A package created by William Edmands. Supported now by Josie Hayes.
Development version 0.2.1 archived on the Zenodo repository.
Performs simultaneous data conversion, preprocessing and statistical analysis and instrument stoppage/signal drift monitoring throughout the course of a metabolomic profiling experiment. The simExTargId package performs raw data to mzXML conversion (MSConvert), peak-picking, retention time alignment, grouping (xcms), ESI adducts/in-source fragment artefact identification (CAMERA), peak table output pre-processing, automatic outlier detection by PCA and automatic statistical analysis dependent on co-variates supplied (MetMSLine). Signal drift monitoring and possible MS2 target list identification is facilitated during a metabolomic profiling experiment by two shiny user interfaces. An email notification system using the sendmailR package is implemented in simExTargId to warn the user(s) of instrumental stoppages, outlying quality control samples and signal attenuation/drift. The simExTargId package is still in active development but nevertheless should still be at a functional and useful stage for many metabolomic investigators.
The latest development version and all package dependencies can be installed with one-line of code directly from GitHub using the devtools package. First ensure devtools is installed, instructions can be found here: https://github.com/hadley/devtools
devtools::install_github('JosieLHayes/simExTargId', dependencies=c("Depends", "Imports", "Suggests"))
The vignette can be viewed here with example raw data acquired on a Thermo FT-ICR mass spectrometer (this requires playing with the creation time of the raw data files to simulate a real-time data collection).
The simExTargId package has thus far only been tested with Agilent (.d) and Thermo (.RAW/.raw) data files on a computer running Windows but depending on interest could be readily extended to other instrument manufacturers and may well already function on an OSX or Linux based operating system.
The R package utilizes MSConvert (ProteoWizard), xcms, CAMERA, MetMSLine, Shiny and many other packages to attempt to implement real-time simultaneous metabolomic MS1-profiling Experiment and statistically relevant MS2 Target Identification.
A major impetus for development of this package was to provide an email-based early warning system for LC-MS instrumental stoppages/errors but also for more subtle changes such as instrument drift and (PCA-based) outlying pooled quality control samples for example. When collecting a dataset of precious/limited samples such as those with a low volume/quantity provided (e.g. mouse sera, dried blood spot punches) it can be particularly poignant if un-noticed instrumental stoppages leads to degradation of your samples or unwanted variation/batch effects.
Instrumental drifts and potentially MS2 targets can be identified in real time using the two shiny apps developed for simExTargId peakMonitor
and targetId
.
Narrowing the temporal gap to truly "online" target feature MS2 fragmentation is also a primary goal of simExTargId.
All peak-picking, retention time alignment and grouping is performed by xcms, then ESI adducts and isotopes are detected by CAMERA in real time. Following this pre-processing of the peak-table is performed by MetMSLine. Automatic PCA-based outlier detection is performed and a warning email sent to user(s) if a QC sample is outlying. Furthermore, automatic real-time univariate co-variate based statistical analysis is then performed.
Example PCA scores iterations 1 and 2 - averaged pooled Quality Control samples should cluster tightly near the origins of the first and second principal components
The simExTargId workflow can be left running during the MS1-profiling data collection providing a degree of reassurance that serious instrumental difficulties will not go unnoticed (even whilst present in the laboratory and otherwise distracted) and also outlying samples and statistically relevant LC-MS feature targets can be identified and additional experiments/reinjections appended to the experimental worklist before the end.
Example signal attenuation warning email - pooled quality controls can be used to monitor a database (.csv) of previously identified metabolites if a maximum percentage (>20%) of the monitored metabolites are greater than 20% attenuated (these defaults can be altered by user) then a warning email is sent. The output of the peakMonitor
function can be visualized in a shiny application.
SimExTargId also attempts to inculcate a rigorous (and broadly accepted) experimental design and also a directory sub-structure for each experiment to help organize data and files for the user. Sub-directories, tables and plots are then generated in real time. For example, each time a new iteration of simExTargId takes place updated PCA plots and tables will appear in each sub-directory. The idea is to preserve the reproducibility by the recording of parameters used (and session information) and providing an easy to navigate and intuitive sub-directory structure.
Example sub-directory structure generated by simExTargId upon initiation - the "data" directory should contain files which should not be modified such as the worklist/co-variates table received from the collaborators or the worklist file for example, the "doc" directory should contain all documentation i.e. manuscript files and method details, the "output" directory contains all output generated by simExTargId or by R code created after data collection, the "R" directory is used for all R code and workspace (.RData/.RDS) files. All mzXML files are saved in a separate directory due to their large size and to keep the size of analysis directory as small as possible.
A technical note describing the simExTargId package has been pre-published on the site BioRxiv: simExTargId: An R package for real-time LC-MS metabolomic data analysis, instrument failure/drift notification and MS2 target identification William M. B. Edmands, Stephen M. Rappaport bioRxiv 151159; doi: https://doi.org/10.1101/151159
Or please cite the DOI provided by the Zenodo archive.
The simExTargId package is licenced under the GPLv3 (http://www.gnu.org/licenses/gpl.html).