Skip to contents

Changelog

Source: NEWS.md

patRoon 3.0

This release adds a significant amount of new functionality and changes. Please see the updated Handbook, pre-print, and sections below for more information. Users of previous patRoon versions should inform themselves with the important changes highlighted in the next section.

Upgrading notes

This releases adds changes to various functions to improve consistency and support new functionality. While care was taken to avoid breaking changes, some changes will affect existing workflows. Below is a summary of important points.

  • It is important to clear the cache when updating to this release. This can be done by running clearCache("all") or manually removing the cache.sqlite file from your project directory.

  • The analysis information table format was changed to support new functionality, file formats and cleanup.

    1. Rename the "group" column to "replicate"
    2. Adjust the path column to the new format: path_raw, path_centroid, path_profile and path_ims. Often, renaming path to path_centroid is sufficient.

  • The following packages should be updated or installed

    • patRoonData: to ensure compatibility with patRoon 3.0, e.g. remotes::install_github("rickhelmus/patRoonData").
    • patRoonExt: to ensure the latest versions of external dependencies and availability of IMS specifics, e.g. remotes::install_github("rickhelmus/patRoonExt").
    • Rmstoolkitlib: this is needed if you want to use the new mstoolkit raw data backend. This should be installed before installing patRoon, e.g. remotes::install_github("rickhelmus/Rmstoolkitlib").
    • patRoonDataIMS: to install demo IMS-HRMS data, e.g. remotes::install_github("rickhelmus/patRoonDataIMS").
    • See the Installation chapter in the updated Handbook for more details and installation options.

  • The generateAnalysisInfo() function was updated for the new format. The fromXXX function arguments need to be set properly. See ?generateAnalysisInfo for more details.

  • The convertMSFiles() interface has various small changes to simplify its usage and extend support. For instance:

    # pre 3.0
convertMSFiles(anaInfo = anaInfo, from = "bruker", to = "mzML", algorithm = "pwiz", centroid = "vendor",
               overWrite = FALSE)
# 3.0
convertMSFiles(anaInfo, typeFrom = "raw", formatFrom = "bruker", typeTo = "centroid", formatTo = "mzML",
               algorithm = "pwiz", overwrite = FALSE)

    Please see ?convertMSFiles for more details.

  • MS peak list generation: the previous backends (“algorithms”) are now deprecated and no algorithm argument should be specified. The interface was also simplfied. See (?generateMSPeakLists for details). E.g.:

    # pre 3.0
mslists <- generateMSPeakLists(fGroups, "mzr", maxMSRtWindow = 5, precursorMzWindow = 4, avgFeatParams = ...,
                               avgFGroupParams = ...)
# 3.0
mslists <- generateMSPeakLists(fGroups, maxMSRTWindow = 5, avgFeatParams = ..., avgFGroupParams = ...)

  • The filtering of MS peak lists is simplified and is now done per MS level (see ?MSPeakLists for details):

    # pre 3.0
mslists <- filter(mslists, relMSMSIntThr = 0.05, topMSMSPeaks = 25, ...)
# 3.0
mslists <- filter(mslists, msLevel = 2, relMinIntensity = 0.05, topMost = 25, ...)

  • The estimation of identification confidence levels was extended beyond suspect screening workflows. The annotateSuspects() function was renamed to estimateIDConfidence(). This function should also be used to estimate ID levels for feature annotation candidates. See ?estimateIDConfidence for details. Eg:

    # pre 3.0
fGroups <- annotateSuspects(fGroups, formulas = formulas, compounds = compounds, MSPeakLists = mslists,
                            IDFile = "idlevelrules.yml")
# 3.0
formulas <- estimateIDConfidence(formulas, IDFile = "idlevelrules.yml")
compounds <- estimateIDConfidence(compounds, MSPeakLists = mslists, formulas = formulas, IDFile = "idlevelrules.yml")
fGroups <- estimateIDConfidence(fGroups, formulas = formulas, compounds = compounds, MSPeakLists = mslists,
                                IDFile = "idlevelrules.yml")

  • The default rules for ID level estimation was changed for level 3a. Either re-generate the file by running genIDLevelRulesFile("idlevelrules.yml") or update the existing file by changing the individualMoNAScore and libMatch thresholds to 0.7.

  • The report configuration file was updated. Run the following to update an existing file:

    genReportSettingsFile("report.yml", baseFrom = "report.yml")

  • The default numeric limits were centralized (discussed further below). To generate a new file for configuration:

    See ?limits for details.

  • Several defaults and names for function arguments were changed (discussed further below).

Please read the remaining sections below and updated documentation for more details and changes. Furthermore, it may be useful to compare the script output of newProject() to see how these changes may affect your workflow. Finally, feel free to contact or create a GitHub issue when in doubt.

Major new functionality

IMS workflows

The 3.0 release is primarily centered around supporting ion mobility separation (IMS) coupled to HRMS workflows. This technique brings several important benefits for NTA, such as an additional parameter for feature annotation (collision cross section or CCS), cleaning up of HRMS data to improve feature detection and annotation, improved separation of compounds with equal or close m/z and support of rapid and selective MS/MS (PASEF). The complete workflow was upgraded to support IMS-HRMS data and make use of the benefits from this technique. Please see the new IMS chapter in the Handbook for further details.

msdata interface

The new msdata raw data interface was introduced to efficiently read complex raw data. The primary motivation was to support the complexity and specific formatting of IMS-HRMS data, but the applied optimizations also benefit classical HRMS workflows. More information on msdata and its configuration is detailed in the Handbook and reference manual (?msdata).

NOTE: due to (slight) changes rounding numerical data there may be slight differences in EICs, spectra and feature data compared to the old interface.

piek and greedy algorithms

This release adds the newly developed piek algorithm, which is a simple, flexible and fast algorithm for detecting features in both HRMS and IMS-HRMS data. Furthermore, the greedy algorithm was added to provide fast grouping of features across samples and also supports both HRMS and IMS-HRMS data. Both algorithms are embedded in patRoon 3.0 and therefore don’t need any installation of external software. See the Handbook and reference manual (?findFeaturesPiek and ?groupFeaturesGreedy) for further details.

Sample metadata and analysis information

A lot of existing functionality was upgraded to support the use of sample metadata in post-processing operations, such as easily making sample groups from metadata for plotting and aggregation of data. Furthermore, the (previously very basic) support for applying linear regression to eg prioritize feature data was improved considerably, and was primarily motivated to improve feature prioritization of unknown transformation products (see 10.1021/acs.est.4c09121). Finally, functionality was added to easily subset features from metadata and update the analysis information in feature objects, e.g. to add or modify sample metadata. The Handbook was updated with a new subsection to demonstrate the use of sample metadata. Furthermore, see below for more details which functions were changed.

Transformation products

Two new algorithms were added for finding transformation products: "ann_form" (generateTPsAnnForm()) and "ann_comp" (generateTPsAnnComp()). These algorithms use data from a thorough feature annotation of unknown features to find candidate TPs and are eg scored on their ‘fit’ to given parents. The algorithms were based on the work described in 10.1021/acs.est.4c09121.

Other new functionality

Importing feature data

Support was added to import features from external data sources (which has been requested for a very long time!). The new functions importFeaturesTable() and importFeaturesGroupsTable() provide a way to import tabular data with raw or sample grouped features, respectively. See the updated Handbook and reference manual (?importFeaturesTable and ?importFeatureGroupsTable) for more details.

New data filters

Several new data filters were added throughout the workflow:

  • absMinMaxIntensity and relMinMaxIntensity: filter features by their maximum intensity across samples (suggested by Nienke Meekel).
  • Peak abundance filters for MS peak lists to eliminate low abundance peaks across averaged spectra.
  • minRelCumIntensity spectrum averaging and filter() parameter to filter MS peaks by their cumulative intensity in the spectrum (suggested by Leon Saal, issue #150).
  • maxMZOverPrec: eliminate peaks with higher m/z values than precursors in MS2 data.
  • fragFormulas and lossFormulas: filter candidates by matching fragment or neutral loss formulae.

MS2 background subtraction and ID confidence estimation for unknowns

This release adds approaches to subtract MS2 background and estimate identification confidence levels unknowns, as described in 10.1021/acs.est.4c09121. The latter functionality allows the estimation of ID levels for formula and compound candidates obtained during feature annotation. The approach is similar (but optimized) compared to the previously introduced system for suspect screening workflows. Similar filters existing to remove candidates with poor ID levels. See the updated Handbook and reference manual (?getBGMSMSPeaks and ?estimateIDConfidence) for more details.

Improvements for reporting and project creation

The code behind newProject() and report() was considerably optimized and improved. The newProject() function was updated to support all the new functionality introduced in patRoon 3.0. The reporting interface was improved in various ways, and the output files are smaller due to lzstring compression and optimizations were applied for self-contained report files.

Updating feature group properties

The new updateGroups() method function re-calculates feature group properties (RT, m/z and mobility). This function is primarily intended to be used after filtering steps of feature data. Re-calculating aforementioned properties can then be improved by exclusion of e.g. noisy features.

Important changes

Centralized default numeric limits

The defaults for common numerical limits and thresholds (retention time, m/z and mobility) are now moved to a separate ‘limits file’. This simplifies the configuration of these parameters and avoids the need to configure them throughout all functions in the workflow. As a result, some default values may have been slightly changed for consistency. See the Limits section in the updated Handbook for more information.

Changed defaults

  • generateFormulas(): calculateFeatures now defaults to FALSE: this usually yields equally valid results and is generally much faster.
  • generateCompoundsMetFrag()
    • the default database was changed from PubChem to PubChemLite, which is much faster and generally more suitable for non-target analysis.
    • MetFusion scoring is not used anymore by default. The scoreTypes function argument can be configured to include it again.
  • generateCompoundsLibrary()
    • specSimParamsLib now defaults to the value for specSimParams
    • specSimParams now defaults to excluding the precursor mass peak
  • Identification confidence levels: increased 3a library match threshold to a more reasonable value (from 0.4 to 0.7)

Changed naming

  • The use of “replicate groups” was changed to “replicates” throughout the whole code and documentation for consistency and clarity. Some important changes include
    • The group column in the analysis information should now be named replicate. If only the old column name is present it will still be used, with a warning.
    • replicateGroups() was renamed to replicates()
    • replicateGroupSubtract() was renamed to replicateSubtract()
    • The rGroups argument to [ and filter() was renamed to replicates.
    • The topMostByRGroup EIC parameter was renamed to topMostByReplicate
  • The annotateSuspects() function was renamed to estimateIDConfidence() for consistency with ID level estimation functionality for formulae and compounds.
  • overWrite –> overwrite for consistency, applies to predictCheckFeaturesSession() and file conversion functions
  • generateMSPeakLists(): maxMSRtWindow was renamed to maxMSRTWindow for consistency
  • plotInt(): xnames –> xNames
  • EIC parameters (getDefEICParams()): retWindow –> window
  • log P values are now consistently stored in "logP" columns in results for compound annotations and transformation products.

Analysis information and use of sample metadata

Several changes were made to the analysis information tables to improve support for different file formats and allow the use of sample metadata (as introduced before). See the updated Handbook for more details.

  • The path column is replaced by four different columns: path_raw, path_centroid, path_profile and path_ims, which can be used to specify the paths to different types of data. If only a path column is present, it will be used as path_centroid with a warning.
  • The group column should now be named replicate. If only a group column is present, it will be used as replicate with a warning.
  • Any additional columns with free-form data can be included and can be used to post-process feature data (see below).
  • A new function was added to adjust analysis information during the workfow (analysisInfo()<-).
  • Similarly, the new reorder argument to [ can be used to reorder the sample analyses, which may e.g. be useful for plotting.
  • The generateAnalysisInfo() function was updated for the changes: please see ?generateAnalysisInfo for its new interface.
  • Using sample metadata for grouping and aggregation
    • New function arguments to group & aggregate data (e.g. groupBy, aggregate) were added to unique(), overlap() and plotting functions.
    • In some cases these replace older function arguments
    • as.data.table()
      • the average argument can now be used to average by grouping specified by metadata columns.
      • average can now be set to fGroups to average all properties for each feature group (works if features=TRUE or features=FALSE).
      • The anaInfoCols argument can be used to add metadata to the output table.
  • The subset operator for features and feature groups ([) now supports subsetting by sample metadata and other data in the analysis information with the new ni argument. This approach was based on how data is subset with data.table.
  • Using sample metadata for linear regression of feature intensities
    • The as.data.table() and plotInt() functions were updated to improve support for linear regression. These functions can now use any sample metadata column(s) to easily perform a linear regression versus feature intensities/areas.
    • plotInt(): see the xBy, groupBy and regression function arguments.
    • as.data.table(): see the regressionBy function argument
    • as.data.table(): the conc_reg column was renamed to x_reg.
  • Internal changes
    • the analysis information is now internally stored as a data.table for efficiency.
    • the featureGroups objects do not store analysis information internally anymore, but instead refer to the analysis information stored in the respective features object.

