Gap Filling

Gap filling integrate signal in the m/z-rt area of a feature (i.e., a chromatographic peak group) for samples in which no chromatographic peak for this feature was identified and add it to the chromPeaks() matrix. Such filled-in peaks are indicated with a TRUE in column "is_filled" in the result object's chromPeakData() data frame.

The method for gap filling along with its settings can be defined with the param argument. Two different approaches are available:

param = FillChromPeaksParam(): the default of the original xcms code. Signal is integrated from the m/z and retention time range as defined in the featureDefinitions() data frame, i.e. from the "rtmin", "rtmax", "mzmin" and "mzmax". This method is not suggested as it underestimates the actual peak area and it is also not available for object being an XcmsExperiment object. See details below for more information and settings for this method.
param = ChromPeakAreaParam(): the area from which the signal for a feature is integrated is defined based on the feature's chromatographic peak areas. The m/z range is by default defined as the the lower quartile of chromatographic peaks' "mzmin" value to the upper quartile of the chromatographic peaks' "mzmax" values. The retention time range for the area is defined analogously. Alternatively, by setting mzmin = median, mzmax = median, rtmin = median and rtmax = median in ChromPeakAreaParam, the median "mzmin", "mzmax", "rtmin" and "rtmax" values from all detected chromatographic peaks of a feature would be used instead. In contrast to the FillChromPeaksParam approach this method uses (all) identified chromatographic peaks of a feature to define the area from which the signal should be integrated.

expandMz,expandMz<-: getter and setter for the expandMz slot of the object.

expandRt,expandRt<-: getter and setter for the expandRt slot of the object.

ppm,ppm<-: getter and setter for the ppm slot of the object.

Usage

fillChromPeaks(object, param, ...)

# S4 method for class 'XcmsExperiment,ChromPeakAreaParam'
fillChromPeaks(
  object,
  param,
  msLevel = 1L,
  chunkSize = 2L,
  BPPARAM = bpparam()
)

FillChromPeaksParam(
  expandMz = 0,
  expandRt = 0,
  ppm = 0,
  fixedMz = 0,
  fixedRt = 0
)

fixedRt(object)

fixedMz(object)

ChromPeakAreaParam(
  mzmin = function(z) quantile(z, probs = 0.25, names = FALSE),
  mzmax = function(z) quantile(z, probs = 0.75, names = FALSE),
  rtmin = function(z) quantile(z, probs = 0.25, names = FALSE),
  rtmax = function(z) quantile(z, probs = 0.75, names = FALSE)
)

