MsQuality 1.0.0
Data quality assessment is an integral part of preparatory data analysis to ensure sound biological information retrieval.
We present here the MsQuality package, which provides functionality to
calculate quality metrics for mass spectrometry-derived, spectral data at the
per-sample level. MsQuality relies on the
mzQC framework of quality metrics defined
by the Human Proteome Organization-Proteomics Standards Intitiative (HUPO-PSI).
These metrics quantify the quality of spectral raw files using a controlled
vocabulary. The package is especially addressed towards users that acquire
mass spectrometry data on a large scale (e.g. data sets from clinical settings
consisting of several thousands of samples): while it is easier to control
for high-quality data acquisition in small-scale experiments, typically run
in one or few batches, clinical data sets are often acquired over longer
time frames and are prone to higher technical variation that is often
unnoticed. MsQuality tries to address this problem by calculating metrics that
can be stored along the spectral data sets (raw files or feature-extracted
data sets). MsQuality, thus, facilitates the tracking of shifts in data
quality and quantifies the quality using multiple metrics. It should be thus
easier to identify samples that are of low quality (high-number of missing
values, termination of chromatographic runs, low instrument sensitivity, etc.).
We would like to note here that these metrics only give an indication of
data quality, and, before removing indicated low-quality samples from the
analysis more advanced analytics, e.g. using the implemented functionality
and visualizations in the MatrixQCvis package, should be scrutinized.
Also, data quality
should always be regarded in the context of the sample type and experimental
settings, i.e. quality metrics should always be compared with regard to the
sample type, experimental setup, instrumentation, etc..
The MsQuality package allows to calculate low-level quality metrics that
require minimum information on mass spectrometry data: retention time,
m/z values, and associated intensities.
The list included in the mzQC framework is excessive, also including
metrics that rely on more high-level information, that might not be readily
accessible from .raw or .mzML files, e.g. pump pressure mean, or rely
on alignment results, e.g. retention time mean shift, signal-to-noise ratio,
precursor errors (ppm).
The MsQuality package is built upon the Spectra and the MsExperiment
package.
Metrics will be calculated based on the information stored in a
Spectra object and the respective dataOrigin entries are used to
distinguish between the mass spectral data of multiple samples.
The MsExperiment serves as a container to
store the mass spectral data of multiple samples. MsQuality enables the user
to calculate quality metrics both on Spectra and MsExperiment objects.
In this vignette, we will (i) create some exemplary Spectra and MsExperiment
objects, (ii) calculate the quality metrics on these
data sets, and (iii) visualize some of the metrics.
Other R packages are available in Bioconductor that are able to assess
the quality of mass spectrometry data:
artMS
uses MaxQuant output and enables to calculate several QC metrics, e.g.
correlation matrix for technical replicates, calculation of total sum of
intensities in biological replicates, total peptide counts in biological
replicates, charge state distribution of PSMs identified in each biological
replicates, or MS1 scan counts in each biological replicate.
MSstatsQC
and the visualization tool
MSstatsQCgui
require csv files in long format from spectral processing tools such as
Skyline and Panorama autoQC or MSnbase objects. MSstatsQC enables to
generate individual, moving range, cumulative sum for mean, and/or
cumulative sum for variability control charts for each metric. Metrics
can be any kind of user-defined metric stored in the data columns for a given
peptide, e.g. retention time and peak area.
MQmetrics
provides a pipeline to analyze the quality of proteomics data sets from
MaxQuant files and focuses on proteomics-/MaxQuant-specific metrics, e.g.
proteins identified, peptides identified, or proteins versus
peptide/protein ratio.
MatrixQCvis
provides an interactive shiny-based interface to assess data quality at various
processing steps (normalization, transformation, batch correction, and
imputation) of rectangular matrices. The package includes several diagnostic
plots and metrics such as barplots of intensity distributions, plots to
visualize drifts, MA plots and Hoeffding’s D value calculation, and
dimension reduction plots and provides specific tools to analyze data sets
containing missing values as commonly observed in mass spectrometry.
proBatch
enables to assess batch effects in (prote)omics data sets and corrects these
batch effects in subsequent steps. Several tools to visualize data quality are
included in the proBatch packages, such as barplots of intensity
distributions, cluster and heatmap analysis tools, and PCA dimension
reduction plots. Additionally, proBatch enables to assess diagnostics at
the feature level, e.g. peptides or spike-ins.
