mass2adduct - Finding molecular adducts in mass spectrometry data

In mass spectrometry imaging, adducts can form between target molecules (e.g. metabolites) and other substances such as matrix or salt ions. This package presents tools for counting and identifying possible adducts in MS data, and accompanies Janda et al. (2021).

Read the paper here.

Install and load package

You will need the devtools package, if you don't have it already:

install.packages("devtools") # Install from CRAN
library(devtools) # Load the package
install.packages(c("knitr","rmarkdown")) # If you want to build vignette

If you want to build the vignette you will need knitr and rmarkdown R packages, as well as pandoc.

Install mass2adduct from Github:

install_github("kbseah/mass2adduct")
install_github("kbseah/mass2adduct", build_vignettes=TRUE, build=TRUE, build_opts=c("--no-resave-data","--no-manual")) # Build vignette "mass2adduct" (not done by default)

The vignette contains a detailed explanation of the mass2adduct package with a worked example, and it is recommended to read it if you want to understand the reasoning behind the package.

Load mass2adduct package after it is installed

library(mass2adduct)

Import data

Import MS imaging intensity data

MSI data exported from the MSiReader software with the "intensity export" function, or from some other third party software that can output plain-text comma-delimited tables (CSV format). You can find an example CSV file in the folder inst/extdata. Import the data into R as a data.frame:

d <- msimat("msi.csv", sep=";")
class(d) # "msimat"
print(d) # Report a summary
plot(d) # Mass spectrum of total intensities per peak

If the data matrix is very large, consider reformatting it (see "Reformatting large CSV files" below).

Import Cardinal objects

Cardinal is an R package for working with MSI data. Cardinal MSProcessedImagingExperiment or MSContinuousImagingExperiment objects (requires Cardinal v2.2 and above) can be converted to mass2adduct's msimat format. However, the data must first be pre-processed, with peaks already binned with the peakBin() function from Cardinal.

# Read ImzML file
d <- readImzML("msi_file")
# Pre-process and peak bin with existing peaklist
d_peaks <- d %>% normalize() %>% smoothSignal() %>% reduceBaseline() %>% peakBin(peaklist) %>% process
# Convert to msimat format for mass2adduct
d_msimat <- cardinal2msimat(d_peaks)
# Compare peak histograms from Cardinal vs mass2adduct, should look the same
plot(d_peaks)
plot(d_msimat)

Import list of masses directly

If you do not intend to use pixel intensity values for downstream analyses, you can import a list of mass values as a plain text file (one per line):

peaks <- scan("mass_example",what=numeric())

Tabulate all pairwise mass differences

Take all possible pairs of masses and calculate the mass difference for each pair. These mass differences represent potential molecular adducts.

d.diff <- massdiff(peaks) # From MSI data
peaks.diff <- massdiff(peaks) # From simple numeric vector of mass values

Output is a massdiff object with three elements: the two parent masses and the difference between them.

Bin mass differences into a histogram

The calculated mass differences are misleadingly precise, because measurement error and uncertainty are not taken into account. They should be binned into a histogram with a user-specified bin width, that depends on the known mass precision of your instrument. In this example, we use the value of 0.01:

d.diff.hist <- hist(d.diff) # Produces a histogram object
plot(d.diff.hist) # Plot the histogram

The resulting object d.diff.hist is a standard R histogram object. You can "zoom" into specific regions of the plot with the xlim parameter in the R plot command.

plot (d.diff.hist, xlim=c(100,150))

Count occurrences of known adducts

The package comes with a small data set of common molecular adducts, called adducts (use the command help(adducts) to see a description). Users can supply their own custom sets of adducts as long as they are in the same format (as a data.frame with three columns: name, formula, mass)

The following function looks for known adducts by finding the closest-matching bin in the mass difference histogram produced above. It reports the number of counts (i.e. how many pairs of MS peaks have that mass difference) and the quantile.

