Hi,
Check out the mzClustGeneric in xcms/R/mzClust.R,
this is the function doing the actual work.
This is called from group(method="mzClust") and takes a simple mz/sample matrix.
groups <- mzClustGeneric(peaks[,c("mz","sample")],
sampclass=classlabel,
mzppm=mzppm,mzabs=mzabs,
minsamp=minsamp,
minfrac=minfrac)
This would be a new method for findPeaks(), which would be quite welcome!
For direct infusion with several scans (as opposed to single scans, which MSW expects)
I could envision that it is not too unlike centWave,
but without the wavelets (because there are no gaussian peak shapes):
It'd take a (very narrow) scanrange, detect ROIs, but skip the wavelet analysis,
and then apply one of the mzCenterFun to get a more accurate m/z than
you can expect from a single scan. A question would be whether and how
to integrate the intensities (Max ? Avg ?). Does that make sense ?
Yours,
Steffen