a possible bug in `subset` function

[Zhengnan Cheng]

Dear Cardinal developers,

 

I hope this message finds you well. I am writing to bring to your attention a bug that I have encountered while using the Cardinal R package. The `subset` function is much slower in the newest version than the old one.

Here is my test code:

library(Cardinal) |> suppressWarnings() |> suppressMessages()

file <- 'test.imzML'

print('readMSIData')

system.time({

    mse <- Cardinal::readMSIData(file, attach.only = FALSE)

})

print(mse)

print('subsetPixels')

system.time({

    submse <- subsetPixels(mse, run == runNames(mse)[1])

})

print('subsetFeatures')

system.time({

    submse <- subsetFeatures(mse, mz == mz(mse)[1])

})

sessionInfo()

Cardinal_3.0.1: 

[1] "readMSIData"

   user  system elapsed

  2.261   0.810   3.135

An object of class 'MSContinuousImagingExperiment'

  <1779 feature, 75924 pixel> imaging dataset

    imageData(1): intensity

    featureData(0):

    pixelData(0):

    metadata(12): spectrum representation ibd binary type ... files

        name

    run(1): test

    raster dimensions: 299 x 386

    coord(2): x = 1..299, y = 1..386

    mass range:  125.0001 to 1281.6032

    centroided: TRUE

[1] "subsetPixels"

   user  system elapsed

  1.016   0.475   1.505

[1] "subsetFeatures"

   user  system elapsed

  0.046   0.000   0.047

R version 4.2.3 (2023-03-15)

Platform: x86_64-conda-linux-gnu (64-bit)

Running under: CentOS Linux 7 (Core)

locale:

 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              

 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    

 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8  

 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                

 [9] LC_ADDRESS=C               LC_TELEPHONE=C            

[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C      

attached base packages:

[1] stats4    stats     graphics  grDevices utils     datasets  methods  

[8] base    

other attached packages:

[1] Cardinal_3.0.1      S4Vectors_0.36.2    EBImage_4.40.0    

[4] BiocParallel_1.32.6 BiocGenerics_0.44.0 ProtGenerics_1.30.0

loaded via a namespace (and not attached):

 [1] magrittr_2.0.3    MASS_7.3-58.3     mclust_6.0.0      viridisLite_0.4.1

 [5] lattice_0.21-8    jpeg_0.1-10       rlang_1.1.1       fastmap_1.1.1    

 [9] tools_4.2.3       parallel_4.2.3    grid_4.2.3        biglm_0.9-2.1    

[13] Biobase_2.58.0    nlme_3.1-162      png_0.1-8         irlba_2.3.5.1    

[17] DBI_1.1.3         cli_3.6.1         matter_2.0.1      htmltools_0.5.5  

[21] abind_1.4-5       digest_0.6.31     Matrix_1.5-4.1    htmlwidgets_1.6.2

[25] bitops_1.0-7      fftwtools_0.9-11  codetools_0.2-19  signal_0.7-7    

[29] RCurl_1.98-1.12   tiff_0.1-11       sp_1.6-1          compiler_4.2.3  

[33] locfit_1.5-9.7

Cardinal_3.4.3: 

[1] "readMSIData"

   user  system elapsed

  5.024   0.423   5.579

An object of class 'MSContinuousImagingExperiment'

  <1779 feature, 75924 pixel> imaging dataset

    imageData(1): intensity

    featureData(0):

    pixelData(0):

    metadata(1): parse

    run(1): test

    raster dimensions: 299 x 386

    coord(2): x = 1..299, y = 1..386

    mass range:  125.0001 to 1281.6032

    centroided: TRUE

[1] "subsetPixels"

   user  system elapsed

 92.075   1.266  93.563

[1] "subsetFeatures"

   user  system elapsed

  0.530   0.002   0.533

R version 4.3.2 (2023-10-31)

Platform: x86_64-conda-linux-gnu (64-bit)

Running under: CentOS Linux 7 (Core)

locale:

 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              

 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    

 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8  

 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                

 [9] LC_ADDRESS=C               LC_TELEPHONE=C            

[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C      

time zone: Asia/Shanghai

tzcode source: system (glibc)

attached base packages:

[1] stats4    stats     graphics  grDevices utils     datasets  methods  

[8] base    

other attached packages:

[1] Cardinal_3.4.3      S4Vectors_0.40.2    EBImage_4.44.0    

[4] BiocParallel_1.36.0 BiocGenerics_0.48.1 ProtGenerics_1.34.0

loaded via a namespace (and not attached):

 [1] nlme_3.1-164       cli_3.6.2          rlang_1.1.3        DBI_1.2.2        

 [5] mclust_6.1         tiff_0.1-12        png_0.1-8          RCurl_1.98-1.14  

 [9] htmltools_0.5.8    ontologyIndex_2.12 locfit_1.5-9.9     Biobase_2.62.0    

[13] grid_4.3.2         abind_1.4-5        MASS_7.3-60.0.1    bitops_1.0-7      

[17] fastmap_1.1.1      compiler_4.3.2     codetools_0.2-20   irlba_2.3.5.1    

[21] CardinalIO_1.0.0   htmlwidgets_1.6.4  fftwtools_0.9-11   lattice_0.22-6    

[25] digest_0.6.35      signal_1.8-0       viridisLite_0.4.2  matter_2.4.1      

[29] parallel_4.3.2     magrittr_2.0.3     Matrix_1.6-5       tools_4.3.2      

[33] jpeg_0.1-10        biglm_0.9-2.1    

I found that in the old version, loading MSI data into memory (using `attach.only = FALSE`) greatly improved the speed of subset operations, but this improvement is not significant in the newest version. The same problems occurred when using slicing methods ([]).

Your prompt attention to this matter would be highly appreciated.  Thank you for your time and effort in maintaining the package. 
 
Best regards,
Zhengnan Cheng