MS peak lists

  • The generateMSPeakLists() function now uses the msdata interface to read raw data. The old backends (mzR and DataAnalysis) are now deprecated and should not be used anymore. For this reason, the algorithm function argument from generateMSPeakLists() was removed and the generateMSPeakListsMzR(), generateMSPeakListsDA(), generateMSPeakListsDAFMF() functions are now marked as deprecated.
  • The precursorMzWindow is replaced by the fixedIsolationWidth function argument, with slightly different behavior. In most cases fixedIsolationWidth=FALSE is recommended: this automatically detects the used isolation window for DDA and disables it for DIA.
  • The averaging of spectra is now intensity weighted.
  • The parameters to average spectra were changed and extended (see ?getDefAvgPListParams)
    • The default method was changed from hierarchical clustering ("hclust") to distance based averaging ("distance_mean"), which is faster and more suitable for large datasets such as IMS data.
    • The "distance" averaging method was renamed to "distance_point".
    • The absMinAbundance and relMinAbundance parameters were added to filter MS peaks by their abundance in the averaged spectra.
    • Several IMS specific parameters were added.

Other

  • File conversion & raw data
  • Features
    • The importFeatures() and importFeatureGroups() functions are more consistent in function argument order and supported input data sources. See the updated documentation (?importFeatures and ?importFeatureGroups) for more details.
    • SAFD algorithm (findFeaturesSAFD())
      • The msdata interface is now used to load raw data instead of interfacing with the MS_import.jl package. As a result:
        • MS_import.jl does not need to be installed anymore.
        • Reading data is expected to be faster.
        • This adds support for other data formats such as mzML.
        • Very basic support for files with ion mobility data: these can be used for feature detection, but the IMS dimension will be ignored.
      • The profPath function argument is removed, as this information should now be specified in the analysis information (see above).
      • Added prefCentroid, centroidMethod and centroidDM function arguments
    • Suspect screening+sets workflows: the set aggregated annotation similarities are now reported as max value instead of mean.
    • Internal changes
      • The groupInfo slot of featureGroups is now stored as a data.table to improve efficiency.
  • Feature annotation
    • generateFormulasDA() is now deprecated and should not be used anymore. Use GenForm or SIRIUS for formula generation.
    • estimateIDConfidence() (previously annotateSuspects()) now copies annotation similarities from feature annotations instead of calculating them from scratch. As a result:
      • The specSimParams argument was removed.
      • IMPORTANT: estimateIDConfidence() should be used to calculate annotation similarities for formulas and compounds objects in advance.
      • See the updated Handbook section on identification confidence estimation for more details.
  • Transformation products
    • generateComponentsTPs(): the internal representation of results are stored differently.
      • The component result table contains for each row a result for one feature group (similar to other components results).
      • For TP algorithms that work with possibly multiple candidates for a feature group, the candidates are stored in a separate table in the candidates column.
      • The as.data.table() function for TP components does not include candidates anymore by default: set candidates=TRUE to include them.
    • generateTPs()
      • consistent calculation and configuration of log P / retDir calculations for all algorithms. Hence, results may differ compared to earlier patRoon versions.
      • Advanced parameters for calculation of log P values, structural similarity etc should now be configured through the TPStructParams function argument (see ?getDefTPStructParams).
      • The retention direction calculation now also takes into account a minimum log P difference (besides minimum retention time). This will typically lead to different (but more reasonable) results. This can be disabled through the aforementioned TPStructParams.

Minor changes and new functionality

  • File conversion & raw data
    • Additional checks are performed to verify paths to raw instrument data.
    • convertMSFilesPWiz(): minIntensity function argument to eliminate low intensity mass peaks and therefore reduce file size.
    • getBPCs(): don’t return m/z information anymore. This was due to the change to use the msdata interface, which does not (yet) provide this information.
    • patRoon.checkCentroided option is removed: checks are much faster now and part of regular file reading operations.
  • Features
    • Made feature qualities more configurable and extendable (suggested by Leon Saal, issue #133/PR #134)
    • The delete() method function for suspect screening results can now remove suspect hits through the k function argument.
    • Loading OpenMS peak intensities is now much faster and removed now unneeded intSearchRTWindow function argument.
    • Feature ID column is now always of character type.
    • Optimized suspect screening.
    • The unique and total suspects/suspect hits are now (consistently) printed.
    • plotInt(): added areas, averageFunc, xlim and ylim function arguments.
    • plotChroms3D() function to plot 3D feature data (retention, intensity and m/z or mobility).
    • filter(): removeISTDs and onlyHits are now cached like other filters.
    • as.data.table()
      • Columns with text data are now collapsed instead of removed if features=TRUE and results are averaged.
      • Fold change calculation is now possible if features==TRUE.
      • The adduct column is split per set and renamed to group_adduct if features=TRUE
      • The intensity and area columns are now suffixed (_intensity/_area) if features=FALSE.
      • added susp_bestEstIDLevel column.
      • moved all docs to new page with all as.data.table() methods: see ?`feature-table`.
    • overlap(): which can be set to NULL to consider all compared data groups.
    • Gathering of feature data for EIC construction was considerably optimized.
    • getPICSet() can now import features from non-centroid or IMS data
    • getPICSet() and calculatePeakQualities() now gain EICParams function arguments to configure EIC construction.
    • A new ion_mz column is added to feature tables and annotations, which represents the original m/z of the ion in sets workflows.
    • The class for suspect screening+sets workflows (featureGroupsScreeningSet) doesn’t have setObjects anymore, which makes common operations faster
    • Changes related to support IMS workflows
      • the features class gained hasIMS and fromIMS slots and method accessors.
      • Several new advanced and IMS specific EIC parameters: see ?getDefEICParams.
      • plotChroms(): support annotation of mobility values (besides RT and m/z).
      • IMS argument for many functions ([, filter(), plotting functions, as.data.table(), …) which configures the usage of features with and/or without IMS data.
      • filter(): applyIMS to selectively apply filtering to features with or without IMS data.
      • filter(): withIMSPrecursor filter to only keep features with IMS precursors.
  • Feature annotation
  • Components
    • components from nontarget: renamed “rt” column to “ret” for consistency.
  • Transformation products
    • generateComponentsTPs()
      • slightly changed function argument order.
      • The fragment and neutral loss matches are now calculated for all candidates (as done previously) and per candidate.
      • filter() can now filter fragment and neutral loss matches per candidate.
      • The componentsTPs gained a few new slots, primarily for reporting.
      • Optimized calculation of total fragment and neutral loss matches
  • Misc
    • patRoon.cache.maxEntries package option (issue #139).
    • Updated PubChem Transformations to 0.2.3.
    • The default metabolic logic transformations are now accessible through the TPLogicTransformations() function.
    • Optimized future parallelization.
    • Disabled parallelization for reporting by default, as it can (still) result in worse performance (e.g. seen on Windows).
    • MS2Quant metadata is now also stored for sets workflow data (MS2QuantMeta slots for fGroupsScreeningSet, formulasSet and compoundsSet).
    • plotUpSet() The nsets argument now defaults to NULL, meaning it will be set to the number of compared items.

Fixes

  • Features
    • Fixed: syncing of xdata slots imported XCMS features that are not sorted by analysis (eg after peak filling) (issue #141).
    • normInts() changes (feedback Jan Specker)
      • Fixed: groupNorm=TRUE now correctly handles zero values which would result in NaN values.
      • Changed: norm_conc is set to one if all are values are NA (default for newProject()).
    • Fixed: workaround for bug in plotting feature optimization results (issue #145).
    • Fixed: featureGroups annotations slot is properly updated if all features are removed from a set.
    • Fixed: as.data.table() method for features: properly handle empty objects.
    • Fixed: KPIC2 grouping would error when there are no results.
    • Fixed: KPIC2/XCMS/XCMS3 grouping would error when there are no results.
    • Fixed: getXCMSnExp() with loadRawData=FALSE now sets additional metadata so that e.g. analysis subsetting works.
    • Fixed: feature filter didn’t hash qualityRange.
    • Fixed: normInts() produced invalid output if no internal standards were found.
    • Fixed: Sets worklfows: RF_SMILES is now stored correctly per set.
    • Fixed: suspect list verification only checked part of the columns.
    • Fixed: plotVenn() now warns when there are too many groups, instead of erroring.
    • Fixed: Properly select all suspect column in sets workflows, which previously might yield in improper results when calculating suspect annotation data, filtering and reporting.
  • Feature annotation
    • Fixed: SIRIUS with calculateFeatures=TRUE could sometimes fail due to file name truncation.
    • Fixed: filter() method for MSPeakListsSet: properly handle case where sets argument is combined with other filters.
    • Fixed: avoid error when consensus() is called with empty objects for a set (issue #144).
    • Fixed: generateCompoundsMetFrag(): don’t try to do multiprocess parallelization for non-local databases (e.g. PubChem) to avoid connection errors.
    • annotationBy filter for MS peak lists: clarify and verify that it only works on group averaged peak lists and cannot be combined with reAverage=TRUE.
    • Fixed: plotScores(): does now take mar/mai of par() parameters into account.
  • Components
    • Fixed: reporting components from cliqueMS and suspect screened features would fail (reported by Jan Specker).
    • Fixed: feature componentization would fail if no annotations were found (reported by Jan Specker).
  • TPs
    • Fixed: The removeTPIsomers filter for transformationProductsStructure didn’t actually apply the removeDuplicates filter.
    • generateComponentsTPs(): if no TP annotations are available then total fragment/neutral loss matches are reported as NA instead of zero.
  • Misc
    • Fixed: report(): don’t show (empty) annotations column in feature data if no feature annotations are present.
    • Fixed: report(): TP graphs were generated for components with absent (parent) fGroups.
    • Fixed: report(): candidate column in CSV of quant/tox prediction tables now don’t contain image link columns.
    • Fixed: report(): the onlyUsedScorings options was ignored.
    • Fixed: report(): Report settings manual section was slightly outdated with missing settings.
    • Fixed: The removeTPIsomers filter for transformationProductsStructure didn’t actually apply the removeDuplicates filter.
    • Fixed: newProject(): don’t break long lines of text strings.
    • Fixed: newProject(): XCMS3/KPIC2 groupFeatures tried to do RT alignment in sets worfklow.
    • Fixed: Sets names are now checked to not contain any special characters (besides underscores). Automatic labels are now separated by underscores instead of dots.

patRoon 2.3.4

Important changes

  • Fixed: previous versions incorrectly assumed that MS2Tox returned LC50 values as natural logarithms instead of log10. (reported by Alessia Ore)

Other changes and fixes:

  • Fixed: reAverage = TRUE was not handled correctly for the delete() method for MSPeakListsSet
  • Speed up ‘unsetting’ of large formulas/compounds objects, which affects eg plotting TP similarities, annotatedPeakList() etc (reported by Alessia Ore)
  • report(): correctly handle removed suspect hits while reporting TP similarities (reported by Alessia Ore)
  • Added visual clue to checkFeatures() and checkComponents() to see whether a feature or component is marked to be removed (pull request #117 as suggested by Leon Saal)
  • loadMSLibrary(): don’t always set Ion_mode of records to positive and guess missing Ion_mode data (issue #119)
  • Fixed: generateFormulasGenForm() topMost was not considered for cached results
  • Small doc updates for MSPeakLists
  • Fixed: the score filters for the filter() method of formulas/compounds could in rare cases not be applied correctly for consensus and/or sets results
  • MSP MS libraries: don’t be case sensitive when checking the “Num Peaks” field (Bas van de Velde)
  • Small fixes in example code from the Handbook (reported by Boris Droz and Jan Specker)
  • Changed cache for filtering of feature groups to reduce cache size (#98)
  • Updated PubChem Transformations to 0.2.0
  • Replaced the featuresSets method for groupFeatures with (updated) featuresSets algorithm specific groupFeatures methods (groupFeaturesOpenMS etc), so that the latter can be called directly in sets workflows (#128)
  • Fixed regression: groupQualityRange filter would always remove all feature groups (reported by Geert Franken)

patRoon 2.3.3

Import changes

Minor changes

Fixes

  • Fixed: The removeTPIsomers filter for transformationProductsStructure didn’t actually apply the removeDuplicates filter.
  • Fixed: if updateScores=TRUE for the methods addFormulaScoring(), predictRespFactors() and predictTox() then NaN scores could be introduced if the formulaScore is zero.
  • Fixed: if updateScores=TRUE for the method addFormulaScoring() then the score would be updated twice.
  • Fixed: generateTPsLibrary()/generateTPsLibraryFormula() specifying >1 generation did not yield in additional TP searches if parents!=NULL
  • Fixed: report() would error with components generated by CAMERA (reported by Nienke Meekel)
  • Fixed: handle OpenBabel crashes on malformed input (reported by Nienke Meekel)

patRoon 2.3.2

  • Fixed: as.data.table() method for featureGroups: normConcToTox argument was ignored (not for featureGroupsScreening)
  • Various reporting fixes for screening results from suspect lists without formula/SMILES data.
  • generateAnalysisInfo(): try to equalize the output and input directory order
  • Added get and plot methods for total ion chromatograms (TICs) and base peak chromatograms (BPCs) for the analysisInfo data.frame, features and featureGroups class as: getTICs(), getBPCs(), plotTICs() and plotBPCs().
  • Updated PubChem transformations to 0.1.9
  • Fix: reporting compounds consensus with library algorithm would fail (issue #110)
  • SIRIUS formula annotations: calculate ion_formula_mz column instead of taking it from SIRIUS data, as it rarely may not be available (issue #111)
  • Fix: improve handling SIRIUS results for features with non-standard adduct assignments (issue #111)
  • Store individual plots in MS2QuantMeta slots
  • force rJava GC to improve stability with MS2Quant predictions

patRoon 2.3.1

When updating to this release, it is important to remove any cached data, i.e. by running clearCache("all") or manually removing the cache.sqlite file from your project directory.

