R/class_peakPantheRAnnotation.R

#' @title An S4 class to represent peakPantheR annotation results
#'
#' @description The \code{peakPantheRAnnotation} class is designed to run and
#' store peakPantheR parallel annotation results. Instances of the class are
#' created with the \code{peakPantheRAnnotation} constructor function, which
#' initialises an object of proper dimension with \code{spectraPaths} (set
#' samples to process) and \code{targetFeatTable} (set compounds to target).
#' \code{spectraPaths} is a character vector of spectra file paths.
#' \code{targetFeatTable} is a \code{\link{data.frame}} of compounds to target
#' as rows and parameters as columns: \code{cpdID} (int), \code{cpdName} (str),
#' \code{rtMin} (float in seconds), \code{rt} (float in seconds, or \emph{NA}),
#' \code{rtMax} (float in seconds), \code{mzMin} (float), \code{mz} (float or
#' \emph{NA}), \code{mzMax} (float).
#'
#' @details The \code{validObject} method ensures the conformity of an object to
#' the \code{peakPantheRAnnotation-class}. The number of compounds is based on
#' \code{cpdID()} length, and the number of samples is based on
#' \code{filepath()} length. Slot type is not checked as \code{setClass}
#' enforces it. peakTables and EICs type are checked on the first list element.
#' \code{annotationTable(object, column)} where \emph{column} is a column from
#' \emph{peakTable}, returns a data.frame of values with the samples as rows,
#' ROI as columns.
#'
#' @slot cpdID A character vector of compound IDs, of length number of compounds
#' @slot cpdName A character vector of compound names, of length number of
#' compounds
#' @slot ROI A data.frame of Regions Of Interest (ROI) with compounds as row and
#' ROI parameters as columns: \code{rtMin} (float in seconds), \code{rt} (float
#' in seconds, or \emph{NA}), \code{rtMax} (float in seconds), \code{mzMin}
#' (float), \code{mz} (float or \emph{NA}), \code{mzMax} (float).
#' @slot FIR A data.frame of Fallback Integration Regions (FIR) with compounds
#' as row and FIR parameters as columns: \code{rtMin} (float in seconds),
#' \code{rtMax} (float in seconds), \code{mzMin} (float), \code{mzMax} (float).
#' @slot uROI A data.frame of updated Regions Of Interest (uROI) with compounds
#' as row and uROI parameters as columns: \code{rtMin} (float in seconds),
#' \code{rt} (float in seconds, or \emph{NA}), \code{rtMax} (float in seconds),
#' \code{mzMin} (float), \code{mz} (float or \emph{NA}), \code{mzMax} (float).
#' @slot filepath A character vector of file paths, of length number of spectra
#' files
#' @slot cpdMetadata A data.frame of compound metadata, with compounds as row
#' and metadata as columns
#' @slot spectraMetadata A data.frame of sample metadata, with samples as row
#' and metadata as columns
#' @slot acquisitionTime A character vector of acquisition date-time (converted
#' from POSIXct) or NA
#' @slot uROIExist A logical stating if uROI have been set
#' @slot useUROI A logical stating if uROI are to be used
#' @slot useFIR A logical stating if FIR are to be used
#' @slot TIC A numeric vector of TIC or NA, of length number of spectra files
#' @slot peakTables A list of peakTable data.frame, of length number of spectra
#' files. Each peakTable data.frame has compounds as rows and peak annotation
#' results as columns.
