and R 4.0, I'm encountering some problems. I've located the problem to originate from get3Dmap only returning a matrix of zeros. The dimensions of the matrix seem to be correct. When I use mzR::peaks on the same data file I get lot's of non-zero values.

I can provide a link to an example .mzML file if that helps.

Thanks for your help!
Joerg

[sneumann]

Hi, I can confirm, the following "test" fails:

library(mzR)
ms <- openMSfile(list.files(system.file("microtofq", package = "msdata"), pattern="MM14.mzML",full.names=TRUE, recursive = TRUE))
threedee <- get3Dmap(ms, scans=1:100, lowMz=100, highMz=1000, resMz=1)
sum(threedee) > 1

on

mzR_2.30.0
R version 4.2.0 (2022-04-22)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 20.04.4 LTS

It does work on

R version 3.6.3 (2020-02-29)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 16.04.7 LTS

[sneumann]

It is called from

mzR/R/methods-mzRpwiz.R

Line 1 in 458704d

setMethod("get3Dmap", "mzRpwiz",

Actual calculation happens in C++:

mzR/src/RcppPwiz.cpp

Line 990 in 4745773

Rcpp::NumericMatrix RcppPwiz::get3DMap ( std::vector<int> scanNumbers, double whichMzLow, double whichMzHigh, double resMz )

My current suspicion is the updated pwiz version.

[sneumann]

So, some git bisect later:

2aed7135fee1fc23daa795b189bb62d335a3bbbf is the first bad commit
commit 2aed7135fee1fc23daa795b189bb62d335a3bbbf
Date:   Mon Oct 25 10:58:52 2021 +0200

        refactor: patially rewrite code to exctract header info
    
        - Create index list in R instead of C++.
        - Use `size_t` instead of `int` where suitable.
        - Ensure that correct data types are passed between R and C++.
        - Use new (own) implementation to extract the acquisition number from the
          spectrum ID instead of `id::translateNativeIDToScanNumber` from proteowizard
          to avoid segfaults on macOS (https://github.com/sneumann/xcms/issues/422).

 DESCRIPTION         |   2 +-
 NEWS                |   9 +++++
 R/methods-mzRpwiz.R |  10 +++---
 src/RcppPwiz.cpp    | 101 ++++++++++++++++++++++++++++++++++------------------
 src/RcppPwiz.h      |   8 +++--
 5 files changed, 89 insertions(+), 41 deletions(-)

[jorainer]

Uh - then it seems that's my fault. I'll have a look and fix.

[jorainer]

Sorry, super-stupid misspelling from my side. Wrote DetailLevel_FullMetadata instead of DetailLevel_FullData in the spectrum call, thus only header information were imported but no m/z and intensity values. It's fixed now in PR #269 that also adds a unit test for the get3Dmap function.

[sneumann]

Thank Jörg for reporting, thanks to git bisect run to identify the causal commit (my first guess was wrong), and Jo for the super fast fix. Pushed upstream as 2.31.1 Yours, Steffen