  • Fixed: predictRespFactors() could incorrectly cache results and perform concentration conversions twice for calibrants under some circumstances (reported by Drew Szabo)
  • Fixed: verifyDependencies() could throw errors when external dependencies were not found.
  • Fixed: predictRespFactors()/predictTox(): better handle objects without results
  • Metadata of MS2Quant is now stored in the MS2QuantMeta slots (suggested by Drew Szabo)
  • Fixed: calculateTox()/calculateConcs(): only consider relevant feature annotations
  • Fixed: calculateConcs(): avoid warnings when there are no feature groups
  • as.data.table() methods for featureGroups/featureGroupsScreening:
    • Fixed: avoid errors with features==TRUE and collapseSuspects=NULL
    • Fixed: several fixes when merging predicted concentrations/toxicities if features==TRUE and/or collapseSuspects=NULL
    • Fixed: replicate_groups with incorrect data was included with features==TRUE and average==TRUE
    • Fixed: predicted concentrations are now properly averaged with features==TRUE and average==TRUE
  • Updated PubChem transformations to 0.1.8
  • Improved documentation for collapseSuspects argument for as.data.table() method for suspect screening results

patRoon 2.3

This release adds significant new functionality, several important changes and several bug fixes thanks to user feedback.

Users of previous patRoon versions should inform themselves with the important changes highlighted in the next section. Furthermore, it is important to remove any cached data, i.e. by running clearCache("all") or manually removing the cache.sqlite file from your project directory.

Important new functionality and changes

Revamped installation of patRoon and its dependencies

This release adds new way to install and update patRoon and its dependencies. The most important changes are

  • Introduction of patRoon bundles: these are standalone installations of R, patRoon and its R package dependencies, and all other external dependencies such as MetFrag, OpenJDK, OpenMS etc. This is primarily useful for users not familiar to R or wanting to quickly try a new patRoon release.
  • Semi automated installations with the patRoonInst auxiliary package. This package will install patRoon and its dependencies automatically, which prevents the need to manually install packages from different sources (BioConductor, GitHub etc). Furthermore, this package also installs patRoonExt, another axuliary package that bundles most external dependencies (e.g. MetFrag, PubChemLite, OpenMS).
  • The old patRoon_install script is now replaced by the above installation methids, and is therefore considered deprecated, will be removed in the future and should therefore not be used anymore.

For more information, please read the updated installation chapter in the handbook, and see the project pages of patRoonInst and patRoonExt.

IMPORTANT If you installed patRoon via the legacy installation script, please read the installation chapter to disable/remove this installation prior to updating patRoon!

Prediction of feature toxicities and concentrations (MS2Tox/MS2Quant integration)

The second milestone of this release is the integration of the MS2Tox and MS2Quant R packages, which support machine learning approaches to predict the toxicity and concentration of features. The integration adds the following functionality to patRoon:

  • Automated prediction of toxicity (fish LC50) and response factors/concentrations for features from SIRIUS+CSI:FingerID fingerprints or SMILES.
  • The predictions can be made from suspect data and formula/compound annotation candidates. These can be combined and aggregated when calculating toxicities/concentrations for features.
  • The as.data.table() function and reporting interface were updated to inspect the predicted toxicities/concentrations.
  • Various new filters were added to prioritize data on calculated toxicities, response factors and concentrations.
  • Various small usability improvements to simplify calibrations and concentration units.

Please see the relevant section in the handbook, and the project pages of MS2Tox and MS2Quant for more details.

Minor new functionality and changes

  • loadMSLibrary(): improve compatibility with more .msp flavors (issue #72).
  • newProject(): save/load parameters to reproduce subsequent project creations (issue #61)
  • groupFeaturesOpenMS(): now supports handling large numbers of analyses on Windows (reported by Geert Franken, fix thanks to https://github.com/OpenMS/OpenMS/issues/6845).
  • Add package option patRoon.checkCentroided to control whether analyses files are verified to be centroided (suggested by Geert Franken).
  • Updated PubChem transformations to v0.1.7
  • Reference documentation: all generics are now documented, mainly to ensure that default arguments are listed again in the function documentation (reported by Geert Franken)
  • overlap(): the which param can now also be a list to compare groups of replicate groups (similar to plotVenn())
  • consensus() method for featureGroups: new verifyAnaInfo flag to optionally skip if the analysis information are equal for all compared objects. This is mainly useful when the data is the same but in different formats.
  • Compatibility with OpenMS 3.0
  • Windows CI is now performed on GitHub actions instead of AppVeyor.
  • genReportSettingsFile(): baseFrom argument to update old report settings files.
  • generateFormulasSIRIUS(): new getFingerprints and token arguments to download CSI:FingerID fingerprints for formula candidates. This was primarily implemented to support calculating toxicities/concentrations from formula annotations.

Fixes

  • Fixed: in rare cases EICs were incorrectly loaded from cache
  • Fixed: report() now correctly handles SIRIUS compounds results and suspects without SMILES
  • Fixed: report layout is now compatible with bslib 0.5.0 (reported by Alessia Ore)
  • Fixed: annotatedBy filter for MSPeakLists could remove precursor peaks in MS/MS data regardless if retainPrecursorMSMS=TRUE
  • Fixed: report(): The suspect(s) column for compound annotation results was always empty
  • Fixed: the reAverage argument was ignored by the filter() method of MSPeakLists when checking if cached data is available (issue #87)
  • Fixed: if reAverage=TRUE to the filter() method of MSPeakLists then the peak IDs were not regenerated (issue #87)
  • Fixed: Suspect screening results were incorrectly merged with >2 sets (issue #90)
  • Fixed: plotSpectrum() method for formulas/compounds didn’t expand plot height for formula annotations with only one mass peak
  • Fixed: annotatedPeakList()/plotSpectrum() methods for compounds didn’t label mass peaks with compounds algorithm if formulas were provided but no formula candidate was present.
  • Fixed: in some specific conditions the plot() would throw an error (“cannot coerce type ‘S4’ to vector of type ‘double’”)
  • Fixed: generateFormulas() generic definition had wrong argument order
  • Fixed: getSIRIUSToken() resulted in errors if the password input was cancelled.

patRoon 2.2

This release adds significant new functionality, several important changes and many bug fixes thanks to user feedback.

Users of previous patRoon versions should inform themselves with the important changes highlighted in the next section. Furthermore, it is important to remove any cached data, i.e. by running clearCache("all") or manually removing the cache.sqlite file from your project directory.

Important new functionality and changes

New reporting interface

The most significant change in this release is the addition of redesigned reporting functionality. Some key functionality and changes:

  • The html interface was completely redesigned to provide a modern, responsive and easier to use interface, which is powered by the bslib and reactable R packages.
  • The browsing and exploring of reported data is made significantly easier by centralizing all workflow data (features, annotations, TPs etc). Furthermore, tabular data can be easily filtered and can be grouped by properties such as suspects, parents, replicate groups etc.
  • All plots are now stored as SVG vector graphics, which are generally smaller in size, faster to create and can be zoomed in without loss of quality.
  • The generation of plots and other reporting data was optimized, and can be further speed up by parallelization.
  • Caching of report data was significantly optimized, which makes re-generating reports with partially changed data/parameters much faster.
  • Due to these optimizations, the default number of annotation candidates was increased from 5 to 25.
  • Chromatograms can now also be plotted for individual features. In addition, it is also possible to add plots of EICs in analyses where no feature was found, which makes it easy to spot any features that were not detected (or filtered out).
  • The configuration of reporting parameters was simplified and is now achieved through a YAML file.
  • And many other improvements…

An example can be seen from the report output of the tutorial.

The new reporting interface is used with the new report() method function. All the documentation was updated to reflect these changes. The now ‘legacy’ report interface (reportCSV(), reportPDF() and reportHTML()) is still available for backwards compatibility and may still be of interest as the new interface currently only supports the HTML format.

The new reporting functionality obviously did not yet underwent years of usage and feedback. Hence, please report any bugs and suggestions you may have!

Usage of SIRIUS version 5

Active logins are now necessary to use webservices such as CSI:FingerID, see e.g. https://boecker-lab.github.io/docs.sirius.github.io/account-and-license/ This release of patRoon adds support to make logging in more easy and adds several compatibility fixes for the latest SIRIUS version. The new utility function getSIRIUSToken() can be used to obtain a necessary login token. The new token argument for generateCompoundsSIRIUS() can be used to automatically log in. The newPorject() function was extended to use this new functionality.

Docker images moved

The Docker images are now served by the GitLab server of the University of Amsterdam. To pull the latest images you can run the following command:

docker pull uva-hva.gitlab.host:4567/r.helmus/patroon/patroonrs

The changes are reflected in the installation section of the handbook.

TP formula libraries

A new algorithm for generateTPs() was added: library_formula. This algorithm is similar to the library algorithm, but only works with chemical formulae. This is especially useful if only formula data is available for parents and/or TPs. The genFormulaTPLibrary() utility function can be used to automatically generate a formula library from given transformation rules. More information can be found in the updated handbook and reference manual (?generateTPsLibraryFormula).

Other

Other important changes include:

  • Features
    • Common parameters for creation of extracted ion chromatograms (EICs), such as topMost and onlyPresent, are now combined in a parameter list. The parameter list is specified with the new EICParams argument to functions such as plotChroms() and report(). A list with default parameter values is generated with the getDefEICParams() function. More information can be found in the reference manual: ?EICParams.
    • The order of some of the arguments to the plotChroms() method for featureGroups was changed.
    • makeSet() method for featureGroups (and related functions adducts() and selectIons()): the original set specific feature groups are now combined to create the final feature groups, instead of grouping features from all sets at once. This prevents rare cases where features with different adduct assignments in the same set would be grouped together (i.e. if their neutral mass would be the same). Note that this change probably will produce slightly different results. This change required the addition of a new slot annotationsChanged to featureGroupsSet for internal usage by the adducts()<- method.
  • Feature annotations
    • For sets workflows, scorings that are considered set specific (e.g. MS/MS match) are now not averaged anymore. Instead, these scorings are stored per set, which improves estimation of set specific ID levels. The old behaviour can be enabled by setting the new setAvgSpecificScores arguments of generateFormulas()/generateCompounds() to TRUE.
  • Chemical data from e.g. suspects and TPs can now be automatically ‘neutralized’ by addition/subtraction of protons, by setting the neutralChemProps/neutralizeTPs arguments. Whether a structure was neutralized is marked by the new molNeutralized column.
    • If neutralizeTPs is set and a neutralization of a TP results in a duplicate structure (i.e. in case the algorithm also generated the neutral form of the TP) then the neutralized TP is removed.