To install this package, start R and enter:
if (!requireNamespace("BiocManager", quietly = TRUE))
install.packages("BiocManager")
if (!requireNamespace("remotes", quietly = TRUE))
install.packages("remotes")
## to install from Bioconductor
BiocManager::install("MsQuality")
## to install from GitHub
BiocManager::install("tnaake/MsQuality")This will install this package and all eventually missing dependencies.
MsQuality is currently under active development. If you discover any bugs,
typos or develop ideas of improving MsQuality feel free to raise an issue
via GitHub or send a mail to the
developer.
Spectra and MsExperiment objectsLoad the Spectra package.
library("Spectra")
library("MsExperiment")
library("MsQuality")Spectra and MsExperiment objects from mzML filesThere are several options available to create a Spectra object. One way, as
outlined in the vignette of the Spectra package is
by specifying the location of mass spectrometry raw files in mzML, mzXML or
CDF format and using the MsBackendMzR backend. Here we load the example
files from the sciex data set of the msdata package and create a Spectra
object from the two provided mzML files. The example is taken from the
Spectra vignette.
## this example is taken from the Spectra vignette
fls <- dir(system.file("sciex", package = "msdata"), full.names = TRUE)
sps_sciex <- Spectra(fls, backend = MsBackendMzR())The data set consists of a single sample measured in two different
injections to the same LC-MS setup. An empty instance of an
MsExperiment object is created and populated with information on the samples
by assigning data on the samples (sampleData), information on the
mzML files (MsExperimentFiles) and spectral information (spectra).
In a last step, using linkSampleData, the relationships between the samples
and the spectral information are defined.
## this example is taken from the Spectra vignette
lmse <- MsExperiment()
sd <- DataFrame(sample_id = c("QC1", "QC2"),
sample_name = c("QC Pool", "QC Pool"),
injection_idx = c(1, 3))
sampleData(lmse) <- sd
## add mzML files to the experiment
experimentFiles(lmse) <- MsExperimentFiles(mzML_files = fls)
## add the Spectra object to the experiment
spectra(lmse) <- sps_sciex
## use linkSampleData to establish and define relationships between sample
## annotations and MS data
lmse <- linkSampleData(lmse, with = "experimentFiles.mzML_file",
sampleIndex = c(1, 2), withIndex = c(1, 2))Spectra and MsExperiment objects from (feature-extracted) intensity tablesAnother common approach is the creation of Spectra objects from a
DataFrames using the MsBackendDataFrame backend.
We will use here the data set of Lee et al. (2019), that contains metabolite level
information measured by reverse phase liquid chromatography (RPLC) coupled
to mass spectrometry and hydrophilic interaction liquid chromatography (HILIC)
coupled to mass spectrometry
(derived from the file STables - rev1.xlsx in the Supplementary Information).
In a separate step (see documentation for Lee2019_meta_vals and
Lee2019), we have created a list containing Spectra objects for
each samples (objects sps_l_rplc and sps_l_hilic) and MsExperiment
objects containing the data of all samples (objects msexp_rplc and
msexp_hilic). We will load here these objects:
data("Lee2019", package = "MsQuality")The final data set contains 541 paired samples (i.e. 541 samples derived from RPLC and 541 samples derived from HILIC).
We will combine the sps_rplc and sps_hilic objects in the following and
calculate on this combined document the metrics.
sps_comb <- c(sps_rplc, sps_hilic)The most important function to assess the data quality and to calculate the
metrics is the calculateMetrics function. The function takes
a Spectra or MsExperiment object as input, a character vector of metrics
to be calculated, and, optionally a list of parameters passed to the
quality metrics functions.