#' @slot dataPoints A list of length number of spectra files. Each list element
#' is a \emph{ROIsDataPoint} list of \code{data.frame} of raw data points for
#' each ROI/uROI (retention time 'rt', mass 'mz' and intensity 'int' (as column)
#' of each raw data points (as row))
#' @slot peakFit A list of length number of spectra files. Each list element is
#' a \emph{curveFit} list of \code{peakPantheR_curveFit} or NA for each ROI
#' @slot isAnnotated A logical stating if the annotation has taken place
#'
#' \subsection{Details:}{
#' The \emph{peakTables} \code{data.frame} are structured as follow:
#' \tabular{ll}{
#' cpdID \tab database compound ID\cr
#' cpdName \tab compound name\cr
#' found \tab was the peak found\cr
#' rt \tab retention time of peak apex (sec)\cr
#' rtMin \tab leading edge of peak retention time (sec) determined at 0.5\% of
#' apex intensity\cr
#' rtMax \tab trailing edge of peak retention time (sec) determined at 0.5\% of
#' apex intensity\cr
#' mz \tab weighted (by intensity) mean of peak m/z across scans\cr
#' mzMin \tab m/z peak minimum (between rtMin, rtMax)\cr
#' mzMax \tab m/z peak maximum (between rtMin, rtMax)\cr
#' peakArea \tab integrated peak area\cr
#' maxIntMeasured \tab maximum peak intensity in raw data\cr
#' maxIntPredicted \tab maximum peak intensity based on curve fit\cr
#' is_filled \tab Logical indicate if the feature was integrated using FIR
#' (Fallback Integration Region)\cr
#' ppm_error \tab difference in ppm between the expected and measured m/z\cr
#' rt_dev_sec \tab difference in seconds between the expected and measured rt\cr
#' tailingFactor \tab the tailing factor is a measure of peak tailing.It is
#' defined as the distance from the front slope of the peak to the back slope
#' divided by twice the distance from the center line of the peak to the front
#' slope, with all measurements made at 5\% of the maximum peak height. The
#' tailing factor of a peak will typically be similar to the asymmetry factor
#' for the same peak, but the two values cannot be directly converted\cr
#' asymmetryFactor \tab the asymmetry factor is a measure of peak tailing. It is
#' defined as the distance from the center line of the peak to the back slope
#' divided by the distance from the center line of the peak to the front slope,
#' with all measurements made at 10\% of the maximum peak height. The asymmetry
#' factor of a peak will typically be similar to the tailing factor for the same
#' peak, but the two values cannot be directly converted\cr
#' }
#' }
#'
#' @return (peakPantheRAnnotation)
#' @examples
#' if(requireNamespace('faahKO')){
#' ## Initialise a peakPantheRAnnotation object with 3 samples and 2 targeted
#' ## compounds
#'
#' # Paths to spectra files
#' library(faahKO)
#' spectraPaths <- c(system.file('cdf/KO/ko15.CDF', package = 'faahKO'),
#'                     system.file('cdf/KO/ko16.CDF', package = 'faahKO'),
#'                     system.file('cdf/KO/ko18.CDF', package = 'faahKO'))
#'
#' # targetFeatTable
#' targetFeatTable <- data.frame(matrix(vector(),2,8,dimnames=list(c(),
#'                     c('cpdID','cpdName','rtMin','rt','rtMax','mzMin','mz',
#'                     'mzMax'))), stringsAsFactors=FALSE)
#' targetFeatTable[1,] <- c('ID-1', 'Cpd 1', 3310., 3344.888, 3390., 522.194778,
#'                         522.2, 522.205222)
#' targetFeatTable[2,] <- c('ID-2', 'Cpd 2', 3280., 3385.577, 3440., 496.195038,
#'                         496.2, 496.204962)
#' targetFeatTable[,c(3:8)] <- vapply(targetFeatTable[,c(3:8)], as.numeric,
#'                                     FUN.VALUE=numeric(2))
#'
#' annotation <- peakPantheRAnnotation(spectraPaths=spectraPaths,
#'                                     targetFeatTable=targetFeatTable)
#'
#' annotation
#' # An object of class peakPantheRAnnotation
#' #  2 compounds in 3 samples.