Minor new functionality

  • newProject(): added possibility to exclude analyses out of folder (issue #60, #63)
  • Features
    • as.data.table() for featureGroups
      • if regression=TRUE: add column with p values
      • if features=TRUE: add replicate group column
    • plotChroms(): analysis, groupName and intMax arguments
  • Feature annotation
    • A delete() method function was added to modify MS peak lists
    • GenForm: thrMS, thrMSMS, thrComb and maxCandidates arguments, which can be used to tweak calculations for features with many candidates, e.g., to limit calculation times.
  • Suspects
    • Multiple conditions for ID level estimation can now be combined with the and keyword in the YAML configuration file. This is especially useful when combined with the or keyword.
  • Componentization
    • Feature components: add adduct_abundance column
    • plotInt() method for components: index argument can now also be component name
  • TPs
    • generateTPsCTS(): support new PFAS libraries (set "pfas_environmental" or "pfas_metabolism" as the transLibrary argument).
    • generateTPsLibrary(): the matchParentsBy argument now also accepts "formula" and "name".
    • TP libraries may contain a retDir column that specifies the retention time direction of the TP compared to its parent (alternative to specifying log P values).
    • New argument matchGenerationsBy to the library (and library_formula) algorithm for generateTPs(), which controls how parents/TPs are matched when searching multiple transformation generations.
    • Added maxExpGenerations argument to generateTPsBiotransformer to avoid excessive TP hierarchy expansions.
    • generateTPsCTS(): support new PFAS libraries (set "pfas_environmental" or "pfas_metabolism" as the transLibrary argument).
    • generateComponentsTPs() the formulaDiff column now splits elemental losses and gains, similar as the plotGraph() method already did for TPs.
    • A delete() method function was added to modify transformationProducts
  • plotInt() methods: plotArgs and linesArgs to pass additional arguments to plot()/lines(). The latter replaces the dots argument.
  • plotGraph() methods
    • width and height arguments.
    • methods for transformationProductsStructure now draw structures in SVG format to improve quality
  • clearCache(): vacuum option to speed up clearing large cache files.

Minor changes

  • Features
    • as.data.table() for featureGroups with regression=TRUE: treat missing features as NA
    • plotChord() method for featureGroups: significantly optimized some old code
    • plotChroms() / EIC loading
      • refactor and minor improvements
      • The plot y limit is now determined from EIC data to improve accuracy
      • various optimizations to load (cached) EIC data
  • Feature annotations
    • plotScores()
      • split bars for sets
      • only split bars if results are present for >1 sets and/or consensus algorithms
    • plotSpectrum() for sets workflows better handles missing data from one or more sets when making a comparison, which avoids empty plots in such cases.
    • Several optimizations for annotatedPeakLists(), especially with sets workflows.
  • Suspects
    • Annotation similarities are now calculated with spectral similarity C++ code used by other functionality in patRoon, which is faster and allows more configuration options. Consequently, the specSimParams argument replaces the relMinMSMSIntensity and simMSMSMethod arguments.
    • annotateSuspects(): log if the suspect formula/compound data could not be matched with feature annotations
  • TPs
    • The format of the formulaDiff column in TP component results was changed to more easily identify elemental losses/gains.
    • The fromTPs slot was added to TP components and is TRUE if a transformationProducts object was used during componentization. This is primarily for internal use.
  • plotGraph() methods: show empty plot instead of throwing an error if results are empty
  • Loosened strictness of centroided data verification to speed it up, especially when dealing with many analyses.
  • Updated PubChem transformations to April 2023 release (0.1.5)
  • Updated MetFrag to 2.5.0
  • Validation of formula data in e.g. suspect lists is now much faster when prefCalcChemProps=FALSE

Fixes

  • reportHTML()
    • Fixed: EICOnlyPresent argument to reportHTML() is effective again
    • Fixed: reportHTML() could show plots of wrong TP results
  • newProject()
  • Features
    • Fix: OpenMS featureXML files exported for feature grouping now contain analysis file names, which prevents warnings about MS runs not being annotated.
    • Fixed: blank filter didn’t properly handle differing blank assignments per analysis
    • selectIons() does not throw an error anymore if there is no suitable adduct/isotope information in the given components, which would result in incorrect behavior with sets mode if e.g. no annotations were found for one of the sets.
    • Fixed: predictCheckFeaturesSession() marked passing peaks to be removed instead of the other way around (issue #59)
    • Fixed: selectIons() didn’t properly handle empty components objects
    • Fixed: chromPeaks() from xcms was sometimes not found (issue #68)
    • Fixed: calculatePeakQualities() would throw errors for empty feature results (reported by Louise Malm)
    • Analysis information
      • Ensure no duplicate analysis names are present.
      • Allow NA values for blanks
    • plotChroms() / EIC loading: group rectangle with topMost set didn’t consider retention times and intensities from other features
    • Fixed: The traceSNRFiltering argument could not be set for findFeaturesOpenMS()
    • plotChroms() now better supports plotting chromatograms for analysis without feature data (ie when onlyPresent=FALSE) in sets workflows. This now correctly works for cases where a feature is completely absent in a set.
  • Feature annotation
    • Fixed regression where the filter() method for MSPeakLists where precursor isolation (isolatePrec argument) also applied to MS/MS data (issue #56).
    • Fixed: filtered sets data (peak lists/annotations) could sometimes lead to errors
    • Fixed: generateCompoundsMetFrag() didn’t properly detect changes in local database files when considering cached data.
    • Fixed: in rare case MSPeakLists without any results could lead to errors.
    • Fixed: scoreTypes slot could contain scorings not actually used, e.g. if the scoreTypes argument to generateCompoundsMetFrag() contained scorings not actually present in the used database.
    • Custom MetFrag scorings specified that were not in compoundScorings() are now saved in compounds results and recognized by e.g. score normalization and plotScores()
    • Fixed: addFormulaScoring(): updateScore argument was ignored and treated always as TRUE
    • SIRUS
      • Fixed: Features with data offsets are now properly loaded.
      • Fixed: zero intensity precursor peaks returned by SIRIUS compound annotation were not removed, resulting in errors with annotateSuspects() (issue #54). Suspects
    • Fixed: warnings generated during suspect screening for very large suspect list could lead to very high memory usage and R errors.
    • remove bogus higherThanNext setting from estimated ID level 4a
    • Fixed: screenSuspects() would fail if the adduct column contains partially NA data.
    • Formulae with isotopes in e.g. suspect lists are now not normalized anymore, as this would remove the isotope designations
    • Fixed: unset() for featureGroupsScreeningSet resulted in loss of group quality scores and internal standard assignments
    • Fix: If a suspect list does not contain SMILES and formulas then InChIs were not used to calculate the missing formula data. (issue #54)
    • annotateSuspects()
      • Fixes for consensus annotation results (issue #54)
      • Sets workflows now separate log files for each set.
      • Fixed: annotation similarities didn’t properly handle results from generateCompoundsLibrary() if the library did not contain peak formula annotations.
    • Fixed: In sets workflows the fragments_formula was not split per set in the screening results.
  • TPs
    • Fixed: Set specific spectral similarities were not assigned correctly during TP componentization if a feature group occurs multiple times in the same component
  • Fix: multiprocessing with classic: don’t try to capture output when logging is disabled
  • Small fixes and improvements for verification of parameter lists
  • Fixed: convertMSFiles() if the analysisInfo argument is set and outPath is set with a length >1 then the wrong output path could be used.
  • Installation script: increase download timeout to avoid (unclear) errors when the script is downloading large file (issue #76).

patRoon 2.1

This release extends version 2.0 with new functionality, several important changes and bug fixes. The newProject() function was updated for the new functionality. Please see the updated Handbook and sections below for more information.

Users of previous patRoon versions should inform themselves with the important changes highlighted in the next section. Furthermore, it is highly recommended to remove any cached data, i.e. by running clearCache("all") or manually removing the cache.sqlite file from your project directory.

Important new functionality and changes

  • Transformation product (TP) screening
    • The generateTPs() function now supports an additional algorithm that interfaces with the Chemical Transformation Simulator (CTS). An important advantage of this algorithm is that it supports several abiotic transformation pathway libraries.
    • Functionality was added to generate interactive plots of transformation pathways using the plotGraph() generic function. Furthermore, this function can incorporate componentization results to easily display which TPs are present in the screening results.
    • A new class, transformationProductsStructure, is now used to store results for algorithms that provide structural information (biotransformer, library and cts). This better harmonizes the functionality between algorithms (e.g. with the filter() method function).
    • plotVenn(), plotUpSet() and consensus() methods are now available to compare and combine TP data.
    • TPs with equal structures but originating from different pathways are now handled differently to ease data interpretation
      • The names for these TPs are now the same (but still unique per parent).
      • These TPs are only included once in components, reports, suspect list conversion etc. to simplify data processing.
      • For this reason convertToMFDB()/generateComponentsTPs() don’t include any columns anymore that are specific to the transformation pathway.
    • Hierarchy expansion takes place for BioTransformer results to estimate full pathways. Please see the reference manual (?generateTPsBiotransformer()) for details.
  • Feature intensity normalization
    • The functionality to normalize feature intensities was significantly extended in this release of patRoon. A new method function, normInts() now supports various normalization methods, such as normalization by internal standards and the TIC. With internal standard normalization, the plotGraph() function can be used to interactively evaluate which internal standards were automatically assigned to each feature group.
    • Major changes
      • The normInts() function now handles all normalization and stores normalized intensities/areas in the feature data.
      • Functions that can use normalized data (as.data.table(), plotInt() etc) now have a new normalized argument, which should be TRUE to use normalized data. The normFunc argument to these functions was removed since it is not necessary anymore.
      • If normalized=TRUE and normInts() was not called on the feature data, a simple automatic default normalization is done. This is primarily for backwards compatibility.
    • Minor changes/additions
      • as.data.table() can now report normalized values for averaged feature data (if (features && average && normalized) == TRUE)
      • removeISTDs argument for filter() to remove feature groups that are assigned as internal standards.
      • The analysis information can contain a normalization concentration column (norm_conc) that influences normalization calculations. The generateAnalysisInfo() function can now initialize this data.
      • ISTDs and ISTDAssignments slots and their accessor methods internalStandards() and internalStandardAssignments() to store/access the internal standard assignment data.
  • MS libraries
    • This release adds support for loading and post-processing MS libraries (e.g. MSP or JSON files from MassBank) and using them for compound annotation. An important advantage is that annotation may be more reliable since experimental spectra are matched, and the process is much faster since no online database search or in-silico annotation need to be performed. The rules for identification level estimation were updated accordingly to support the new spectral match score (libMatch).
  • The functionality to automatically curate and calculate chemical properties such as SMILES, InChI and formulas for e.g. suspect lists was significantly changed. More data is now verified, and several optimizations were implemented to better handle large suspect lists or MS libraries. Note that minor changes in neutralMass values may be observed. For more details please see the reference manual (e.g. ?screenSuspects).

Other new functionality

  • Transformation products (TPs)
    • A new filter() method was defined for the transformationProducts class to filter generic properties.
    • New calcSims argument to the generateTPs functions: if TRUE then structural similarities will be calculated between parents/TPs.
    • The library algorithm now caches its results and supports multiple transformation generations (generations argument).
  • reportHTML()
    • Improved layout for TP reporting
    • Plotting of transformation hierarchies (requires setting the new TPs argument).
    • All reported feature information (chromatograms etc) are now placed inside a new menu
  • generateFormulasSIRIUS()/generateCompoundsSIRIUS(): projectPath and dryRun arguments. These are mainly for internal use.
  • getEICs() utility to obtain raw EIC data (suggested by Ricardo Cunha).