Spectra and MsExperiment objectsCurrently, the following metrics are included:
qualityMetrics(msexp_rplc)## [1] "rtDuration" "rtOverTicQuantiles"
## [3] "rtOverMsQuarters" "ticQuartileToQuartileLogRatio"
## [5] "numberSpectra" "medianPrecursorMz"
## [7] "rtIqr" "rtIqrRate"
## [9] "areaUnderTic" "areaUnderTicRtQuantiles"
## [11] "extentIdentifiedPrecursorIntensity" "medianTicRtIqr"
## [13] "medianTicOfRtRange" "mzAcquisitionRange"
## [15] "rtAcquisitionRange" "precursorIntensityRange"
## [17] "precursorIntensityQuartiles" "precursorIntensityMean"
## [19] "precursorIntensitySd" "msSignal10xChange"
## [21] "ratioCharge1over2" "ratioCharge3over2"
## [23] "ratioCharge4over2" "meanCharge"
## [25] "medianCharge"The following list gives a brief explanation for these metrics. Further
information may be found at the
HUPO-PSI mzQC project page or in the
respective help file for the quality metric (accessible by e.g. entering
?rtDuration to the R console). Currently, all quality metrics can be
calculated for both Spectra and MsExperiment objects.
mode = "TIC_change";The most important function to assess the data quality and to calculate the
metrics is the calculateMetrics function. The function takes
a Spectra or MsExperiment object as input, a character vector of metrics
to be calculated, and, optionally a list of parameters passed to the
quality metrics functions.
When passing a Spectra/MsExperiment object to the function, a data.frame
returned by calculateMetrics with the metrics specified by the argument
metrics. By default, qualityMetrics(object) is taken to specify
the calculation of quality metrics. calculateMetrics also accepts a list
of parameters passed to the individual quality metrics functions. For
each quality metrics functions, the relevant parameters are selected based on
the accepted arguments.
Additional arguments can be given to the quality metrics functions.
For example, the function ticQuartileToQuartileLogRatio function has the
arguments relativeTo, mode, and msLevel. relativeTo specifies to which
quantile the log TIC quantile is relatively related to (either to the 1st
quantile or the respective previous one). mode (either "TIC_change" or
"TIC") specifies if the quantiles are taken from the changes between TICs
of scan events or the TICs directly. One Spectra/MsExperiment object may
also contain more than one msLevel, e.g. if it also contains information on
MS\(^2\) or MS\(^3\) features. If the user adds the arguments
relativeTo = "Q1", mode = "TIC", msLevel = c(1L, 2L)),
ticQuartileToQuartileLogRatio is run with the parameter combinations
relativeTo = "Q1", mode = "TIC", msLevel = c(1L, 2L).
The results based on these parameter combinations are returned and the used parameters are returned as attributes to the returned vector.
Here, we would like to calculate the metrics of all included quality
metrics functions (qualityMetrics(object)) and additionally pass the
parameter relativeTo = "Q1" and relativeTo = "previous". For computational
reasons, we will restrict the calculation of the metrics to the first
sample and to RPLC samples.
## subset the Spectra objects
sps_comb_subset <- sps_comb[grep("Sample.1_", sps_comb$dataOrigin), ]
## for RPLC and HILIC
metrics_sps_Q1 <- calculateMetrics(object = sps_comb_subset,
metrics = qualityMetrics(sps_comb_subset),
relativeTo = "Q1", msLevel = 1L)
metrics_sps_Q1## rtDuration rtOverTicQuantiles.0% rtOverTicQuantiles.25%
