Finding Peaks in Raw Data
Returns position, signal height and approximate width at half maximum peak height.
peaks(x, y = NULL, minPH, minPW, thr, stepF = 0.49)
x, y |
Position and height of signal. Any reasonable way of
defining the coordinates is acceptable. See function
|
minPH |
Mimimum height of peak to be reported. |
minPW |
Minimum width of peak at half maximum to be reported. |
thr |
Threshold below which the signal is not processed. |
stepF |
|
The function is especially useful for signals in which both very broad and very narrow peaks are of interest. The peaks may lie very close to each other or might even be superpositioned on top of each other, e.g. peaks on broader shoulders. The algorithm is also very useful when the resolution of the signal is poor and the noise is small.
The function is looking for peaks without any preceding baseline substraction or smoothing, which could distort the spectrum.
The selection criteria minPH and minPW and the values
for the calculated peak widths are only approximate.
dataframe consisting of
x |
Position of peak |
y |
Signal height |
w |
Approximate width at half maximum of peak |
In the function, the main selection criterium for the peaks is the height of the peaks, the second optional criterium is the width of the peaks.
Rene Locher
n <- 200
freq <- 1:n
theory <- sin(freq/n*4*pi)*cos(freq/n*3*pi)
spec <- theory + 0.1*rnorm(n)
plot(spec,type="b")
lines(theory,lwd=2)
pts <- peaks(spec, minPH=0.7)
points(pts,col="red",cex=1.2, pch=20)
## peaks after smoothing the spectrum
spec.sm <- loess.smooth(freq, spec, span=0.2,
degree = 2, evaluation = 100)
lines(spec.sm$x, spec.sm$y, col="steelblue", lwd=2)
pts <- peaks(spec.sm, minPH=0.4)
points(pts,col="green",cex=1.2,pch=20)
## Analyses of Mass Spectrum between 12000 and 100'000
## without smoothing, without baseline substraction
data(MS)
MS1 <- log10(MS[MS$mz>12000&MS$mz<1e5,])
P <- peaks(MS1, minPH=0.02, minPW=0.001)
plot(MS1, type="l", xlab="log10(mz)", ylab="log10(I)")
points(P,col="blue",cex=1.6)Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.