Minor changes

  • Transformation products (TPs)
    • biotransformer
      • Does not automatically calculate structural similarities anymore. This now requires setting the calcSims argument to TRUE (see above).
      • Simplified/Harmonized several column names
      • Converted ID and parent IDs to integer values. This was primarily done for consistency with other algorithms.
      • Removed several unnecessary parent_ columns (parent_SMILES, etc)
      • The steps argument was renamed to generations for consistency with other algorithms.
    • library algorithm: naming of TPs is similarly done as other algorithms. The library TP names are now stored in the name_lib column.
  • Removed the onlyLinked argument from the plotGraph() generic. This was done as the new plotGraph() methods don’t support this argument. Note that the argument was only removed from the generic, the original plotGraph() methods still support the argument.
  • generateCompoundsSIRIUS(): removed unused errorRetries argument
  • Updated used versions of PubChemLite, MetFrag and pugixml
  • Support for the MetFrag OECD PFAS compound database (https://zenodo.org/record/6385954). Should be configured like PubChemLite.
  • More consistent installation suggestion message when an R package from GitHub is found missing
  • Optimized topMost filter applied during MS peak list averaging

Fixes

  • Transformation products (TPs)
    • Fixed: convertMFDB() now always collapses duplicates, not just for biotransformer results.
    • Fixed: biotransformer: retDir is now derived from the original parent, i.e. not its direct parent
  • Fixed: reportHTML() now properly subscripts negative element counts in formulas
  • Fixed: reportHTML() improve handling of missing or split compound identifiers when generating URL links
  • Fixed: annotatedPeakList() for compounds: avoid _unset suffixes in mergedBy column from data of sets workflows
  • Fixed: newProject(): loading analysis info from CSV now works again on Windows
  • More workarounds to avoid NA exit codes on Linux systems
  • Fixed: generateCompoundsSIRIUS(): topMost argument was used where topMostFormulas was supposed to be used
  • Fixed: as.data.table() method for featureAnnotations would throw an error for empty results with OM=TRUE

patRoon 2.0.2

  • Fixed: removeBlanks feature groups filter would not handle analyses with multiple blanks.
  • Made enviPick optional dependency and added instructions to install from GitHub, as it was removed from CRAN.

patRoon 2.0.1

  • Fixed: newProject() would not add suspect annotation to the output script if the example suspect list or sets workflows were chosen.
  • Fixed: Default optimization range for KPIC2 min_width was incorrect (PR #31, thanks to @@coltonlloyd)
  • installPatRoon() improvements in determining what is already installed
  • Fixed: Group qualities/scores were not transferred to new featureGroups objects after calling screenSuspects() or unset()
  • Checking of MS file extensions (e.g. for generateAnalysisInfo()) is now case insensitive (see issue #34 and #43)
  • Fixed: newProject(): reportCSV() call in generated script included non-existing MSPeakLists argument.
  • Fixed: Suspect screening would result in an error if the adduct argument was specified (Corey Griffith)
  • Refactoring of the reference documentation pages (issue #35)
    • Workflow data generation functions and their algorithms specific counterparts (e.g. findFeatures, findFeaturesKPIC2) are now documented on separate pages.
    • The plotting functions for featureGroups are now documented in a separate page (?feature-plotting)
    • Many small textual improvements were made in the process
  • Fixed: the findFeaturesKPIC2() and importFeaturesKPIC2() now have correct casing (was lower case ‘f’)
  • Fixed: some function arguments for convertMSFiles() were not properly verified
  • More checks to verify if input mzML/mzXML data actually exists
  • Improved reference documentation for analysis-information (issue #33)
  • Handbook: detailed overview of all workflow functions and classes (issue #41, special thanks to @@hechth)
  • Fixed: annotations slot for featureGroups was not updated when removing groups with delete() (except sets workflows)
  • Fixed: better handle missing spectrum data with spectrumSimilarity()
  • Made nontarget an optional dependency and install it from GitHub with CI and in the installation docs (see issue #48)
  • Fixed: MS files were not always correctly found
  • Fixed: newProject() ignored group/blank input for sets workflows
  • Fixed: better error handling for suspect lists with only one valid column
  • Improve suspect list handling when input mz/rt columns are not numeric
  • checkFeatures(): don’t show multiple rows if a suspect was matched with multiple feature groups. This change removed the option to show specific suspect columns.
  • Fixed: checkFeatures() errored if Plot mode was ‘Top most replicates’ or ‘All’
  • Fixed: several issues with topMost plotting of EICs for sets data
  • Fixed: plotChroms(): peak area filling (showPeakAreas=TRUE) didn’t work if the peak height exceeded ylim
  • screenSuspects() with sets workflows: don’t warn about set specific suspect data if all data is NA
  • checkFeatures()/checkComponents() now cleanup unavailable selections when saving the session
  • Fixed: reportHTML(): Don’t try to report TP components if no data is available
  • formulasSet method for plotSpectrum(): don’t try to plot a comparison plot for candidates without MS/MS data
  • reportHTML() don’t try to plot a comparison plot for formula candidates without MS/MS data

patRoon 2.0

This release adds a significant amount of new functionality and changes. Please see the updated Handbook and sections below for more information.

Users of previous patRoon versions should inform themselves with the important changes highlighted in the next section. Furthermore, it is highly recommended to remove any cached data, i.e. by running clearCache("all") or manually removing the cache.sqlite file from your project directory.

Important changes

  • Features
    • XMCS(3): Renamed argument exportedData to loadRawData.
    • The mzWindow and EICMzWindow arguments were renamed to mzExpWindow / EICMzExpWindow and are now with slightly different meaning (please see the reference manual).
    • OpenMS: minFWHM/maxFWHM defaults lowered for findFeatures and feature optimization.
  • Annotation
    • ggplot2 support for several plotting functions (i.e. useGGPlot2 argument) is removed (not often used and a maintenance burden).
    • the precursor argument to the plotSpectrum(), annotatedSpectrum() and plotScores() methods for formulas now expects the neutral formula instead of the ionized formula. This change was done for consistency with compound annotations and sets workflows.
    • The way of obtaining candidate formulae from different features in the same group (i.e. feature formula consensus) was changed.
      • Fixes were applied to improve thresholding with featThreshold.
      • A second and new threshold, featThresholdAnn, only takes annotated features into account.
      • The default of featThreshold is now 0, for featThresholdAnn it is the same as the previous default for featThreshold.
      • Candidate results: renamed analysis column to analysis_from and added analyses column that lists all analyses from which the consensus was made.
      • if multiple annotations are available for a single MS/MS peak (eg due to differences between feature annotations) then only the annotation with lowest m/z deviation is kept (and a warning is emitted).
      • Scores of annotated fragments from different features are now averaged.
    • The storage classes and interface for formula and compound annotation was harmonized
      • The formulas and compounds classes now derive from the new featureAnnotations class. Most of the functionality common to formulas/compounds are defined for this class.
      • Storage of formula annotation results mostly follow the format that was already used for compound annotations.
      • The maxFormulas/maxFragFormulas argument for as.data.table() were removed, as these don’t make much sense with the new format.
      • The elements filter now applies to neutral formulae for both formula and compound annotations (fragElements still applies to the ionized fragment formula).
    • Formula annotation with Bruker now require MSPeakLists. Since all algorithms now require peak lists, generateFormulas now has a mandatory MSPeakLists argument (similar to generateCompounds).
    • Formula candidates (formula and compound annotations) are now reported in the ion_formula (ionized) and neutral_formula (neutral) columns. Similarly, the formula_mz column was renamed to ion_formula_mz.
  • Suspect screening
    • The methodology to match m/z values of suspects and features was changed. This was mainly for consistency and compatibility with sets workflows. Please see the updated section on suspect screening in the Handbook.

Major new functionality

Transformation product screening

The most important new functionality in patRoon 2.0 are transformation product (TP) screening workflows. This release adds functionality to predict TPs (with BioTransformer or metabolic logic) or search TPs in PubChem or custom databases. Furthermore, other data such as MS/MS similarity or feature classification data can be used to relate parent/TP features. Other TP screening functionality includes TP prioritization and automatic generation of TP compound database for MetFrag annotation. The workflows follow the classical design of patRoon, where flexible workflows can be executed with a combination of established algorithms and new functionality. For more information, please see the dedicated chapter about TP screening in the Handbook.

Sets workflows

Another major change in this release is the addition of sets workflows. These workflows are typically used to simultaneously process positive and negative ionization data. Advantages of sets workflows include simplification of data processing, combining positive and negative data to improve e.g. feature annotations and easily comparing features across polarities. A sets workflow requires minimal changes to a ‘classical workflow’, and most of the additional work needed to process both polarities is done automatically behind the scenes. For more information, please see the dedicated chapter about sets workflows in the Handbook.

Features

The following new feature detection/grouping algorithms were integrated: SIRIUS, KPIC2 and SAFD. Furthermore, integration with MetaClean was implemented for the calculation of peak qualities and machine learning based classification of pass/fail peaks. In addition, the peak qualities are used to calculate peak scores, which can be used for quick assessment and prioritization.

Data curation

Interactive curation of feature data with checkChromatograms() was replaced with checkFeatures(), which is much faster, is better suitable for larger datasets, customizable and has an improved user interface. Furthermore, this tool can be used for training/assessing MetaClean models. Similarly, checkComponents() is a function that allows interactive curation of component data.

The delete() generic function allows straightforward deletion of parts of the workflow data, such as features, components and annotations. Furthermore, this function makes it easy to implement customized filters.

Adduct annotation

The algorithms of OpenMS (MetaboliteAdductDecharger) and cliqueMS were integrated for additional ways to detect adducts/isotopes through componentization. Furthermore, the new selectIons() method uses these annotations to prioritize features (e.g. by only retaining those with preferable adducts). In addition, this function stores the adduct annotations for the remaining feature groups, which can then be automatically used for e.g. formula and compound annotation.

Other new functionality

  • newProject
    • Updated for new functionality such as sets and TP workflows and adduct annotation.
    • Completely re-designed code generation to improve extensibility. The generated code will have a slightly different layout and some parameter defaults were changed.
  • Features
    • as.data.table()
      • intensity normalization (normFunc argument)
      • customized averaging (averageFunc argument)
      • calculation of Fold-changes (FCParams argument)
      • report peak qualities/scores (qualities argument)
    • New plotVolcano() method function to plot fold changes.
    • topMostByRGroup/EICTopMostByRGroup arguments for plotting/reporting EIC data of only the top most intense replicate(s).
    • reportHTML() now only plots the EIC of the most intense feature of each replicate group (i.e. EICTopMostByRGroup=TRUE and EICTopMost=1).
    • XCMS3
      • loadRawData argument for feature grouping and comparison()
      • ... argument for findFeaturesXCMS3
      • preGroupParam to specify grouping parameters used prior to RT alignment (suggested by Ricardo Cunha)
    • The internal XCMS feature (group) objects are synchronized as much as possible when feature data is changed.
    • Feature groups: print feature counts with show() and filter() methods.
    • OpenMS feature finding: useFFMIntensities argument to speed up intensity loading (experimental).
    • reportHTML() now reports general feature information in a separate tab.
    • Feature groups: results argument to [ (subset) and filter() to quickly synchronize feature groups between objects (e.g. to quickly remove feature groups without annotation results).
  • Annotation
    • The methodology of plotSpectrum() to automatically calculate the space necessary for formula annotation texts and candidate structures was improved. Annotation texts are now automatically resized if there is insufficient space, and the maximum size and resolution for candidate structures can be controlled with the maxMolSize/molRes parameters.
    • filter() method for MSPeakLists: minMSMSPeaks filter to only retain MSMS peak lists with a minimum number of peaks.
    • filter() method for MSPeakLists: annotatedBy filter to only keep peaks with formula/compound annotations.
  • Suspect screening
    • The screenSuspects() method now supports the amend argument, which allows combining results of different screenSuspects() calls (see the Handbook for details).
  • Components
    • A new algorithm, specclust, which generates components based on hierarchically clustering feature groups with high MS/MS similarities.