## Sample.1_RPLC 18.214 0 0.08098166
## Sample.1_HILIC 16.000 0 0.34375000
## rtOverTicQuantiles.50% rtOverTicQuantiles.75%
## Sample.1_RPLC 0.08098166 0.1495004
## Sample.1_HILIC 0.34375000 0.5375000
## rtOverTicQuantiles.100% rtOverMsQuarters.Quarter1
## Sample.1_RPLC 1 0.02893379
## Sample.1_HILIC 1 0.15625000
## rtOverMsQuarters.Quarter2 rtOverMsQuarters.Quarter3
## Sample.1_RPLC 0.08806413 0.3102009
## Sample.1_HILIC 0.47500000 0.6216875
## rtOverMsQuarters.Quarter4 ticQuartileToQuartileLogRatio.Q2/Q1
## Sample.1_RPLC 1 NaN
## Sample.1_HILIC 1 -Inf
## ticQuartileToQuartileLogRatio.Q3/Q1
## Sample.1_RPLC NaN
## Sample.1_HILIC NaN
## ticQuartileToQuartileLogRatio.Q4/Q1 numberSpectra
## Sample.1_RPLC NaN 190
## Sample.1_HILIC NaN 165
## medianPrecursorMz rtIqr rtIqrRate areaUnderTic
## Sample.1_RPLC 198.05 5.10125 18.42686 655624525
## Sample.1_HILIC 179.10 7.44700 11.14543 149945016
## areaUnderTicRtQuantiles.25% areaUnderTicRtQuantiles.50%
## Sample.1_RPLC 25612593.2 389734297
## Sample.1_HILIC 927493.9 100961764
## areaUnderTicRtQuantiles.75% areaUnderTicRtQuantiles.100%
## Sample.1_RPLC 233673968 5692460
## Sample.1_HILIC 40996727 5460065
## extentIdentifiedPrecursorIntensity medianTicRtIqr
## Sample.1_RPLC 36467.884 7496.006
## Sample.1_HILIC 8713.906 2195.400
## medianTicOfRtRange mzAcquisitionRange.min mzAcquisitionRange.max
## Sample.1_RPLC 13858.935 60.1 1377.6
## Sample.1_HILIC 2086.417 73.0 784.1
## rtAcquisitionRange.min rtAcquisitionRange.max
## Sample.1_RPLC 0.986 19.2
## Sample.1_HILIC 1.100 17.1
## precursorIntensityRange.min precursorIntensityRange.max
## Sample.1_RPLC 100.6492 349282909
## Sample.1_HILIC 100.0895 92021751
## precursorIntensityQuartiles.Q1 precursorIntensityQuartiles.Q2
## Sample.1_RPLC 1667.3911 6746.195
## Sample.1_HILIC 300.5038 2304.383
## precursorIntensityQuartiles.Q3 precursorIntensityMean
## Sample.1_RPLC 75003.90 3450655.4
## Sample.1_HILIC 33382.25 908757.7
## precursorIntensitySd msSignal10xChange ratioCharge1over2
## Sample.1_RPLC 26563228 56 NaN
## Sample.1_HILIC 7729451 47 NaN
## ratioCharge3over2 ratioCharge4over2 meanCharge medianCharge
## Sample.1_RPLC NaN NaN NaN NA
## Sample.1_HILIC NaN NaN NaN NA
## attr(,"relativeTo")
## [1] "Q1"
## attr(,"msLevel")
## [1] 1metrics_sps_previous <- calculateMetrics(object = sps_comb_subset,
metrics = qualityMetrics(sps_comb_subset),
relativeTo = "previous", msLevel = 1L)
metrics_sps_previous## rtDuration rtOverTicQuantiles.0% rtOverTicQuantiles.25%
## Sample.1_RPLC 18.214 0 0.08098166
## Sample.1_HILIC 16.000 0 0.34375000
## rtOverTicQuantiles.50% rtOverTicQuantiles.75%
## Sample.1_RPLC 0.08098166 0.1495004
## Sample.1_HILIC 0.34375000 0.5375000
## rtOverTicQuantiles.100% rtOverMsQuarters.Quarter1
## Sample.1_RPLC 1 0.02893379
## Sample.1_HILIC 1 0.15625000
## rtOverMsQuarters.Quarter2 rtOverMsQuarters.Quarter3
## Sample.1_RPLC 0.08806413 0.3102009
## Sample.1_HILIC 0.47500000 0.6216875
## rtOverMsQuarters.Quarter4 ticQuartileToQuartileLogRatio.Q2/Q1
## Sample.1_RPLC 1 NaN
## Sample.1_HILIC 1 -Inf
## ticQuartileToQuartileLogRatio.Q3/Q2
## Sample.1_RPLC 5.831233
## Sample.1_HILIC Inf
## ticQuartileToQuartileLogRatio.Q4/Q3 numberSpectra
## Sample.1_RPLC 8.915513 190
## Sample.1_HILIC 8.982638 165
## medianPrecursorMz rtIqr rtIqrRate areaUnderTic
## Sample.1_RPLC 198.05 5.10125 18.42686 655624525
## Sample.1_HILIC 179.10 7.44700 11.14543 149945016
## areaUnderTicRtQuantiles.25% areaUnderTicRtQuantiles.50%
## Sample.1_RPLC 25612593.2 389734297
## Sample.1_HILIC 927493.9 100961764
## areaUnderTicRtQuantiles.75% areaUnderTicRtQuantiles.100%
## Sample.1_RPLC 233673968 5692460
## Sample.1_HILIC 40996727 5460065
## extentIdentifiedPrecursorIntensity medianTicRtIqr
## Sample.1_RPLC 36467.884 7496.006
## Sample.1_HILIC 8713.906 2195.400
## medianTicOfRtRange mzAcquisitionRange.min mzAcquisitionRange.max
## Sample.1_RPLC 