Note that quantile values will usually be quite high because the majority of mass differences have zero to few counts.

head(adductMatch (d.diff.hist)) # Show the top ten matches to known adducts

List the highest-counted mass differences and any matches to known adducts

The procedure above (adductMatch) only looks for known adducts. What about peaks in the histogram which may not have a good match with the reference list? The function topAdducts performs the complementary function: Rank mass differences by the number of times they are observed, and report any matches to known adducts.

topAdducts(d.diff.hist,n=10) # Report top ten hits

Find mass peaks associated with specific mass difference

Find parent and derivative mass peaks associated with a specific molecular adduct, e.g. DHB-H2O (mass 136.01600), within a window of 1 mDa, i.e. +/- 0.5 mDa. Alternatively, the window can be set as a proportion of the mass peak center in ppm units.

d.diff.DHBH2O <- diffGetPeaks(d.diff, mass=136.01600, mDa=1)

Output is a subset of the original d.diff object.

Test for significant correlations between mass peaks

Test for spatial correlations between mass peaks in MS imaging data (imported with the msimat function).

For example, we wish to see if what we believe to be pairs of parent- and derivative-ion masses tend to occur together. If they are truly related by molecular adduct formation, then their abundances should be correlated. These putative pairs are found with the diffGetPeaks function (above).

d.diff.DHBH2O.corr <- corrPairsMSI(d, d.diff.DHBH2O)
d.diff.DHBH2O.corr <- corrPairsMSI(d, d.diff.DHBH2O, how="parallel", ncores=4) # If using a Unix system, code can be parallelized

Output is a data.frame with p-values for each ion pair. By default the cutoff for significance is p=0.05 with Bonferroni correction.

For large data sets, where the tables would not fit into memory, it is possible to break up the problem into "chunks" processed serially. Use the function corrPairsMSIchunks instead of corrPairsMSI. See the help message for details.

Reformatting large CSV files

Plain-text CSV files of MSI data exported by software such as SCILS or MSIreader can be large, on the order of several Gb. For many MSI data sets, a lot of this is "wasted" because the majority of peaks are only detected in a minority of pixels. Low-abundance peaks can be thrown out through filtering, but the majority of entries in the data matrix may still be zeroes.

Reading these large files directly into R is inefficient and may not be possible because of insufficient memory. To be more economical, the data can be represented in a "triplet" format rather than matrix.

A Perl script msimunging.pl (in the inst/ subdirectory of the package source) is provided to do this format conversion before importing the files into R. It can convert between plain CSV format, the so-called "triplet" format, and a binary format readable by Perl (mostly used for testing). Note that it is nonetheless still limited by the amount of memory you have on your system (e.g. if your CSV file is 40 Gb and your RAM is 32 Gb, then it will probably not work).

The script also provides the option to filter peaks by their frequency of nonzero pixels.

Conversion to triplet format output gives five files, which correspond to the inputs required by the msimat() function in the R package - rows,cols,vals,peaks,spots. Importing a data set with those parameters, and without specifying anything to csv, will store the matrix internally in triplet form, which will also take up less memory.

Usage information is shown by typing msimunging.pl --help.

Help and documentation

Further documentation for all the above functions can be found with the help or ? functions in R. Overview of the entire package can be viewed with help(mass2adduct). List of exported functions is given by library(help="mass2adduct"). If you used the build_vignettes=TRUE option when installing the package, you can view the mass2adduct vignette with vignette("mass2adduct",package="mass2adduct").

Please report any problems to the package maintainer, either by email or with the issue tracker on Github.

Citation

Janda M, Seah BKB, Jakob D, Beckmann J, Geier B, Liebeke M (2021) Determination of abundant metabolite matrix adducts illuminates the dark metabolome of MALDI-mass spectrometry imaging datasets. Analytical Chemistry. doi:10.1021/acs.analchem.0c04720

Department of Symbiosis, Max Planck Institute for Marine Microbiology