Minor changes

  • Features
    • groupFeatures: renamed the feat argument to obj.
    • Improved performance for some feature group filters.
    • reportHTML(): EICs are shared between tabs to avoid duplicated plotting
    • The features object embedded in featureGroups objects is now synchronized, and any features not present in any group are removed accordingly. This reduces memory usage and indirectly causes reportCSV() to only report features still present.
    • plotInt(): now has xnames and showLegend arguments to adjust plotting.
  • Annotation
    • The [ (subset) and filter() methods for MSPeakLists now only re-average peak lists if the new reAverage argument is set to TRUE (default FALSE). This change was mainly done as (1) the effects are usually minor and (2) re-averaging invalidates any formula/compound annotations done prior to filtering.
    • filter() method for MSPeakLists: the withMSMS filter is now applied after all other filters.
    • MetFrag: the raw unprocessed annotation formulas are now additionally stored in the fragInfo tables (formula_MF column).
    • MetFrag: the precursor ion m/z is now taken from peak list data instead of the feature group to improve annotation.
    • MetFrag: the useSmiles parameter is now set to true as this seems to improve results sometimes.
    • as.data.table() method for formulas: if average=TRUE then all column data that cannot be reasonably averaged are excluded.
    • annotatedPeakList(): also add annotation columns for missing results (for consistency).
    • Compound MS/MS annotations do not include peak intensities anymore (already stored in MS peak lists).
    • The minMaxNormalization argument to the consensus() method for compounds was removed (unused).
    • filter() for formulas/compounds: if algorithm consensus results are filtered with scoreLimits, and a score term exists multiple times for a candidate, only one of the terms needs to fall within the specified limits for the candidate to be kept (was all).
    • plotSpectrum() for compounds: plotStruct is now defaulted to FALSE.
    • MSPeakLists data now store an unique identifier for each mass peak in the ID column. These IDs are used by e.g. formula/compound annotations, and stored in the PLID column in their fragInfo data. This replaces the PLIndex column in fragInfo data, which was only row based, and therefore invalidated in case peak lists were filtered afterwards.
  • Adducts
    • GenFormAdducts() and MetFragAdducts() now additionally return adducts in generic format and use cached data for efficiency.
    • err argument to as.character() to control if an error or NA should returned if conversion fails.
    • as.adduct() now removes any whitespace and performs stricter format checks to make conversion more robust.
    • Standardized GenForm/MetFrag element addition/subtraction data to improve consistency for conversions (eg NH4 –> H4N).
    • Conversion from/to adduct formats of OpenMS (MetaboliteAdductDecharger) and cliqueMS.
    • calculateIonFormula() and calculateNeutralFormula() now Hill sort their result
    • The embedded GenForm code was updated to the latest version.
  • Suspect screening
    • as.data.table(): Suspect screening specific columns are now prefixed with susp_.
    • The suspFormRank and suspCompRank suspect annotation data columns were renamed to formRank/suspCompRank (the previous change made prefixing unnecessary).
    • Several updates for Bruker TASQ import.
    • logPath argument for annotateSuspects() to specify the file path for log files are disable logging completely.
    • suspect names are now trimmed to 150 characters to avoid logging issues on e.g. Windows
  • Components
    • Intensity clusters now use fastcluster for hierarchical clustering.
    • Changed column rt to ret for consistency.
    • show(): show unique feature group counts.
    • filter(): allow negative rtIncrement values.
    • nontarget: replaced extraOpts argument with ....
    • nontarget: store links as character string indices instead of numeric indices.
    • RAMClustR: moved position of ionization argument to improve consistency.
    • The ‘reduced components’ mechanism, where a components object was returned without any algorithm specific data (using the componentsReduced class) when filtering/subsetting components, was removed. This system was quite unintuitive and imposed unnecessary limitations. Instead, functions that cannot work after component data is changed (e.g. those specific to intensity clustering) will throw an error if needed.
    • The objects returned from intclust components are now derived from a general componentsClust class, which is shared with specclust components. The common functionality for both algorithms is implemented for this class.
  • Misc
    • show() methods now print class inheritance tree
    • The progressr package is not used anymore, thus, it is not necessary to set up progress bars with future based multiprocessing.
    • newProject: Moved order of componentization step (now before annotation & suspect screening).
    • Plots of chromatograms, spectra etc that are without data now reflect this in the generated plot.

Fixes

  • Features
    • Blank filter: don’t subtract blanks from each other
    • Fixed: when xlim/ylim was used with plotChroms then peaks were not always correctly filled
    • retMin argument to plot() method for featureGroupsComparison wasn’t properly used/defaulted.
  • Annotations
    • Fixed: plotSpectrum() if xlim is set and this yields no data then an empty plot is shown.
    • Fixed: plotSpectrum() automatic ylim determination was incorrect if only one peak is shown.
    • Fixed: consensus from feature formulas possibly could have fragment m/zs not in group MS/MS peak lists.
    • Fixed: consensus from feature formulas possibly could have fragment m/zs that deviated from those in in group MS/MS peaklists.
    • Fixed: formula algorithm consensus wrongly ranked candidates not ubiquitously present in all algorithms.
    • Fixed: the scoreLimits filter for formulas could ignore results not obtained with MS/MS data.
    • Fixed: MetFrag was using a wrong/inconsistent cache name.
    • Fixed: as.data.table(compounds, fragments=TRUE) returned empty results for candidates without fragment annotations.
    • Fixed: topX arguments for the MSPeakLists method for filter() would re-order peak lists, thereby invaliding any annotations.
    • Fixed: conversion of adducts with multiple additions/subtractions to GenForm/MetFrag format failed.
    • Fixed: Hill ordering: H wasn’t sorted alphabetically if no C is present.
    • Several fixes were applied to improve handling of SIRIUS ‘adduct fragments’.
    • formula/compound annotation consensus ranking is now properly scaled.
    • Fixed: generateMSPeakListsDAFMF() potentially used wrong DA compound data in case features were filtered.
  • Suspect screening
    • Fixed: numericIDLevel() now properly handles NA values.
    • importFeatureGroupsBrukerTASQ(): Improved handling of absent analyses in imported results files.
    • Fixed: Automatic m/z calculation for suspects
      • Improperly handled electron masses for adducts involving element subtract (e.g. [M-H]-), resulting in ~1.5 mDa deviations
      • Adduct conversion didn’t handle multiple molecules (e.g. [2M+H]+) and multiple charges (e.g. [M+2H]2+)
  • Components
    • RAMClustR: ensure that columns are the right type if all values are NA.
    • CAMERA: correctly handle cases when minSize filter results in zero components.
    • plotGraph(): improve error handling with empty objects.
  • Misc
    • Future multiprocessing: make sure that logs are created even when an error occurs.
    • Classic multiprocessing: intermediate results are cached again.
    • Fixed: parallelization issues with cached data (thanks to https://blog.r-hub.io/2021/03/13/rsqlite-parallel/)
    • newProject(): correctly handle DIA with Bruker MS peak lists.

patRoon 1.2.1

  • Fixed: XCMS feature grouping didn’t work when the peakgroups alignment method was used (fixes issue #22)
  • Fixed: (harmless) mapply warning was shown with newProject()
  • newProject(): don’t show Remove button in analyses select screen when the script option is selected, as this will not work properly.
  • IPO: add default limits for OpenMS traceTermOutliers
  • IPO optimization fix: integer parameters are properly rounded
  • Fixed: generateFeatureOptPSet("xcms3", method="matchedFilter") would return a parameter set with step instead of binSize (issue #23)
  • Fixed: newProject() would generate an ID levels configuration file even when no suspect list was selected.
  • MCS calculation: handle NULL values that may occasionally be returned by rcdk::get.mcs
  • Fixed: intensity filter failed if previous filters lead to zero feature groups.
  • Fixed: reportHTML() annotation table was paged.
  • Fixed: Check final path lengths of log files and truncate where necessary (reported by Corey Griffith)
  • Fixed: in some cases the checking of analysisInfo validity may result in an error (reported by Tiago Sobreira)
  • Fixed: convertMSFiles() error with dirs=TRUE (reported by Tiago Sobreira)
  • Small updates/fixes for installPatRoon()
  • Fixed: screenSuspects() did not take onlyHits into account for caching
  • screenSuspects(): The original suspect name is stored in the name_orig column
  • XCMS3 feature group optimization: binSize and minFraction values were rounded while they shouldn’t (issue #27)

patRoon 1.2.0

This releases focuses on a significantly changed suspect screening interface, which brings several utilities to assist suspect annotation, prioritization, mixing of suspect and full NTA workflows and other general improvements.

IMPORTANT: The suspect screening interface has changed significantly. Please read the documentation (?screenSuspects and the handbook) for more details. If you want to quickly update your code without using any new functionality:

Change your existing code, e.g.

scr <- screenSuspects(fGroups, suspectList, ...)
fGroupsScr <- groupFeaturesScreening(fGroups, scr)

to

fGroupsScr <- screenSuspects(fGroups, suspectList, ..., onlyHits = TRUE)

Major changes

  • New suspect screening interface
    • By default, feature groups without suspect hit are not removed (unless onlyHits=TRUE). This allows straightforward mixing of suspect and full non-target workflows.
    • The feature groups are not renamed tot the suspect name anymore. If you quickly want to assess which suspects were found, use the screenInfo() or as.data.table() methods.
    • Subsetting of suspsect screening results can be done with the suspects argument to [, e.g. fGroupsScr[, suspects = "carbamazepine"]
    • A new method, annotateSuspects(), allows combining the annotation workflow data (peak lists, formulas, compounds) to perform a detailed annotation for the suspects found during the workflow. This method calculates properties such as
      • Rankings: where is the suspect formula/compound ranked in the candidates from the workflow data
      • Annotation similarity: how well does the MS/MS spectrum of the suspect matches with formula/compound annotations.
      • An estimation of identification levels to assist in quickly evaluating how well the suspect was annotated. The rules for identification levels are fully configurable.
    • A dedicated filter() method for suspect screening results, which allows you to easily prioritize data, for instance, by selecting minimum annotation ranks and similarities, identification levels and automatically choosing the best match in case multiple suspects are assigned to one feature (and vice versa).
    • A dedicated as.data.table() method and reporting functionality for suspect screening results to quickly inspect their annotation data.
    • Please refer to the updated suspect screening sections in the handbook and ?screenSuspects and ?annotateSuspects for more information.
  • Changes to suspect lists
    • Whenever possible, suspect information such as formulae, neutral masses, InChIKeys etc will be calculated for the input suspect list (obtainable afterwards with screenInfo()).
    • The suspect names will be checked to be file compatible, and automatically adjusted if necessary.
    • If MS/MS fragments are known for a suspect (formula or m/z), these can be included in the suspect list to improve suspect annotation.
    • The old suspect screening support for features objects was removed. The same and much more functionality can be obtained by the workflow for feature groups.
  • The reportCSV() function was simplified and uses as.data.table() to generate the CSV data. This should give more consistent results.
  • The individualMoNAScore MetFrag scoring is now enabled by default.

Other changes

patRoon 1.1

  • IMPORTANT: The plotEIC(), groups() and plotSpec() methods were renamed to plotChroms(), groupTable() and plotSpectrum(). This was done to avoid name clashes with XCMS and CAMERA. The old functions still work (with a warning), but please update your scripts as these will be removed in the future.
  • IMPORTANT: Major changes to the parallelization functionality
    • patRoon now supports an additional method to perform parallelization for tools such as MetFrag, SIRIUS etc. The main purpose of this method is to allow you to perform such calculations on external computer clusters. Please see the updated parallelization section in the handbook for more details.
    • The logPath and maxProcAmount arguments to functions such generateFormulas, generateCompounds etc were removed. These should now solely be configured through package options (see ?patRoon).
    • The patRoon.maxProcAmount package option was renamed to patRoon.MP.maxProcs.
  • Changes related to SIRIUS
    • IMPORTANT: Support for SIRIUS 4.5.0. Please update to this version since these changes break support for older versions.
    • Fix: SIRIUS formula calculation with calculateFeatures=TRUE would try to calculate formulas for features even if not present (eg after being removed by subsetting or filtering steps).
    • The SIRBatchSize argument to formula and compound generation functions was renamed to splitBatches and its meaning has slightly changed. Please see the reference manual (e.g. ?generateFormulas) for more details.
  • Changes related to MetFrag
    • Paths to local database files for MetFrag are now normalized, which makes handling of relative paths more reliable.
    • Changes in the specified local MetFrag database files are now inspected to improve caching.
    • Consistency: generateCompoundsMetfrag was renamed to generateCompoundsMetFrag.
  • Optimized loading of spectra and EIC data.
  • New utility functions
  • Finding features with OpenMS: potentially large temporary files are removed when possible to avoid clogging up disk space (especially relevant on some Linux systems where /tmp is small).
  • Several packages such as XCMS are not attached by default, which significantly speeds up loading patRoon (e.g. with library()).
  • The compoundViewer() function was marked as defunct, as it hasn;t been working for some time and its functionality is largely replaced by reportHTML().
  • generateComponentsNontarget(): update homolog statistics for merged series.
  • checkChromatograms(): fix error when fGroups has only one replicate group
  • convertMSFiles(): If algorithm="pwiz" and vendor centroiding is used then any extra filters are now correctly put after the peakPicking filter.
  • getXCMSnExp() is now properly exported and documented.

patRoon 1.0.4

  • Small compatibility fixes for macOS
  • Updated support for latest PubChemLite
  • The annoTypeCount score for annotated compounds with PubChemLite is now not normalized by default anymore when reporting results.
  • reportHTML() now correctly handles relative paths while opening the final report in a browser.

patRoon 1.0.3

  • componentsNT: include algorithm data returned by nontarget::homol.search in homol slot (suggested by Vittorio Albergamo)
  • several convertMSFiles() fixes (issue #14)
    • prevent error when no input files are found
    • only allow one input/output format (didn’t properly work before)
    • recognize that Waters files are directories
    • cwt option is now available for conversion with ProteoWizard
  • minor fixes for subsetting XCMS features objects
  • Fixed: generateCompoundsMetFrag(): compound names could be sometimes be interpreted as dates (reported by Corey Griffith)
  • Fixed: on very rare cases empty peaklists could be present after averaging.
  • Fixed: SIRIUS annotation didn’t use set adduct but used default instead
  • SIRIUS results are better handled if choosen adduct is not [M+H]+ or [M+H]+
  • More fixes for loading data.table objects properly from cache.
  • RStudio Docker image: see the updated installation instructions in the handbook (thanks to Thanh Wang for some last minute fixes!)