13858.935 60.1 1377.6
## Sample.1_HILIC 2086.417 73.0 784.1
## rtAcquisitionRange.min rtAcquisitionRange.max
## Sample.1_RPLC 0.986 19.2
## Sample.1_HILIC 1.100 17.1
## precursorIntensityRange.min precursorIntensityRange.max
## Sample.1_RPLC 100.6492 349282909
## Sample.1_HILIC 100.0895 92021751
## precursorIntensityQuartiles.Q1 precursorIntensityQuartiles.Q2
## Sample.1_RPLC 1667.3911 6746.195
## Sample.1_HILIC 300.5038 2304.383
## precursorIntensityQuartiles.Q3 precursorIntensityMean
## Sample.1_RPLC 75003.90 3450655.4
## Sample.1_HILIC 33382.25 908757.7
## precursorIntensitySd msSignal10xChange ratioCharge1over2
## Sample.1_RPLC 26563228 56 NaN
## Sample.1_HILIC 7729451 47 NaN
## ratioCharge3over2 ratioCharge4over2 meanCharge medianCharge
## Sample.1_RPLC NaN NaN NaN NA
## Sample.1_HILIC NaN NaN NaN NA
## attr(,"relativeTo")
## [1] "previous"
## attr(,"msLevel")
## [1] 1Alternatively, an MsExperiment object might be passed to
calculateMetrics. The function will iterate over the samples (referring
to rows in sampleData(msexp))) and calculate the quality metrics on the
corresponding Spectras.
There are in total 541 samples
respectively in the objects msexp_rplc and msexp_hilic. To improve
the visualization and interpretability, we will only calculate the metrics
from the first 20 of these samples.
## subset the MsExperiment objects
msexp_rplc_subset <- msexp_rplc[1:20]
msexp_hilic_subset <- msexp_hilic[1:20]
## define metrics
metrics_sps <- c("rtDuration", "rtOverTicQuantiles", "rtOverMsQuarters",
"ticQuartileToQuartileLogRatio", "numberSpectra", "medianPrecursorMz",
"rtIqr", "rtIqrRate", "areaUnderTic")
## for RPLC-derived MsExperiment
metrics_rplc_msexp <- calculateMetrics(object = msexp_rplc_subset,
metrics = qualityMetrics(msexp_rplc_subset),
relativeTo = "Q1", msLevel = 1L)
## for HILIC-derived MsExperiment
metrics_hilic_msexp <- calculateMetrics(object = msexp_hilic_subset,
metrics = qualityMetrics(msexp_hilic_subset),
relativeTo = "Q1", msLevel = 1L)When passing an MsExperiment object to calculateMetrics a data.frame
object is returned with the samples (derived from the rownames of
sampleData(msexp)) in the rows and the metrics in columns.
We will show here the objects metrics_rplc_msexp and metrics_hilic_msexp
## [1] "metrics_rplc_msexp"## [1] "metrics_hilic_msexp"The quality metrics can be most easily compared when graphically visualized.
The MsQuality package offers the possibility to graphically display the
metrics using the plotMetric and shinyMsQuality functions. The
plotMetric function will create one plot based on a single metric.
shinyMsQuality, on the other hand, opens a shiny application that allows
to browse through all the metrics stored in the object.
As a way of example, we will plot here the number of features. A high number of missing features might indicate low data quality, however, also different sample types might exhibit contrasting number of detected features. As a general rule, only samples of the same type should be compared to adjust for sample type-specific effects.
metrics_msexp <- rbind(metrics_rplc_msexp, metrics_hilic_msexp)
plotMetric(qc = metrics_msexp, metric = "numberSpectra")Similarly, we are able to display the area under the TIC for the retention time quantiles. This plot gives information on the perceived signal (TIC) for the differnt retention time quantiles and could indicate drifts or interruptions of sensitivity during the run.
plotMetric(qc = metrics_msexp, metric = "ticQuartileToQuartileLogRatio")Alternatively, to browse through all metrics that were calculated in an
interactive way, we can use the shinyMsQuality function.