# S4 method for class 'FillChromPeaksParam'
expandMz(object)

# S4 method for class 'FillChromPeaksParam'
expandMz(object) <- value

# S4 method for class 'FillChromPeaksParam'
expandRt(object)

# S4 method for class 'FillChromPeaksParam'
expandRt(object) <- value

# S4 method for class 'FillChromPeaksParam'
ppm(object)

# S4 method for class 'FillChromPeaksParam'
ppm(object) <- value

# S4 method for class 'XCMSnExp,FillChromPeaksParam'
fillChromPeaks(object, param, msLevel = 1L, BPPARAM = bpparam())

# S4 method for class 'XCMSnExp,ChromPeakAreaParam'
fillChromPeaks(object, param, msLevel = 1L, BPPARAM = bpparam())

# S4 method for class 'XCMSnExp,missing'
fillChromPeaks(object, param, BPPARAM = bpparam(), msLevel = 1L)

Arguments

object: XcmsExperiment or XCMSnExp object with identified and grouped chromatographic peaks.
param: ChromPeakAreaParam or FillChromPeaksParam object defining which approach should be used (see details section).
...: currently ignored.
msLevel: integer(1) defining the MS level on which peak filling should be performed (defaults to msLevel = 1L). Only peak filling on one MS level at a time is supported, to fill in peaks for MS level 1 and 2 run first using msLevel = 1 and then (on the returned result object) again with msLevel = 2.
chunkSize: For fillChromPeaks if object is an XcmsExperiment: integer(1) defining the number of files (samples) that should be loaded into memory and processed at the same time. This setting thus allows to balance between memory demand and speed (due to parallel processing). Because parallel processing can only performed on the subset of data currently loaded into memory in each iteration, the value for chunkSize should match the defined parallel setting setup. Using a parallel processing setup using 4 CPUs (separate processes) but using chunkSize = 1will not perform any parallel processing, as only the data from one sample is loaded in memory at a time. On the other hand, settingchunkSize` to the total number of samples in an experiment will load the full MS data into memory and will thus in most settings cause an out-of-memory error.
BPPARAM: Parallel processing settings.
expandMz: for FillChromPeaksParam: numeric(1) defining the value by which the mz width of peaks should be expanded. Each peak is expanded in mz direction by expandMz * their original m/z width. A value of 0 means no expansion, a value of 1 grows each peak by 1 * the m/z width of the peak resulting in peaks with twice their original size in m/z direction (expansion by half m/z width to both sides).
expandRt: for FillChromPeaksParam: numeric(1), same as expandMz but for the retention time width.
ppm: for FillChromPeaksParam: numeric(1) optionally specifying a ppm by which the m/z width of the peak region should be expanded. For peaks with an m/z width smaller than mean(c(mzmin, mzmax)) * ppm / 1e6, the mzmin will be replaced by mean(c(mzmin, mzmax)) - (mean(c(mzmin, mzmax)) * ppm / 2 / 1e6) mzmax by mean(c(mzmin, mzmax)) + (mean(c(mzmin, mzmax)) * ppm / 2 / 1e6). This is applied before eventually expanding the m/z width using the expandMz parameter.
fixedMz: for FillChromPeaksParam: numeric(1) defining a constant factor by which the m/z width of each feature is to be expanded. The m/z width is expanded on both sides by fixedMz (i.e. fixedMz is subtracted from the lower m/z and added to the upper m/z). This expansion is applied after expandMz and ppm.
fixedRt: for FillChromPeaksParam: numeric(1) defining a constant factor by which the retention time width of each factor is to be expanded. The rt width is expanded on both sides by fixedRt (i.e. fixedRt is subtracted from the lower rt and added to the upper rt). This expansion is applied after expandRt.
mzmin: function to be applied to values in the "mzmin" column of all chromatographic peaks of a feature to define the lower m/z value of the area from which signal for the feature should be integrated. Defaults to mzmin = function(z) quantile(z, probs = 0.25) hence using the 25% quantile of all values.
mzmax: function to be applied to values in the "mzmax" column of all chromatographic peaks of a feature to define the upper m/z value of the area from which signal for the feature should be integrated. Defaults to mzmax = function(z) quantile(z, probs = 0.75) hence using the 75% quantile of all values.
rtmin: function to be applied to values in the "rtmin" column of all chromatographic peaks of a feature to define the lower rt value of the area from which signal for the feature should be integrated. Defaults to rtmin = function(z) quantile(z, probs = 0.25) hence using the 25% quantile of all values.
rtmax: function to be applied to values in the "rtmax" column of all chromatographic peaks of a feature to define the upper rt value of the area from which signal for the feature should be integrated. Defaults to rtmax = function(z) quantile(z, probs = 0.75) hence using the 75% quantile of all values.
value: The value for the slot.

Value

An XcmsExperiment or XCMSnExp object with previously missing chromatographic peaks for features filled into its chromPeaks() matrix.

The FillChromPeaksParam function returns a FillChromPeaksParam object.

Details

After correspondence (i.e. grouping of chromatographic peaks across samples) there will always be features (peak groups) that do not include peaks from every sample. The fillChromPeaks method defines intensity values for such features in the missing samples by integrating the signal in the m/z-rt region of the feature. Two different approaches to define this region are available: with ChromPeakAreaParam the region is defined based on the detected chromatographic peaks of a feature, while with FillChromPeaksParam the region is defined based on the m/z and retention times of the feature (which represent the m/z and retentention times of the apex position of the associated chromatographic peaks). For the latter approach various parameters are available to increase the area from which signal is to be integrated, either by a constant value (fixedMz and fixedRt) or by a feature-relative amount (expandMz and expandRt).