patRoon 1.0.2

  • Fixed: avoid errors when SIRIUS returns zero results (reported by Vittorio Albergamo)
  • Fixed: plotGraph() didn’t properly handle components without linked series (reported by Vittorio Albergamo)
  • Keep signal to noise data when importing/exporting XCMS data (sn column) (suggested by Ricardo Cunha)
  • Reversed argument order of exportedData/verbose to getXCMSSet() functions to avoid ambiguities
  • Automated tests for importing/exporting XCMS(3) data + small fixes for surfaced bugs
  • generateComponentsNontarget(): allow wider m/z deviation for proper linkage of different series (controlled by absMzDevLink argument).
  • Fixed: addAllDAEICs() sometimes used wrong names for EICs
  • Improved handling of empty feature groups objects when reporting
  • Fixed: reportPDF() may report formula annotated spectra of results not present in input featureGroups
  • Fixed: Loading data.table data from cache now calls data.table::setalloccol() to ensure proper behavior if data.table::set() is called on cached data.
  • Fixed: plotSpec() for compounds with useGGPlot2=TRUE would try to plot formulas for non-annotated peaks (resulting in many extra diagonal lines)
  • Fixed: some functions involved in caching plot data for HTML reports sometimes returned invalid data.
  • Fixed: EICs plotted by reportPDF() where not properly placed in a grid (as specified by EICGrid argument)
  • Small tweaks/fixes for reportHTML()
    • now displays subscripted neutral formulae
    • Fixed: x axis title of EICs in annotation tab was cut-off
    • Fixed: The rt vs mz plot in the summary page now uses minutes for retention times if retMin=TRUE
  • Updates for SIRIUS 4.4.29

patRoon 1.0.1

  • Perform neutral mass calculation for suspect screening with OpenBabel to avoid some possible Java/RCDK bugs on Linux.

patRoon 1.0

June 2020

  • Fixed: newProject() didn’t show polarity selection if only a compound identification algorithm was selected.
  • Updated external dependency versions in installer script.
  • Fixed: groupFeaturesXCMS3() didn’t properly cache results.
  • MSPeakLists: results for averaged peak lists are now the same order as the input feature groups
  • Fixed: XCMS(3) feature group import used wrong variable name internally (reported by Ricardo Cunha)

May 2020

  • IMPORTANT Major changes were made related to SIRIUS support
    • Multiple features can now be annotated at once by SIRIUS (configurable with new SIRBatchSize function argument). This dramatically improves overal calculation times (thanks to Markus Fleischauer for pointing out this possibility!).
    • generateFormulasSirius() and generateCompoundsSirius() are now properly capitalized to generateFormulasSIRIUS() and generateCompoundsSIRIUS()
    • Support for SIRIUS 4.4.
    • If all features are annotated at once then SIRIUS output is directly shown on the console.
    • The amount of cores used by SIRIUS can be specified with the cores function arguments.
    • More extra commandline options can be given to SIRIUS
  • Fixed: groupNames(), analyses() and similar methods sometimes returned NULL instead of an empty character vector for empty objects.
  • plotHeatMap() with interactive=TRUE: switch from now removed d3heatmap package to heatmaply
  • Fixed: reportHTML() didn’t split PubChem URLs when multiple identifiers were reported.
  • PWizBatchSize argument for convertMSFiles()

April 2020

  • extraOptsRT/extraOptsGroup arguments for OpenMS feature grouping to allow custom command line options.
  • importFeatureGroupsBrukerTASQ
    • now correctly takes retention times of suspects into account when creating feature groups.
    • retention times / m/z values are now averaged over grouped suspects.
  • The plot() method for featureGroups now allows drawing legends when colourBy="fGroups" and sets colourBy="none" by default, both for consistency with plotEIC().
  • All documentation is now available as PDF files on the website (https://rickhelmus.github.io/patRoon/)
  • newProject() now uses XCMS3 algorithms instead of the older XCMS interface.
  • Fixed: features in objects generated by xcms (not xcms3) could not be subset with zero analyses (which resulted in errors by e.g. unique() and reportHTML()). Reported by Corey Griffith.

March 2020

  • Fixed: Normalization of scorings for formulae/compounds potentially uses wrong data after subsetting/filtering of formulas/compounds objects
  • Suspect screening
    • Fixed: Errors/Warnings of missing data in suspect list were not shown if using cached data
    • If a value is missing in any of the columns from the suspect list used for automatic ion mass calculation (e.g. SMILES, formula, …) then data from another suitable column is tried.
    • Fixed: invalid neutral mass calculation for suspects with charged SMILES/InChIs
    • Default adduct can be specified in newProject() dialog
  • Small compatibility fix for feature finding with OpenMS 2.5 (reported by Thanh Wang)
  • RAMClustR is now supported from CRAN and no need to install github package anymore
  • pubchemlite identifiers are now URL linked in HTML reports
  • related CIDs are now reported for PubChemLite results.
  • MetFrag compound generation: removed addTrivialNames option as it never worked very well.
  • reportHTML(): only components with reported feature groups are now reported.

February 2020

  • Several small improvements and fixes for TASQ import
  • Suspect screening:
    • now also support chemical formula for automatic m/z calculation
    • more robust loading of suspect lists (e.g. skip suspects with missing/invalid data)

January 2020

  • Ignore user specified scorings for local databases such as CompTox that are not actually present in the DB. This makes it easier to use e.g. different DB versions with differing scorings.
  • Add scorings from wastewater and smoking metadata comptox MetFrag databases
  • Windows install script now install latest (March2019) CompTox
  • Updates for latest PubChemLite relaease (Jan2020)
  • Suspect screening now doesn’t require pre-calculated ion m/z values. Instead, suspect lists can contain SMILES, InChI or neutral mass values which are used for automatic ion m/z calculation. See ?screenSuspects for more details.

December 2019

  • Added missing score terms for latest CompTox MetFrag database
  • labels parameter for formulas/compounds methods of consensus()
  • Fixed: colour assignment for scores plotting of merged formulae/compound results might be incorrect (reported by Emma Schymanski)
  • Fixed: analysis table in newProject() UI only showed partial amount of rows.
  • Fixed: don’t print normalized instead of absolute design parameters when only one parameter is optimized in DoEs (fixes issue #10)

November 2019

  • IMPORTANT The addFormulaScoring() function now uses a different algorithm to calculate formula scores for compound candidates. The score is now based on the actual formula ranking in the provided formulas object, and is fixed between zero (no match) and one (best match).
  • Formula feature consensus:
    • All scorings are now averaged, including those that are not fragment specific (e.g. precursor m/z error)
    • This also improves ranking in certain specific cases
  • Vectorized plotting of MS spectra to make it potentially faster
  • Added PubChemLite support for MetFrag

October 2019

  • Fixed: convertMSFiles correctly checks if input exists
  • Specific optimizations after benchmarking results:
    • maxProcAmount (i.e. number of parallel processes) now defaults to amount of physical cores instead of total number of CPU threads.
    • Decreased batchSize to 8 for GenForm formula calculation.
  • plot() for featureGroups can now highlight unique/shared features across replicates (suggested by V Albergamo)
  • Linking of homologous series:
    • Improved info descriptions for plotGraph()
    • Series are now properly unlinked when merging (was too greedy)
    • Better algorithm to detect conflicting series
    • Fixed bug when updating removed links
  • concs option for generateAnalysisInfo() to set concentration data

September 2019

  • Labels for objects in a featureGroupsComparison can be customized (useful for e.g. plotting)
  • Caching and progress bar support for suspect screening
  • Updated/Fixed JDK installation for installation script
  • Fixed missing pipe operator import (%>%)

August 2019

  • topMost argument for GenForm formula calculation.
  • Added XCMS3 support for finding and grouping features, importing/exisiting data and parameter optimization (i.e. mostly on-par with classic XCMS support).
  • Changed compound result column name from InChi to InChI

June 2019

  • IMPORTANT Several things are renamed for clarity/consistency
    • The column to specify replicate groups for blank subtraction in the analysis information is re-named from ref to blank. Similarly, the refs argument to generateAnalysisInfo() is now called blanks.
    • reportMD() is renamed to reportHTML()
    • filter() method for formulas: minExplainedFragPeaks is now called minExplainedPeaks
    • screenTargets and its targets parameter have been renamed to screenSuspects() / suspects
  • Fixed incorrect selection after feature table (or other interactive tables) have been manually re-ordered (reported by Thanh Wang)
  • groups() and as.data.table() methods for featureGroups: optionally consider feature areas instead of peak intensities.
  • plotSilhouettes() method for compoundsCluster
  • Added rGroups argument to subset operator for featureGroups to subset by replicate groups (equivalent to rGroups argument to filter()).
  • Improved logging of output from CLI tools (e.g. OpenMS, MetFrag, SIRUS, …)

May 2019

  • Formula updates
    • GenForm formula calculation with MSMode="both" (the default): instead of repeating calculations with and without MS/MS data and combining the data, it now simply does either of the two depending on MS/MS data availability. The old behavior turned out to be redundant, hence, calculation is now a bit faster.
    • GenForm now perform precursor isolation to cleanup MS1 data prior to formula calculation. During this step any mass peaks that are unlikely part of the isotopic pattern of the feature are removed, which otherwise would penalize the isotopic scoring. The result is that isotopic scoring is dramatically improved now. This filter step is part of new filter functionality for MSPeakLists, see ?MSPeakLists and ?generateFormulas for more information.
    • When formula consensus are made from multiple features the scorings and mass errors are now averaged (instead of taking the values from the best ranked feature).
    • Improved ranking of candidates from a consensus of multiple formula objects (see ?formulas).
  • Consensus for compounds are now similarly ranked as formulas.
  • More consistent minimum abundance arguments for consensus() (absMinAbundance and relMinAbundance)
  • MetFrag: for-ident database and new statistical scores are now supported
  • as.data.table() / as.data.frame() for featureGroups now optionally reports regression information, which may be useful for quantitative purposes. This replaces the (defunct) regression() method and limited support from screenTargets().
  • plotGraph() method to visually inspect linked homologous series.

April 2019

  • Misc small tweaks and fixes for newProject() (e.g. loading of example data).
  • Improved graphical output of various common plotting functions.
  • Updated tutorial vignette and added handbook

March 2019

  • reportMD(): most time consuming plots are now cached. Hence, re-reporting should be signficiantly faster now.
  • Updates to MS data file conversion:
    • convertMSFiles() now (optionally) takes analysis information (anaInfo) for file input.
    • convertMSFiles() now supports Bruker DataAnalysis as conversion algorithm (replaces now deprecated exportDAFiles() function).
    • MSFileFormats() function to list supported input conversion formats.
    • generateAnalysisInfo() now recognizes more file formats. This is mainly useful so its output can be used with convertMSFiles().
    • convertMSFiles() now has the centroid argument to more easily perform centroiding.
  • Updates to newProject():
    • The analyses selector recognizes more data file formats. This way you can select analyses that have not been converted yet.
    • Data pre-treatment options now include more sophisticated file conversion options (e.g. using ProteoWizard). This and the new analysis selector functionality ensures that data files in all major vendor formats do not have to be converted prior to generating a script.
    • Re-organized tabs to mirror non-target workflow.
    • Suspect screening support.
    • Improved layout of output script.
  • withMSMS filter for MS peak lists.
  • Timeout for formula calculation with GenForm to avoid excessive calculation times.
  • importFeatures() generic function
  • Reporting functions renamed arguments related to compounds reporting (e.g. compoundTopMost to compoundsTopMost)

