locate.peaks {FTICRMS}R Documentation

Locate Peaks in a FT-ICR MS Spectrum

Description

Locates peaks in FT-ICR MS spectra assuming that the peaks are roughly parabolic on the log scale.

Usage

locate.peaks(peak.base, num.pts = 5, R2.thresh = 0.98, 
             oneside.min = 1, peak.method = "parabola", 
             thresh = -Inf)

Arguments

peak.base numeric matrix with two columns containing the masses and the pre-transformed spectrum intensities
num.pts minimum number of points needed to have a peak
R2.thresh minimum R^2 needed to have a peak
oneside.min minimum number of points needed on each side of the local maximum
peak.method how to locate peaks; currently the only options are "parabola" and "locmax"
thresh only local maxes that are larger than this will be checked to see if they are peaks

Details

If peak.method = "parabola", the algorithm works by locating local maxima in the spectrum, then seeing if any num.pts consecutive points with at least oneside.min point(s) on each side of the local maximum have a coefficient of determination (R^2) of at least R2.thresh when fitted with a quadratic. If, in addition, the coefficient of the squared term is negative, then this is declared a peak and the vertex of the corresponding parabola is located. The coordinates of the vertex give the components Center_hat and Max_hat in the return value. The Width_hat component is the negative reciprocal of the coefficient of the squared term.

If peak.method = "locmax", then the algorithm merely returns the set of local maxima larger than thresh, and the Width_hat component of the return value is NA.

Value

A data frame with columns

Center_hat estimated mass of peak
Max_hat estimated intensity of peak
Width_hat estimated width of peak

Note

An extremely large value for Width_hat most likely indicates a bad fit.

Using peak.method = "locmax" will vastly speed up the runtime, but may affect the quality of the analysis.

As noted in both papers by Barkauskas, a typical FT-ICR MS spectrum has far more peaks than can be accounted for by actual compounds. Thus, defining a good value of thresh will vastly speed up the computation without materially affecting the analysis.

Author(s)

Don Barkauskas (barkda@wald.ucdavis.edu)

References

Barkauskas, D.A. (2009) “Statistical Analysis of Matrix-Assisted Laser Desorption/Ionization Fourier Transform Ion Cyclotron Resonance Mass Spectrometry Data with Applications to Cancer Biomarker Detection”. Ph.D. dissertation, University of California at Davis.

Barkauskas, D.A. et al. (2009) “Detecting glycan cancer biomarkers in serum samples using MALDI FT-ICR mass spectrometry data”. Bioinformatics, 25:2, 251–257.

See Also

run.peaks


[Package FTICRMS version 0.7 Index]