Adjusted retention times will be used if available.

Based on the peak finding algorithm that was used to identify the (chromatographic) peaks, different internal functions are used to guarantee that the integrated peak signal matches as much as possible the peak signal integration used during the peak detection. For peaks identified with the matchedFilter() method, signal integration is performed on the profile matrix generated with the same settings used also during peak finding (using the same bin size for example). For direct injection data and peaks identified with the MSW algorithm signal is integrated only along the mz dimension. For all other methods the complete (raw) signal within the area is used.

Slots

expandMz,expandRt,ppm,fixedMz,fixedRt: See corresponding parameter above.
rtmin,rtmax,mzmin,mzmax: See corresponding parameter above.

Note

The reported "mzmin", "mzmax", "rtmin" and "rtmax" for the filled peaks represents the actual MS area from which the signal was integrated.

No peak is filled in if no signal was present in a file/sample in the respective mz-rt area. These samples will still show a NA in the matrix returned by the featureValues() method.

Author

Johannes Rainer

Examples


## Load a test data set with identified chromatographic peaks
library(xcms)
library(MsExperiment)
res <- loadXcmsData("faahko_sub2")

## Disable parallel processing for this example
register(SerialParam())

## Perform the correspondence. We assign all samples to the same group.
res <- groupChromPeaks(res,
    param = PeakDensityParam(sampleGroups = rep(1, length(res))))

## For how many features do we lack an integrated peak signal?
sum(is.na(featureValues(res)))
#> [1] 26

## Filling missing peak data using the peak area from identified
## chromatographic peaks.
res <- fillChromPeaks(res, param = ChromPeakAreaParam())

## How many missing values do we have after peak filling?
sum(is.na(featureValues(res)))
#> [1] 2

## Get the peaks that have been filled in:
fp <- chromPeaks(res)[chromPeakData(res)$is_filled, ]
head(fp)
#>          mz mzmin mzmax       rt    rtmin    rtmax     into intb   maxo sn
#> CP249 286.2 286.2 286.2 3252.556 3236.907 3274.857  1288521   NA  78048 NA
#> CP250 380.1 380.1 380.1 3193.087 3132.054 3216.561  2094879   NA  35928 NA
#> CP251 447.2 447.2 447.2 3858.193 3828.068 3914.140  2156784   NA  61256 NA
#> CP252 497.2 497.2 497.2 3382.447 3347.236 3438.785 10344783   NA 310912 NA
#> CP253 510.2 510.2 510.2 3750.211 3741.213 3799.898  1927180   NA  61368 NA
#> CP254 531.2 531.2 531.2 3340.193 3328.848 3383.620  7589092   NA 254464 NA
#>       sample
#> CP249      1
#> CP250      1
#> CP251      1
#> CP252      1
#> CP253      1
#> CP254      1

## Get the process history step along with the parameters used to perform
## The peak filling:
ph <- processHistory(res, type = "Missing peak filling")[[1]]
ph
#> Object of class "XProcessHistory"
#>  type: Missing peak filling 
#>  date: Wed Oct 30 14:03:21 2024 
#>  info:  
#>  fileIndex: 1,2,3 
#>  Parameter class: ChromPeakAreaParam 
#>  MS level(s) 1 

## The parameter class:
ph@param
#> Object of class:  ChromPeakAreaParam 
#>  Parameters:
#>  - rtmin: function (z) 
#> quantile(z, probs = 0.25, names = FALSE)
#> <environment: 0x5596821f57e0>
#>  - rtmax: function (z) 
#> quantile(z, probs = 0.75, names = FALSE)
#> <environment: 0x5596821f57e0>
#>  - mzmin: function (z) 
#> quantile(z, probs = 0.25, names = FALSE)
#> <environment: 0x5596821f57e0>
#>  - mzmax: function (z) 
#> quantile(z, probs = 0.75, names = FALSE)
#> <environment: 0x5596821f57e0>

## It is also possible to remove filled-in peaks:
res <- dropFilledChromPeaks(res)

sum(is.na(featureValues(res)))
#> [1] 26