SWATH: consider correlation across samples to discriminate between co-eluting compounds

[jorainer]

This is related to issue #375 . Basic functionality is implemented with the reconstructChromPeakSpectra. This function assigns all MS2 chromatographic peaks to an MS1 chromatographic peak if the correlation of their peak shapes is higher than a threshold. This means that also signal from co-eluting compounds can be present in the reconstructed MS2 spectrum (given that the peak shape of the ion for the co-eluting compound is similar the the peak shape of the original compound and that both are in the same isolation window.

One idea to remove such additional/foreign mass peaks from the MS2 spectrum would be to check if the intensity of the mass peak across samples correlates with the intensity of the feature across samples. I would assume this should be the case for mass peaks that are from fragments from the ion represented by the feature.

Note: the reconstructed MS2 spectrum can still contain peaks from other isotopes or adducts of the ion measured by the MS1 chromatographic peak. Not sure how we can get rid of them, or if we actually want to get rid of them...

Comments @michaelwitting ?

[michaelwitting]

I think the grouping of peaks from different adducts, isotopes etc cannot be avoided. In case of adducts, they might fall into a different SWATH pocket.
The major question is for me: When to do the reconstruction? Before or after alignment and correspondence?
I was thinking that it might be nice to first group all MS1 peaks based on the same approach (chromatogram and peak correlation) to at least have a guess of how many peaks are beloning to the same metabolite. Otherwise, we would reconstruct the MS2 spectrum for the monoisotope, the 1st isotope, the 2nd isotope.
The grouping would be completely independent of any annotation, just pure chromatogram and peak correlation and creates something like a "peak group" and then we correlate the MS2 candidates against the highest (?) peak from this group.

Just an idea...

CAMERA can jump then on the "peak group" and do its thing...