findPeaks.centWave-methods {xcms}R Documentation

Feature detection for centroided LC/MS data

Description

Peak density and wavelet based feature detection for centroided LC/MS data

Arguments

object xcmsSet object
scanrange scan range to check for peaks
minEntries minimum number of peaks(centroids) within the range of dev ppm
dev see above
snthresh signal to noise ratio cutoff
noiserange neighbourhood size (number of scans) for each ROI used for noise/baseline analysis
minPeakWidth Minimum extent (number of scans) of a peak. Areas bigger than 2*minPeakWidth are checked whether they contain more than one peak.
scales scales to use for wavelet analysis, the scale range should correspond to the width range of the chromatographic peaks (number of scans)
maxGaussErr maximum root-mean-square error (RMSE) of Gauss fit, RMSE is close to zero if peak is gaussian-shaped
maxGaussOverlap maximal allowed overlapping (peak area)
minPtsAboveBaseLine minimum number of data points above baseline
rbwidth width of running median for baseline estimation
scRangeTol tolerance (in scans) if gaussian peak center is not within the initial ROI
maxDescOutlier Maximum number of outliers on the descent to find peak borders
mzdiff minimum difference in m/z for peaks with overlapping retention times, can be negative to allow overlap
rtdiff minimum difference in retention time for peaks with overlapping m/z, can be negative to allow overlap
sleep number of seconds to pause between plotting peak finding cycles
verbose.columns additional peak meta data columns are returned

Details

This algorithm is most suitable for high resolution centroid LC/TOF-MS data. In the first phase of the method areas of high peak density (characterised by having at least minEntries peaks within dev ppm in consecutive scans) in the LC/MS map are located. In the second phase these regions of interest (ROI) are further analysed. Gaussian curves are fitted to the chromatographic peaks, and areas larger than 2*minPeakWidth scans are checked whether they contain more than one peak (these may also overlap) using wavelet analysis.

Value

A matrix with columns:

mz weighted (by intensity) mean of peak m/z across scans
mzmin m/z of minimum step
mzmax m/z of maximum step
rt retention time of peak midpoint
rtmin leading edge of peak retention time
rtmax trailing edge of peak retention time
into integrated area of original (raw) peak
maxo maximum intensity of original (raw) peak
sn Signal/Noise ratio
egauss RMSE of Gaussian fit
if verbose.columns is TRUE additionally :
mu Gauss parameter $μ$
sigma Gauss parameter $σ$
h Gauss parameter h
f ROI number
scpos Peak position found by wavelet analysis
scmin Left peak limit found by wavelet analysis (scan number)
scmax Right peak limit found by wavelet analysis (scan number)
lmin leading edge of peak (scan number)
lmax trailing edge of peak (scan number)

Methods

object = "xcmsRaw"
findPeaks.centWave(object, scanrange=c(1,length(object@scantime)), minEntries=4, dev=140e-6, snthresh=10, noiserange=20, minPeakWidth=7, scales=c(5,7,9,12), maxGaussErr=0.25, maxGaussOverlap = 0.5, minPtsAboveBaseLine=4, rbwidth=minPeakWidth * 2, scRangeTol=2, maxDescOutlier=floor(minPeakWidth/2), mzdiff=-0.02, rtdiff=-round(2/3 *minPeakWidth *mean(diff(object@scantime))), sleep=0, verbose.columns = FALSE)

Author(s)

Ralf Tautenhahn, rtautenh@ipb-halle.de

See Also

findPeaks-methods xcmsRaw-class


[Package xcms version 1.8.0 Index]