February 2019

  • Compound scorings are now normalized towards their original min/max values. This ensures that the score column of MetFrag results stays correct.
  • plotScores(): Option to only report scorings that have been used for ranking
  • as.data.table()/as.data.frame() method for compounds: optionally normalize scores.
  • reportPDF()/reportMD() now report only 5 top most candidate compounds by default (controlled by compoundsTopMost argument).
  • metadata for MS peak lists
  • plotSpec() now displays subscripted formulae
  • IMPORTANT Several major changes were made to the filter() methods for features and featureGroups. Please carefully read the updated documentation for these methods! (i.e. ?`filter,features-method` and ?`filter,featureGroups-method`).
    • Most argument have been renamed for consistency, simplicity and clarity.
    • The order when multiple filters are specified to the featureGroups method was adjusted, notably to improve reliability of blank filtration. Again, please see ?`filter,featureGroups-method`.
    • The following new filters were added:
      • mass defect range (mzDefectRange argument)
      • maximum relative standard deviation (RSD) of intensities between replicates (maxReplicateIntRSD argument)
      • minimum number of features within analyses (absMinFeatures and relMinFeatures arguments).
      • pre-intensity filters (preAbsMinIntensity and preRelMinIntensity arguments)
      • most existing filters now accept both relative and absolute values.
    • The script generation functionality of newScript() has been updated and supports more filter types.
    • The repetitions argument is not needed anymore for the new algorithm and has been removed.
    • Inf values now should be used to specify no maximum for range filters (was -1).
  • Fixed: GenForm now always uses Hill sorting.
  • annotatedPeakList() method for formulas and compounds. Also used by reportMD for improved annotation peak tables.
  • Tweaked default mzR MS peak lists settings (halved maxRtMSWidth and precursorMzWindow)
  • Fixed: Make sure that MetFrag web doesn’t try to set unsupported database
  • IMPORTANT Several changes were made to improve clarity and consensistency for arguments that specify retention/mz windows or allowable deviations.
    • Functions with changed argument names: generateComponentsNontarget, generateComponentsRAMClustR, generateCompoundsSirius, generateFormulasGenForm, generateFormulasSirius, generateMSPeakListsDA, generateMSPeakListsMzR, importFeatureGroupsBrukerPA
    • The maxRtMSWidth argument to generateMSPeakListsDA, generateMSPeakListsMzR (now maxMSRtWindow) now specifies a retention time window ( +/- retention time feature) instead of total retention width around a feature. Hence, current input values should be halved.
  • CAMERA and RAMClustR components: both now have minSize and relMinReplicates (replaces ubiquitous for CAMERA) arguments. Note that their defaults may filter out (feature groups from) components. See their documentation for more info.
  • Changed capitalisation of MetFrag CL location option from patRoon.path.metFragCL to patRoon.path.MetFragCL. The old name still works for backward compatability.
  • Documented usage of the CompTox database with MetFrag. See ?generateCompounds.
  • Default normalization of MetFrag scorings now follows MetFrag web behaviour.
  • topMostFormulas argument for SIRIUS compound generation.
  • Fixed GenForm ranking in case both MS and MS/MS formulae are present.
  • reportPDF()/reportMD() now report only 5 top most candidate formulae by default (controlled by formulasTopMost argument).
  • Added verifyDependencies() function to let the user verify if external tools can be found.
  • The meaning of the dirs argument to convertMSFiles() was slightly changed: if TRUE (the default) the input can either be paths to analyses files or to directories containing the analyses files.
  • More effective locating ProteoWizard binaries by using the Windows registry.
  • Nicer default graphics for featureGroups method for plot().
  • reportMD(): Don’t plot Chord if <3 (non-empty) replicate groups are available.
  • All filter() methods now support negation by negate argument.

January 2019

  • minSize and ubiquitous arguments for CAMERA component generation. See ?generateComponentsCamera.
  • Various tweaks for plotEIC() and plotSpec() methods
  • Various small additions to newProject()
  • reportMD(): added table with annotated fragments for compounds/formulas
  • consensus() updates
    • consensus() methods now support extracting unique data. This also replaces the unique() method that was defined for featureGroupsComparison.
    • comparison() now automatically determines object names from algorithm (consistency with consensus() method for other objects).
    • Fixed: coverage calculation for consensus formulas now correctly based on precursor overlap (was overlap of precursor+fragment).
  • plotVenn() and plotUpSet() methods to compare different compounds or formulas objects.
  • filter() method for components.
  • DataAnalysis formula generation: fixed neutral formula calculation if MSMode="msms", now needs adduct argument.
  • Neutral loss filter for compounds.
  • IMPORTANT Adduct specification is now simplified and more generic by usage of a new adduct class. This means that generateCompounds() and generateFormulas() now expect slightly differing arguments. Please see their manual pages.
  • Workaround for homologous series generation with nontarget (see https://github.com/blosloos/nontarget/issues/6)
  • Improvements to terminate background commandline processes when e.g. R is terminated.
  • clearCache() now supports removal of caches via regular expressions.
  • Added/Improved topMost and extraOpts arguments for SIRIUS formula/compound generation.
  • Annotated fragments from SIRIUS compounds now correctly contain charged molecular form.
  • filter() method for compounds now support generic scoring filtering and on elements of precursor and fragment formulae.
  • IMPORTANT Several changes were made to the MetFrag compound generation interface in order to simplify it and make it more generic. See ?generateCompounds for more details (notably the Scorings section).
  • More MS peak list updates
    • Precursor peaks are now flagged in MS peak list data and plotSpec()
    • Prune MS peak lists (not MS/MS) if no precursor could be determined (enabled by default, see pruneMissingPrecursorMS option in ?generateMSPeakLists).
    • Better retain precursor peaks after filtering steps: only intensity thresholds may remove precursors (always for MS data, optional for MS/MS with retainPrecursorMSMS function arguments, see ?MSPeakLists and ?generateMSPeakLists).
  • All major workflow classes now have algorithm() and as.data.table()/as.data.frame() methods. The latter replaces and enhances the makeTable() (formulas class) and groupTable() (featureGroups class) methods.

December 2018

  • Moved OpenMS XML writing code from R to C++: significantly reduces time required for grouping large amount of features.
  • Several updates for functionality that uses Bruker DataAnalyses
    • Improved verification and consistency for handling processed data from DataAnalysis
    • Automatic saving & closing of analyses processed with DataAnalysis. Files are now generally closed by default to limit the resources needed by DataAnalysis.
    • revertDAAnalyses() function: brings back set of Bruker analyses to their unprocessed state.
    • Minimum intensity arguments for Bruker DataAnalysis MS peak lists.
    • Slightly different doFMF behaviour for DataAnalysis feature finding.
  • Several important updates were made to fomula calculation functionality.
    • The interface has been simplified as the functionality from the formula and formulaConsensus classes are now merged: there is no need to call consensus() anymore after generateFormulas().
    • Formulae can now directly be calculated for feature groups by using group averaged MS peak lists (by setting calculateFeatures=FALSE). This can greatly speed up calulcation, especially with many analyses.
    • The new filter() and as.data.table()/as.data.frame methods bring new functionalities related to filtering, extracting data and performing several processing steps commonly performed for organic matter (OM) characterization.
    • Other updates on formulas
      • length now returns number of unique precursor formulas (was total number of results)
      • Fixed: Reported fragment formulas from SIRIUS were incorrectly assumed to be charged. Charged fragment formulas are now calculated manually (the neutral form is stored in the frag_neutral_formula column). This ensures correct comparison when a consensus is made.
      • reportCSV() now splits formulas for each feature group in separate CSV files (similar to compounds reporting).
      • Fixed: reportPDF() now actually includes formula annotations in annotated compound spectra when formulas are specified.
      • New oc argument when using GenForm: if enabled only organic formulae are accepted (i.e. with at least one carbon atom). Also incorporated a small fix for the actual GenForm binary to make this option work (https://sourceforge.net/p/genform/tickets/1/).
      • Fixed: coverage calculation of formulae across features treated formulae calculated only from MS data separately.
      • GenForm now also includes precursor annotation from MS/MS data.
  • file argument for clearCache()
  • Updates on MS peak lists
    • More consistent naming for algorithm specific MS peak list generators (i.e. generateMSPeakListsX where X is the algo).
    • Additional MS peak lists are generated by averaging the lists of features within a feature group.
    • generateCompounds() and plotting functionality now uses averaged group peak lists instead of peak list of most intense analysis.
    • plotSpec() method for MSPeakLists: plot (non-annotated) MS and MS/MS spectra.
    • Minimum intensity filter option that is applied after averaging.
    • Now uses “hclust” method for averaging by default, which now uses the fastcluster package.

November 2018

  • Default value for maxRtMSWidth argument used for peak list generation.
  • Fixed: maxRtMSWidth argument for mzR peak list generation had no effect.
  • Preliminary EPA DSSTox support (via LocalCSV).
  • Added addAllDAEICs() function.
  • Renamed mzWidth argument of addDAEIC() to mzWindow.
  • Normalization of compound scores: normalization method can now be set and specified scorings can be excluded.
  • Store/report IUPACName (as compoundName) from MetFrag PubChem data.
  • Renamed trivialName to compoundName for compound tables.
  • convertMSFiles: changed interface with more options, parallelization and ProteoWizard support.
  • Automatic optimization of parameters necessary for feature finding and grouping. Heavily based on the IPO R package. See the ‘feature-optimization’ manual page.
  • IMPORTANT getXcmsSet() is renamed to getXCMSSet()
  • verbose option for findFeatures() / groupFeatures()
  • Changed nintersects default for plotUpSet so that all intersections are plotted by default.
  • plotChord() now properly stops if nothing overlaps.
  • replicateGroupSubtract() now removes replicate groups that were subtracted.
  • Fixed: replicateGroupSubtract() now correctly takes maximum mean intensity for threshold determination when multiple rGroups are specified.
  • Fixed: Wrong compound clusters plotted in reportMD().
  • Fixed: Added timeout after restarting failed command (e.g. MetFrag CL) to prevent rare error “The requested operation cannot be performed on a file with a user-mapped section open”.

October 2018

  • OpenMS features class objects now store number of isotopes found for each feature.
  • IMPORTANT Added all relevant options of FeatureFinderMetabo as function arguments to findFeaturesOpenMS() and renamed/reordered current options for more conistent style. Please check ?findFeatures for the updated function arguments!
  • openReport option for reportMD(). If TRUE the generated report will be opened with a web browser.
  • reportPlots option for reportMD() which collapses reportFGroups, reportChord and reportFormulaSpectra and adds control to plot Venn and UpSet diagrams.
  • plotUpSet() methods to compare feature groups by UpSet plots. See e.g. http://caleydo.org/tools/upset/
  • filter() method for features.
  • EICs now loaded via faster C++ code thats uses mzR instead of XCMS
  • Moved feature intensity loading code for OpenMS features to C++. This results in much faster feature finding.

September 2018

  • Removed filterBy methods: these are now deprecated with new subset operators and groupNames()/analyses() methods. Example: fGroups <- fGroups[, groupNames(compounds)]
  • subset/extraction operators (“[”, “[[” and “$”) for features, featureGroups, MSPeakLists, formulas, formulaConsensus, compounds, compoundsCluster and components classes.
  • analyses() and groupNames() generics to get analyses and feature group names of the data within an object.
  • “[” method for featureGroups: empty feature groups now always dropped, drop argument now ignored.
  • reportMD(): The layout to show compounds, formulas and components is now done with DataTables (DT package). This change allows faster initial loading of results. Furthermore, several small tweaks were done to improve general design.
  • plotSpec() (compounds method): remove unused normalizeScores flag
  • plotSpec() (compounds method): plotting of embedded structure now optional (plotStruct argument)
  • plotSpec() (compounds method): automatic calculation of necessary extra height to plot labels/structure

Augustus 2018

  • The XML code required to load feature (group) data generated by OpenMS is now moved to a C++ interface that uses Rcpp and pugixml. This results in a significant reduction of required processing time. In addition, files are now processed in chunks, allowing even very large feature sets (>10000) without clogging up system memory.
  • Improved general numeric comparisons, resulting in e.g. improved EIC generation.
  • Tweaked OpenMS feature intensity loading: now takes intensity from data point closest to retention time instead of max intensity from datapoints in the search window. Furthermore, the search window for datapoints was reduced and made configurable.

July 2018

  • getMCS() method for compounds
  • plotStructure() method for compounds will draw MCS when mutiple indices are specified

June 2018

  • Added removeRefAnalyses argument to filter() (featureGroups method) to easily remove e.g. analyses that are used as blanks after blank subtraction.
  • Added filterBy() method which removes any feature groups from a featureGroups object of which no results are present in a specified object. Methods are defined for MSPeakLists, formulaConsenus, compounds and components. This method replaces some of the functionality of the filter() method for featureGroups (formConsensus and compounds arguments).
  • Added mz and chromatographic peak width range options to filter() method for feature groups.
  • Moved intensity clustering code (makeHCluster) to new component type (see componentsIntClust class documentation).

May 2018

  • Added compound clustering (see makeHCluster method for compounds). This is an useful tool to get an overview of all the candidate chemical structures after compound identification. The clustering will reduce data complexity. Furthermore, maximum common sucstructures (MCS) can be calculated and plotted for each cluster to get a quick impression of the different structures of candidates.
  • Added function arguments checks using checkmate. This guards all exported functions and methods from wrong user input arguments. If any problems are found (e.g. a wrong data type or range was specified) then the user is informed about this and what kind of input is to be expected.
  • Added workaround (removed latex dependencies added automatically by kableExtra package) that may cause memory leakage when reportMD() is called repeatedly.

April 2018

  • Added unit tests (using testthat) and fixed several small bugs that were revealed in the process.
  • Continuous integration (CI) with running tests on CircleCI (Linux builds) and AppVeyor (Windows builds) and testing coverage on Codecov. Docker images with patRoon and all its dependencies are automatically pushed on Docker Hub.
  • Many small bug fixes.