#' #   updated ROI do not exist (uROI)
#' #   does not use updated ROI (uROI)
#' #   does not use fallback integration regions (FIR)
#' #   is not annotated
#'
#' slotNames(annotation)
#' # [1] 'cpdID'       'cpdName'      'ROI'             'FIR'       'uROI'
#' # [6] 'filepath'    'cpdMetadata'  'spectraMetadata' 'acquisitionTime'
#' # [10] 'uROIExist'  'useUROI'      'useFIR'          'TIC'       'peakTables'
#' # [15] 'dataPoints' 'peakFit'      'isAnnotated'
#'
#' ## Slots shouldn't be accessed directly, accessors are available:
#' cpdID(annotation)
#' # [1] 'ID-1' 'ID-2'
#' cpdName(annotation)
#' # [1] 'Cpd 1' 'Cpd 2'
#' ROI(annotation)
#' #   rtMin       rt rtMax    mzMin    mz    mzMax cpdID cpdName
#' # 1  3310 3344.888  3390 522.1948 522.2 522.2052  ID-1   Cpd 1
#' # 2  3280 3385.577  3440 496.1950 496.2 496.2050  ID-2   Cpd 2
#' FIR(annotation)
#' #   rtMin rtMax mzMin mzMax cpdID cpdName
#' # 1    NA    NA    NA    NA  ID-1   Cpd 1
#' # 2    NA    NA    NA    NA  ID-2   Cpd 2
#' uROI(annotation)
#' #   rtMin rt rtMax mzMin mz mzMax cpdID cpdName
#' # 1    NA NA    NA    NA NA    NA  ID-1   Cpd 1
#' # 2    NA NA    NA    NA NA    NA  ID-2   Cpd 2
#' filepath(annotation)
#' # [1] 'C:/R/R-3.6.0/library/faahKO/cdf/KO/ko15.CDF'
#' # [2]'C:/R/R-3.6.0/library/faahKO/cdf/KO/ko16.CDF'
#' # [3] 'C:/R/R-3.6.0/library/faahKO/cdf/KO/ko18.CDF'
#' cpdMetadata(annotation)
#' # data frame with 0 columns and 2 rows
#' spectraMetadata(annotation)
#' # data frame with 0 columns and 3 rows
#' acquisitionTime(annotation)
#' # [1] NA NA NA
#' uROIExist(annotation)
#' # [1] FALSE
#' useUROI(annotation)
#' # [1] FALSE
#' useFIR(annotation)
#' # [1] FALSE
#' TIC(annotation)
#' # [1] NA NA NA
#' peakTables(annotation)
#' # [[1]]
#' # NULL
#' # [[2]]
#' # NULL
#' # [[3]]
#' # NULL
#' dataPoints(annotation)
#' # [[1]]
#' # NULL
#' # [[2]]
#' # NULL
#' # [[3]]
#' # NULL
#' peakFit(annotation)
#' # [[1]]
#' # NULL
#' # [[2]]
#' # NULL
#' # [[3]]
#' # NULL
#' isAnnotated(annotation)
#' # [1] FALSE
#' }
#'
#' @family peakPantheR
#' @family parallelAnnotation
#'
#' @aliases peakPantheRAnnotation-class
#'
#' @import methods
#'
#' @export
peakPantheRAnnotation   <- setClass("peakPantheRAnnotation",
                                slot = c(cpdID = "character",
                                        cpdName = "character",
                                        ROI = "data.frame",
                                        FIR = "data.frame",
                                        uROI = "data.frame",
                                        filepath = "character",
                                        cpdMetadata = "data.frame",
                                        spectraMetadata = "data.frame",
                                        acquisitionTime = "character",
                                        uROIExist = "logical",
                                        useUROI = "logical",
                                        useFIR = "logical",
                                        TIC = "numeric",
                                        peakTables = "list",
                                        dataPoints = "list",
                                        peakFit = "list",
                                        isAnnotated = "logical"))

Try the peakPantheR package in your browser

Any scripts or data that you put into this service are public.

peakPantheR documentation built on Nov. 8, 2020, 6:38 p.m.