shinyMsQuality(qc = metrics_msexp)All software and respective versions to build this vignette are listed here:
## R version 4.3.0 RC (2023-04-13 r84269)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 22.04.2 LTS
##
## Matrix products: default
## BLAS: /home/biocbuild/bbs-3.17-bioc/R/lib/libRblas.so
## LAPACK: /usr/lib/x86_64-linux-gnu/lapack/liblapack.so.3.10.0
##
## locale:
## [1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
## [3] LC_TIME=en_GB LC_COLLATE=C
## [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: America/New_York
## tzcode source: system (glibc)
##
## attached base packages:
## [1] stats4 stats graphics grDevices utils datasets methods
## [8] base
##
## other attached packages:
## [1] MsExperiment_1.2.0 Spectra_1.10.0 ProtGenerics_1.32.0
## [4] BiocParallel_1.34.0 S4Vectors_0.38.0 BiocGenerics_0.46.0
## [7] MsQuality_1.0.0 knitr_1.42 BiocStyle_2.28.0
##
## loaded via a namespace (and not attached):
## [1] tidyselect_1.2.0 viridisLite_0.4.1
## [3] dplyr_1.1.2 bitops_1.0-7
## [5] fastmap_1.1.1 lazyeval_0.2.2
## [7] RCurl_1.98-1.12 promises_1.2.0.1
## [9] digest_0.6.31 mime_0.12
## [11] lifecycle_1.0.3 cluster_2.1.4
## [13] ellipsis_0.3.2 magrittr_2.0.3
## [15] compiler_4.3.0 rlang_1.1.0
## [17] sass_0.4.5 tools_4.3.0
## [19] igraph_1.4.2 utf8_1.2.3
## [21] yaml_2.3.7 data.table_1.14.8
## [23] labeling_0.4.2 htmlwidgets_1.6.2
## [25] curl_5.0.0 DelayedArray_0.26.0
## [27] RColorBrewer_1.1-3 withr_2.5.0
## [29] purrr_1.0.1 grid_4.3.0
## [31] fansi_1.0.4 xtable_1.8-4
## [33] colorspace_2.1-0 ggplot2_3.4.2
## [35] scales_1.2.1 MASS_7.3-59
## [37] MultiAssayExperiment_1.26.0 SummarizedExperiment_1.30.0
## [39] cli_3.6.1 rmarkdown_2.21
## [41] generics_0.1.3 httr_1.4.5
## [43] cachem_1.0.7 stringr_1.5.0
## [45] zlibbioc_1.46.0 parallel_4.3.0
## [47] AnnotationFilter_1.24.0 BiocManager_1.30.20
## [49] XVector_0.40.0 matrixStats_0.63.0
## [51] vctrs_0.6.2 Matrix_1.5-4
## [53] jsonlite_1.8.4 bookdown_0.33
## [55] IRanges_2.34.0 clue_0.3-64
## [57] crosstalk_1.2.0 plotly_4.10.1
## [59] jquerylib_0.1.4 tidyr_1.3.0
## [61] glue_1.6.2 codetools_0.2-19
## [63] QFeatures_1.10.0 stringi_1.7.12
## [65] gtable_0.3.3 later_1.3.0
## [67] GenomeInfoDb_1.36.0 GenomicRanges_1.52.0
## [69] shinydashboard_0.7.2 munsell_0.5.0
## [71] tibble_3.2.1 pillar_1.9.0
## [73] htmltools_0.5.5 GenomeInfoDbData_1.2.10
## [75] R6_2.5.1 evaluate_0.20
## [77] shiny_1.7.4 lattice_0.21-8
## [79] Biobase_2.60.0 httpuv_1.6.9
## [81] bslib_0.4.2 Rcpp_1.0.10
## [83] xfun_0.39 MsCoreUtils_1.12.0
## [85] fs_1.6.2 MatrixGenerics_1.12.0
## [87] pkgconfig_2.0.3Lee, H.-J., D. M. Kremer, P. Sajjakulnukit, L. Zhang, and C. A. Lyssiotis. 2019. “A Large-Scale Analysis of Targeted Metabolomics Data from Heterogeneous Biological Samples Provides Insights into Metabolite Dynamics.” Metabolomics, 103. https://doi.org/10.1007/s11306-019-1564-8.